reservoir simulation of combined wind energy and compressed air
TRANSCRIPT
RESERVOIR SIMULATION OF COMBINED WIND ENERGY
AND COMPRESSED AIR ENERGY STORAGE
IN DIFFERENT GEOLOGIC SETTINGS
by
Jessica L. Neumiller
A thesis submitted to the Faculty and the Board of Trustees of the Colorado School of
Mines in partial fulfillment of the requirements for the degree of Master of Science
(Petroleum Engineering).
Golden, Colorado
Date ______________
Signed: ________________________Jessica L. Neumiller
Approved: _____________________ Dr. Ramona M. Graves Thesis Advisor
Golden, Colorado
Date _______________
______________________________
Dr. Craig W. Van Kirk Professor and Head,
Department of Petroleum Engineering
ii
ABSTRACT
To meet the inevitable decline in hydrocarbon resources, renewable energy
sources, such as wind energy, should be implemented. This is a promising, but
intermittent energy source. In order to account for wind’s intermittency, large scale
energy storage can exist in combination with wind turbines. One of most advantageous
forms of large scale storage is Compressed Air Energy Storage (CAES). CAES is
designed to store off-peak energy to make it available for use during peak demand
periods. Currently, CAES plants are located in caverns, which are uncommon in
occurrence. In order to make CAES wind farms a reliable energy source, other
geological structures, such as depleted hydrocarbon reservoirs, need to be considered as a
storage option.
This study has used the ECLIPSE 100© black oil simulator to model CAES in its
typical cavern setting, in a hypothetical reservoir setting, and in a potential CAES wind
farm area in Wyoming. The cavern setting is modeled after the Huntorf CAES facility in
Germany. The purpose of the model is to obtain a pressure match based off of the
injection and production schedule of the CAES operations at various permeabilities. This
was done by decreasing the production rates of the facility after confirming that the given
rates were maximums. To model CAES in a reservoir, EZGEN was used to generate an
anticlinal structure with 20% porosity and an original permeability of 100 md. Using the
same rate schedule as with the Huntorf pressure match model, it was determined that a
100 md permeability is unreasonable with these high rates. Permeability had to be
increased to 1,000 md to obtain the modified Huntorf rates. In order to model a more
realistic scenario, wind speed and resulting power data were taken from the Medicine
Bow Wind Project site, near Medicine Bow, Wyoming. Using porosity, permeability,
and injectivity information from the surrounding Greater Green River Basin, four models
iii
were constructed simulating combined wind energy and CAES with different
geographical locations and geological properties. The higher porosity and permeability
models (Moxa 1 and Baxter 1) could obtain higher injection and production rates and
therefore, higher power outputs than the lower porosity and permeability models (Moxa 2
and Baxter 2).
The study showed that CAES can be used in actual reservoir settings. All four
GGRB basin models have high enough power outputs to validate the use of CAES. The
use of ECLIPSE 100© for CAES applications is also confirmed based on the successful
model results for three different settings. Finally, the GGRB has good potential for
combined wind energy and CAES. This study provided a sound first step, but future
work in this area needs to be done. This should include a more intensive reservoir model
with a detailed reservoir characterization. Additionally, the effects of leakoff in the
reservoir, water saturation, fracture effects, and the use of multiple wells and some
horizontal wells should be explored.
iv
TABLE OF CONTENTS
ABSTRACT.......................................................................................................................iiiLIST OF FIGURES...........................................................................................................viiLIST OF TABLES............................................................................................................xivACKNOWLEDGEMENTS..............................................................................................xvi
CHAPTER 1........................................................................................................................1INTRODUCTION...............................................................................................................1
1.1 Energy from Wind.....................................................................................................21.1.1 History of Wind Energy......................................................................................41.1.2 Wind Turbine Design..........................................................................................81.1.3 U.S. Wind Farm Examples................................................................................13
1.2 Scope of Research....................................................................................................151.3 Research Objectives.................................................................................................171.4 Application from Petroleum Industry......................................................................17
CHAPTER 2......................................................................................................................20LITERATURE REVIEW..................................................................................................20
2.1 Benefits of Large-Scale Energy Storage..................................................................212.1.1 Fuel Cells..........................................................................................................232.1.2 The Flywheel.....................................................................................................252.1.3 Superconducting Magnetic Energy Storage.....................................................282.1.4 Supercapacitors................................................................................................302.1.5 Underground Thermal Energy Storage............................................................312.1.6 Pumped Hydroelectric Energy Storage............................................................342.1.7 Compressed Air Energy Storage......................................................................36
2.2 Advantages of CAES over Other Storage Technologies.........................................432.3 Fundamentals of Reservoir Simulation...................................................................48
CHAPTER 3......................................................................................................................52MODEL STUDY OF CAVERN STORAGE....................................................................52
3.1 Description of Huntorf Facility...............................................................................523.2 Cavern CAES Inputs................................................................................................553.3 Cavern CAES Sensitivities and Results..................................................................703.4 Objective Functions.................................................................................................81
v
3.5 Discussion of Cavern CAES Models.......................................................................83CHAPTER 4......................................................................................................................84VERIFICATION OF MODEL USE FOR RESERVOIR STORAGE..............................84
4.1 EZGEN grid input and Model Setup.......................................................................844.2 Reservoir CAES Sensitivities and Results..............................................................884.3 Reservoir CAES Model Comparison.......................................................................954.4 Discussion of Reservoir CAES Models.................................................................105
4.4.1 Comparison of Cavern Models and Reservoir Models...................................1054.4.2 Comparison of Reservoir Models...................................................................106
CHAPTER 5....................................................................................................................108CAES SIMULATION OF THE GREATER GREEN RIVER BASIN...........................108
5.1 Geology of the Greater Green River Basin............................................................1105.2 Model Inputs..........................................................................................................1155.3 Model Results........................................................................................................122
5.3.1 Moxa 1 Model Results....................................................................................1225.3.2 Moxa 2 Model Results....................................................................................1335.3.3 Baxter 1 Model Results...................................................................................1385.3.4 Baxter 2 Model Results...................................................................................141
5.4 Power Analysis......................................................................................................1475.4 Discussion of GGRB Models and Power Implications.........................................150
CHAPTER 6....................................................................................................................153CONCLUSIONS.............................................................................................................153
6.1 Major Results.........................................................................................................1536.2 Model Comparisons...............................................................................................1566.3 Model Conclusions................................................................................................1576.4 Recommendations for Future Work......................................................................1586.5 Final Discussion.....................................................................................................160
NOMENCLATURE........................................................................................................162REFERENCES................................................................................................................163APPENDICES..................................................................................................CD in Pocket
vi
LIST OF FIGURES
Figure 1.1 Wind Resource Map...........................................................................................3
Figure 1.2 Representative Size, Height, and Diameter of Wind Turbines..........................6
Figure 1.3 Decrease in cost of wind-generated Electricity..................................................7
Figure 1.4 Rotor Configurations of the HAWT...................................................................9
Figure 1.5 HAWT displaying major components..............................................................10
Figure 1.6 Sketch of CAES for a Wind Farm....................................................................16
Figure 1.7 Locations of natural gas storage facilities within the U.S................................19
Figure 2.1 Conceptual electricity chain currently used.....................................................20
Figure 2.2 Conceptual electricity chain with the addition of storage................................21
Figure 2.3 Benefits of energy storage along the electricity value chain............................22
Figure 2.4 Load profile of a large-scale energy storage facility........................................23
Figure 2.5 Composition of a fuel cell................................................................................24
Figure 2.6 Cross-section of a typical flywheel..................................................................25
Figure 2.7 A composite flywheel.......................................................................................27
Figure 2.8 SMES system design........................................................................................29
Figure 2.9 Principles behind a capacitor............................................................................30
vii
Figure 2.10 Conceptual design of ATES...........................................................................32
Figure 2.11 Conceptual design of a DTES system............................................................33
Figure 2.12 Operation of a pumped hydroelectric storage facility....................................35
Figure 2.13 Operations of a CAES facility........................................................................37
Figure 2.14 Aerial view of the Huntorf plant....................................................................40
Figure 2.15 Comparison of Huntorf and McIntosh facilities............................................41
Figure 2.16 Power ratings of major storage technologies.................................................43
Figure 2.17 Cost and performance of major storage technologies....................................44
Figure 2.18 Levelized Annual Cost of Bulk Storage Options...........................................45
Figure 2.19 Capital cost of major storage technologies....................................................46
Figure 2.20 Regions of the United States suitable for CAES............................................47
Figure 3.1 Components of the Huntorf facility.................................................................55
Figure 3.2 Daily Power Production and Associated Pressure Response...........................56
Figure 3.3 Actual dimensions of the Huntorf salt caverns................................................58
Figure 3.4 Cross-sectional view of the Huntorf Facility...................................................59
Figure 3.5 Plan view of the Huntorf Facility.....................................................................59
Figure 3.6 Relative permeability curves for water and air used in ECLIPSE 100© input.......................................................................................................................61
Figure 3.7 Pressures, temperatures, and air flow when emptying the caverns..................62
Figure 3.8 Comparison of actual and modeled data for initial model runs.......................70
Figure 3.9 Pressure Match with Varying Pore Volumes and 10,000 md Permeability.....72
viii
Figure 3.10 The injection rate for actual data and modeled data for different pore volumes..................................................................................................................73
Figure 3.11 The production rate for decreased pore volumes for well P-1.......................73
Figure 3.12 The production rate for decreased pore volumes for well P-2.......................74
Figure 3.13 The pressure match obtained with lower production rates and the original pore volumes..........................................................................................................75
Figure 3.14 Change in pressure between modeled and actual values for lower production rates and 10,000 md permeability.......................................................76
Figure 3.15 Percent difference of actual and modeled pressure values for lower production rates and 10,000 md permeability.......................................................77
Figure 3.16 Pressure match with changing production and injection rates and different permeabilities..........................................................................................79
Figure 3.17 Change in actual and modeled pressure for changing injection and production rates with 10,000 md permeability......................................................80
Figure 3.18 Percent difference in actual and modeled pressure for changing injection and production rates with 10,000 md permeability................................80
Figure 4.1 Reservoir structure created with EZGEN for use in ECLIPSE 100© base reservoir model......................................................................................................87
Figure 4.2 Pressure Response for model runs with varying pore volume and 100 md permeability in a reservoir setting.........................................................................89
Figure 4.3 Production rates with varying pore volumes and 100 md permeability for Well P-1 in a reservoir setting..........................................................................89
Figure 4.4 Production rates with varying pore volumes and 100 md permeability for Well P-2 in a reservoir setting...............................................................................90
Figure 4.5 Percent difference of the total production rate for 100 md permeability with varying pore volumes....................................................................................91
ix
Figure 4.6 Pressure response for the 1,000 md model with varying pore volumes in a reservoir setting...................................................................................................92
Figure 4.7 Production rates of model runs with 1,000 md permeability and varying pore volumes for Well P-1 in a reservoir setting...................................................93
Figure 4.8 Pressure response with original pore volume and 10,000 md permeability in a reservoir setting...............................................................................................94
Figure 4.9 Pressure response of the three permeabilities with the original pore volumes in a reservoir setting................................................................................95
Figure 4.10 Production rate comparison for Well P-1 for the three permeabilities with the original pore volumes in a reservoir setting............................................96
Figure 4.11 Production rate comparison for Well P-2 for the three permeabilities with the original pore volumes in a reservoir setting............................................97
Figure 4.12 Percent difference of the total production rate for 100 md and 1,000 md permeability with the original pore volume...........................................................98
Figure 4.13 Pressure response of the 100 md and 1,000 md models with two times the original pore volume in a reservoir setting......................................................99
Figure 4.14 Production rates for Well P-1 for 100 md and 1,000 md permeabilities with two times the original pore volume in a reservoir setting...........................100
Figure 4.15 Production rates for Well P-2 for 100 md and 1000 md permeabilities with two times the original pore volume in a reservoir setting...........................100
Figure 4.16 Percent difference of the total production rate for 100 md and 1,000 md permeability with two times the original pore volume........................................101
Figure 4.17 Pressure response of the 100 md and 1,000 md models with three times the original pore volume in a reservoir setting....................................................102
Figure 4.18 Production rates for Well P-1 for 100 md and 1,000 md permeabilities with three times the original pore volume in a reservoir setting.........................103
x
Figure 4.19 Production rates for Well P-2 for 100 md and 1,000 md permeabilities with three times the original pore volume in a reservoir setting.........................104
Figure 5.1 Map of the Greater Green River Basin with reserve information on existing oil and gas fields Source: (Kirschbaum and Roberts 2005)...................111
Figure 5.2 Major structures within the Greater Green River Basin.................................112
Figure 5.3 Subsurface depth of the Frontier formation within the Greater Green River Basin..........................................................................................................114
Figure 5.4 Location of Tip Top field and Baxter Basin South within the Greater Green River Basin................................................................................................116
Figure 5.5 Reservoir structure created with EZGEN for use in Moxa 1 and Moxa 2 reservoir models...................................................................................................119
Figure 5.6 Reservoir structure created with EZGEN for use in Baxter 1 and Baxter 2 reservoir models...................................................................................................120
Figure 5.7 Moxa 1 injection rates for the 100 MMscf/day injection and 400 MMscf/day production model for one day..........................................................123
Figure 5.8 Moxa 1 production rates for the 100 MMscf/day injection and 400 MMscf/day production model for one day..........................................................124
Figure 5.9 Moxa 1 injection rates for the 100 MMscf/day injection and 400 MMscf/day production model for three days......................................................125
Figure 5.10 Moxa 1 pressure response for the 100 MMscf/day injection and 400 MMscf/day production model for three days......................................................126
Figure 5.11 Zoomed in Moxa 1 pressure response for the 100 MMscf/day injection and 400 MMscf/day production model for three days.........................................126
Figure 5.12 Moxa 1 injection rates for the 100 MMscf/day injection with 1 injection period of 75 MMscf/day and 200 MMscf/day production model for three days.......................................................................................................127
Figure 5.13 Moxa 1 injection rates for the actual initial schedule of 100 MMscf/day injection with 1 injection period of 75 MMscf/day and 200 MMscf/day
xi
production for three days compared to the new schedule with 3 injection periods of 75 MMscf/day for three days..............................................................128
Figure 5.14 Moxa 1 pressure response for the 100 MMscf/day injection and 200 MMscf/day production model for three days with 2 different injection schedules..............................................................................................................129
Figure 5.15 Zoomed in Moxa 1 pressure response for the 100 MMscf/day injection and 200 MMscf/day production model for three days with 2 different injection schedules...............................................................................................129
Figure 5.16 Moxa 1 pressure response for the 50 MMscf/day injection and 200 and 100 MMscf/day production model for three days.........................................130
Figure 5.17 Zoomed in Moxa 1 pressure response for the 50 MMscf/day injection and 200 and 100 MMscf/day production model for three days...........................131
Figure 5.18 Zoomed in Moxa 1 pressure response for the 1 MMscf/day injection and 4 and 2 MMscf/day production model for three days...................................132
Figure 5.19 Moxa 2 injection rates for the 100 MMscf/day injection and 400 MMscf/day production model for one day..........................................................134
Figure 5.20 Moxa 2 production rates for the 100 MMscf/day injection and 400 MMscf/day production model for one day..........................................................135
Figure 5.21 Zoomed in Moxa 2 pressure response for the 100 MMscf/day injection and 400 MMscf/day production model for one day............................................135
Figure 5.22 Zoomed in Moxa 2 pressure response for the 20 MMscf/day injection and 80 and 40 MMscf/day production model for three days...............................136
Figure 5.23 Zoomed in Moxa 2 pressure response for the 1 MMscf/day injection and 4 and 2 MMscf/day production model for three days...................................137
Figure 5.24 Baxter 1 pressure response for the 100 MMscf/day injection and 400 and 200 MMscf/day production model for three days.........................................139
Figure 5.25 Baxter 1 pressure response for the 50 MMscf/day injection and 200 and 100 MMscf/day production model for three days.........................................140
xii
Figure 5.26 Baxter 1 pressure response for the 1 MMscf/day injection and 4 and 2 MMscf/day production model for three days......................................................140
Figure 5.27 Baxter 2 production rates for the 100 MMscf/day injection and 400 MMscf/day production model for three days......................................................142
Figure 5.28 Baxter 2 pressure response for the 100 MMscf/day injection and 400 MMscf/day production model for three days......................................................143
Figure 5.29 Baxter 2 production rates for the 25 MMscf/day injection and 100 MMscf/day production model for three days......................................................144
Figure 5.30 Baxter 2 pressure response for the 25 MMscf/day injection and 100 MMscf/day production model for three days......................................................145
Figure 5.31 Baxter 2 pressure response for the 25 MMscf/day injection and 50 MMscf/day production model for three days......................................................146
Figure 5.32 Baxter 2 pressure response for the 1 MMscf/day injection and 4 and 1 MMscf/day production model for three days......................................................147
xiii
LIST OF TABLES
Table 1.1 Installed wind capacity for world regions...........................................................4
Table 2.1 Properties of various flywheel building materials.............................................26
Table 3.1 Specifications of the Huntorf CAES facility.....................................................53
Table 3.2 Comparison of actual Huntorf data to Cavern CAES base model data.............57
Table 3.3 Gas-deviation factor values for various pressures.............................................65
Table 3.4 Gas Formation Volume Factors for various pressures......................................66
Table 3.5 Lee et al. viscosity calculations.........................................................................67
Table 3.6 Daily schedule for the Huntorf facility..............................................................68
Table 3.7 Conversion factors for oilfield units to metric units..........................................69
Table 3.8 Schedule of injection and production rates with the best pressure match.........78
Table 3.9 Objective functions for Cavern CAES base model and the described sensitivities............................................................................................................82
Table 4.1 Injection and production rates for each well used in Sensitivity C and for Reservoir CAES.....................................................................................................85
Table 4.2 Summary of maximum percent difference between actual production rates and modeled production rates for the three permeability models in a reservoir setting.................................................................................................104
xiv
Table 5.1 Predicted energy values based on average wind speeds for Historical Data (1987 – 1992) and 2004 Data and Actual Energy values collected on wind turbines.......................................................................................................109
Table 5.2 Model inputs for the Moxa and Baxter models...............................................118
Table 5.3 Moxa 1 modeled injection and production rates.............................................122
Table 5.4 Moxa 2 modeled injection and production rates.............................................133
Table 5.5 Baxter 1 modeled injection and production rates............................................138
Table 5.6 Baxter 2 modeled injection and production rates............................................141
Table 5.7 Generated power and from the various production rates of the Moxa and Baxter models...............................................................................................149
xv
Table 5. 8 Comparison of CAES daily power output to amount of natural gas and coal necessary to achieve the same power over the five-hour production period...................................................................................................................149
xvi
ACKNOWLEDGEMENTS
I would like to thank my advisor, Dr. Ramona Graves, for taking me in under not
very ideal circumstances. Thank you for understanding my situation and being willing to
jump right in and participate in my ongoing research. Your knowledge and guidance
have allowed me to produce a thesis that I can be proud of. The members of my
committee also deserve thanks. Thank you all for working with me and providing
valuable insight. Dr. Mark Miller, thank you for sticking with me through this journey
and being willing to talk with me about all facets of my research. Dr. John Curtis, thank
you for being an inspiration in the classroom and sparking my interest in the geologic
side of the petroleum industry.
I would also like to thank my remaining committee member, Dr. Dag Nummedal,
and the entire Colorado Energy Research Institute (CERI) organization. Thank you for
taking interest in my research and providing funding for the duration of my work.
Without your assistance, this project would not have been possible.
John Fanchi also deserves my appreciation for providing me with the idea for this
research and for helping me through the initial stages.
I am also grateful to Fritz Crotogino for all the insight you provided into the
operations of the Huntorf CAES facility. Without your help, I would have been lost in
the data.
Finally, a big thanks go out to my friends and family, especially Mike for
understanding when I wasn’t in the best of moods. Now I fully understand what you
were going through during the final weeks of your thesis writing.
xvii
CHAPTER 1
INTRODUCTION
The need for sustainable energy is an ever-increasing concern. The topic of peak
oil is under constant debate, leaving the energy supply for the future uncertain. The use
of fossil fuels also comes with the problem of a negative impact on the environment
during their extraction and use. In order to compensate for the inevitable decline in
hydrocarbon production and the detriment to the environment, alternative energy sources
should be considered. A promising renewable energy source is the ability to harness
wind through wind turbines and wind farms. Turbines have become highly efficient over
the years and can generate energy at a cost that is comparable with other sources.
However, wind is still an intermittent energy source. A process needs to be in place that
can help increase the efficiency of the wind itself. Excess wind needs to be exploited and
energy must still be accessible when the wind is not blowing.
Large scale energy storage systems offer solutions to accomplishing this task.
One of the most promising forms of large scale storage is a process known as
Compressed Air Energy Storage (CAES). CAES is designed to store off-peak energy to
make it available for use during peak demand periods. During the off-peak periods, a
motor can consume power to compress and store the air in subsurface structures. Then
during peak load periods, the process is reversed allowing the already compressed air to
return to the surface and drive turbines as the air is slowly heated and released. No
additional compression is necessary to drive the turbines because the enthalpy is already
included in the compressed air. Currently, CAES plants are located in caverns, either
1
mined rock caverns or solution-mined salt caverns. This type of setting is ideal for the
use of CAES, but these structures are also low in occurrence. In order to make CAES
wind farms a reliable energy source, other geological structures must be explored, such as
aquifers and reservoirs similar to those found with hydrocarbon production.
1.1 Energy from Wind
In order to understand the use of CAES in collaboration with wind farms, a brief
overview of the energy source is necessary. The potential of the wind, a history of wind
energy and the turbines associated with it, a look at the design of wind turbines and their
main components, and wind farm examples will be examined.
To create economical energy from the wind, strong frequent winds are ideal.
Figure 1.1 shows the wind potential for the U.S. The National Renewable Energy Lab
(NREL) took measurements at monitoring stations at various times throughout the year.
An annual average wind speed was then calculated and used to determine the amount of
energy that could be created based on the rotor of a wind turbine per square meter of area.
This value, expressed as watts per square meter, is used to determine a wind power class
ranging from one to seven. If a site is classified as three or higher, then this area is
suitable for wind farm development. A Class two or higher site has the potential for
running small wind generators (National Renewable Energy Laboratory 2000).
2
Figure 1.1 Wind Resource Map Source: (National Renewable Energy Laboratory 2000)
About 90 % of this available wind energy is located in the Great Plains. By the
year 2015, the U.S. wind capacity is expected to reach 12,000 MW according to the
Energy Information Administration. Worldwide, wind-generated energy is estimated to
be more than 100,000 MW by 2010. Just from 1993 to 2003, a 30% growth has been
observed in the industry. Table 1.1 shows the wind capacity in different parts of the
world and the increases in capacity that have been made (Patel 2006).
3
Table 1.1 Installed wind capacity for world regionsSource: (Patel 2006)
1.1.1 History of Wind Energy
The concept to harness the wind for practical uses has been around since 900 AD,
when the Persians constructed the first windmills. Windmills can still be seen across the
world to aid with mechanical applications, such as pumping water. The use of wind to
generate electricity first surfaced towards the beginning of the 20th century with the
production of small wind electric generators. Like many of today’s wind turbines these
small turbines, most notably the Brush turbine, had rotors with three blades and airfoil
shapes. The Jacobs turbine also surfaced towards the turn of the century and is
considered to be a direct resemblance of some of the modern small turbines. The most
significant U.S. built turbine, came in the late 1930s with the construction of the Smith-
Putnam machine. It was the largest turbine built during that time period and for a number
of years after. It featured a 53.3 m diameter and a power rating of 1.25 MW. The Smith-
4
Putnam was ahead of its time; too little was know about a turbine this size and,
consequently, it suffered a blade failure in 1945 and funding for the project was halted.
After the Smith-Putnam, U.S. interest in wind lost its appeal, but Denmark was busy
designing its own turbines and employing concepts such as aerodynamic stall and an
induction generator. These concepts helped Denmark establish its strong presence in the
wind industry during the 1980s, where it is still a dominant force today. Meanwhile,
during the 1950s, Germany’s Ulrich Hütter founded modern aerodynamic principles that
are still applied today (Manwell, McGowan and Rogers 2002).
Based on growing environmental concerns associated with fossil fuels, the idea of
wind energy re-emerged in the late 1960s. This trend didn’t really catch on in the U.S.
until the Oil Crises of the mid 1970s. The Carter administration pushed for alternative
sources of energy, including wind energy. New turbines were being designed by
government agencies and companies, such as NASA and Boeing, respectively. The
Public Utility Regulatory Policy Act of 1978 created incentives allowing wind turbines to
connect to the grid and forcing utilities to pay for the “avoided cost” associated with each
kWh generated. With these new turbines and incentives, wind energy was becoming
economically feasible (Manwell et al. 2002). In California, wind farms were beginning
to take form; however, many of the machines in these farms were still prototypes and
could not provide the desired energy. Since an investment tax credit was in place, instead
of a production tax credit, faulty machines could be used without any consequence of
economic loss. However, these tax credits were withdrawn by the Reagan administration
in the early 1980s, putting a halt on the wind rush (Gipe 1995). U.S. wind turbine
manufacturers began to go out of business and culminated with the downfall of the
largest U.S. manufacturer, Kennetech Windpower, in 1996. The Danish and German
producers designed turbines of better quality than the U.S. manufacturers, leading to the
complete move of turbine production to Europe. The size of commercial wind turbines
has increased from 50 kW to 2 MW over the past 25 years. Figure 1.2 below displays the
specifications of wind turbines over the years (Manwell et al. 2002).
5
Figure 1.2 Representative Size, Height, and Diameter of Wind TurbinesSource: (Manwell et al. 2002)
The major contributors to the acceleration of wind power use and technology according
to Patel are:
High-strength fiber composites for constructing large, low-cost blades
Falling prices of the power electronics associated with wind power
systems
Variable-speed operation of electrical generators to capture maximum
energy
Improved plant operation, pushing the availability up to 95%
Economies of scale as the turbines and plants are getting larger in size
Accumulated field experience (the learning-curve effect) improving
the capacity factor up to 40%
(Patel 2006)
Figure 1.3 demonstrates item 2 of Patel’s reasoning. Since 1980, the cost of using wind
to generate electricity has decreased significantly, making wind comparable to the cost of
other energy sources, even without tax incentives.
6
Figure 1.3 Decrease in cost of wind-generated Electricity Source: (Patel 2006)
The potential to keep wind on the increase as an alternative energy source is
apparent not only in the U.S. but also worldwide; wind resources could be even more
practical and efficient if issues such as intermittency could be made obsolete with energy
storage options.
1.1.2 Wind Turbine Design
One of the most important components of designing wind turbines is the
understanding of aerodynamics. Obviously, the wind moves the blades of a turbine, but
7
more specifically, it is the concept of lift. As wind travels along the blade, the wind on
the upper surface moves faster than the wind on the lower surface. This creates a lower
pressure on the upper surface and enables the blade to the lifted. The angle at which the
blade is placed determines whether the blade will experience lifting forces or stall. At
increasing angles, the air begins to flow in an irregular vortex or the air becomes
turbulent. This causes the low pressure area on the upper surface to disappear. When
this occurs, the blade is experiencing stall. This scenario is desired during high wind
speeds that the blade is not capable of handling. If stall is not achieved during periods of
high wind speed, then damage to system components can occur.
The wind turbine design that is used almost exclusively today is the horizontal
axis wind turbine (HAWT). The name arose from the concept that the axis of rotation is
parallel to the ground. The HAWT can be designed to have its rotors orientated upwind
or downwind (Figure 1.4).
Besides orientation, HAWT rotors are also classified by their hub design (rigid or
teetering), rotor control (pitch vs. stall), the number of blades (typically two or three), and
how the blades are aligned with the wind (free yaw or active yaw). The basic
components of a HAWT (Figure 1.5) consist of
1. The rotor and the hub used for support of the turbine
2. The drive train, which mainly consists of the gearbox, shafts, coupling,
mechanical brakes, and the generator
3. The turbine electrical switch boxed and control
4. The main frame and Yaw system, which provide a cover for the drive train
and its controls and the support for the upper machinery, respectively
5. The tower and foundation
6. A detached electrical system (grid connection) that supplies all the
components necessary to transfer electricity
8
Figure 1.4 Rotor Configurations of the HAWTSource: (Manwell et al. 2002)
9
Figure 1.5 HAWT displaying major components Source: (Patel 2006)
An overview of each major component will now be provided to supply a basic
understanding of the HAWT. Further details can be found in Manwell et al. (2002) and
Patel (2006).
1) The rotor is oftentimes considered to be the most important part of a turbine, in
terms of cost and performance. It consists of the hub and the blades of the turbine. The
blades utilize airfoils to generate the mechanical power. The width and length of the
blades are determined by the preferred aerodynamic performance, the maximum desired
rotor power, the airfoil properties, and the strength concerns. Most of the models in
10
production today utilize the three-blade design; however, some manufacturers can offer
two-blade models. Single-blade units were once produced, but are no longer in
production. In addition to the three blades, most models operate with the rotors in the
upwind position; however, models can be found that orientate their rotors downwind
(Manwell et al. 2002). Rotors are generally placed upwind because of the turbulence
created by the air current behind the tower. Turbines are typically designed to operate at
their maximum output with wind speeds of 15 m/s (33 mph). They are generally not
designed to operate at higher speeds because of the lack of reliable wind. Therefore it is
uneconomical to design turbines for increased maximum outputs. When the wind does
blow at speeds greater than the maximum output, the turbine must deal with the excess
wind in order to avoid damage (Gipe 1995). Pitch control and stall control are the two
methods available for accomplishing this task. On a pitch control turbine, the electronic
controller checks the power output of the turbine every few seconds. If the power output
becomes too high, then the rotor is signaled to turn slightly out of the wind. When the
wind returns to safe levels, the rotor assumes it original position. A stall controlled
turbine uses the concept of stall to manage high wind scenarios. Its blades are connected
to the hub at a fixed angle. The turbine has been aerodynamically designed so that when
the wind speed becomes too strong, turbulence is created on the non-wind facing side of
the motor blade. This prevents the lifting force from acting on the rotor, creating stall
(Danish Wind Industry Association 2004).
2) The drive train of the HAWT contains the rotating parts of the turbine. It
consists of a low-speed shaft, a gearbox, a high-speed shaft, support bearings, couplings,
a mechanical brake, and the generator (Manwell et al. 2002). The low-speed shaft
connects the hub of the rotor to the gearbox. It also contains pipes for the hydraulic
system that allow for the operation of the aerodynamic brakes. With a typical 1000 kW
turbine, the rotational speed of the low-speed shaft is between 19 – 30 rpm. The purpose
of the gearbox is to force the high-speed shaft to operate about 50 times faster than the
low-speed shaft. The high-speed shaft drives the electrical generator and has an
11
operational speed of around 1,500 rpm. It also contains an emergency mechanical disc
brake to be used during the failure of the aerodynamic brake or when the turbine is being
repaired (Danish Wind Industry Association 2004). The generator used on a HAWT is
either an induction or synchronous generator. Both of these designs allow the generator
to operate at nearly constant rotational speeds when it is connected to a utility grid. Most
turbine manufactures opt to use an induction generator because they are inexpensive and
can be simply connected to a grid. This generator type runs within a limited range of
speeds that are slightly faster than the synchronous generator.
3) The control system is imperative in the areas of turbine operation and
production of power. It consists of sensors, controllers, power amplifiers, and actuators.
Traditional control engineering principles are applied in the design of control systems for
turbines. Three guidelines must be followed and kept in balance when considering
turbine control systems:
1. Setting upper bounds on and limiting the torque and power experienced by the
drive train.
2. Maximizing the fatigue life of the rotor drive train and other structural
components in the presence of changes in the wind direction, speed (including
gusts), and turbulence, as well as start-stop cycles of the wind turbine.
3. Maximizing the energy production.
(Manwell et al. 2002)
4) The mainframe and yaw system provide the housing and orientation systems
necessary for turbine operation. More specifically, the mainframe contributes the
mounting and the proper alignment for the drive train components and the yaw
orientation system properly aligns the rotor with the wind, referred to as yawing
(Manwell et al. 2002). An electronic controller is responsible for the operation of the
yaw mechanism; it senses the wind direction using the attached wind vane. An
anemometer and a wind vane measure the speed and direction of the wind. These
measurements are transmitted to the electronic controller, which tell the turbine to start
12
the blades spinning at a minimum speed and stop their movement at the maximum speed.
The yaw system will then correct for the wind, but it will only yaw a few degrees at a
time (Danish Wind Industry Association 2004). Also included in this category is the
nacelle, which is a weather protecting cover for the above machinery.
5) The tower and foundation of a turbine support the nacelle and rotor. The
primary types of towers used today are the free standing type composed of steel tubes
(tubular towers), lattice or truss towers, and concrete towers. Smaller turbines also have
the option of using guyed towers (Manwell et al. 2002). Tubular towers are safer for
personnel because the ladder inside of the tube provides the access to the top of the
tower. The advantage of the lattice design is its cheap construction. Tower height is
typically 1 to 1.5 times the rotor diameter, but is usually at least 20 m (66 ft). A 1,000
kW turbine will normally have a tower 50 to 80 m (150 to 240 ft) high. Generally, it is
advantageous to have a taller tower because of the increase in wind speeds higher from
the ground (Danish Wind Industry Association 2004).
6) The final major component of a HAWT turbine is the electrical system. It
consists of a number of electrical components responsible for general operation and for
the transmission of electricity to a utility grid. Some examples are cables, switchgear,
transformers, power electronic converters, power factor correction capacitors, and yaw
and pitch motors (Gipe 1995).
1.1.3 U.S. Wind Farm Examples
Although a number of countries exploit wind energy on a much larger scale than
the U.S., the U.S. still has some large scale wind energy projects in operation or the
planning stages. An onshore and offshore wind example will be explored in this
subsection. The U.S. only has two offshore examples, one is Cape Wind, and the other is
off the coast of South Padre Island, both of which are still in the permitting phase. The
13
proposed Cape Wind facility will be explored for this research. The onshore wind
project, Foote Creek Rim, is the largest in the intermountain West. It is located near
Arlington, Wyoming and was the first of its kind in the state. Foote Creek Rim is a
treeless plateau bordering I-80 between Rawlins and Laramie. It has one of the highest
wind speeds in the country, with average wind speeds around 25 mph. This speed is 25 –
70 percent faster than most good wind sites. The turbines of Foote Creek Rim are
designed to operate at wind speeds between 8 and 65 mph. Above 65 mph, the turbines
automatically shut down; this is an important feature because winds in this area can reach
125 mph (Bureau of Land Management 2006). The turbines are also manufactured to
withstand the cold temperatures of the area, which can plummet to 30ºF below zero
(American Wind Energy Association 1999). The first phase of the project opened on
Earth Day, April 22, in 1999. Upon opening, the wind farm had an output of 85 MW
through the operation of 69 600-kW wind turbines, enough energy to supply power to
27,000 homes. After the addition of more phases to the project, Foote Creek Rim now
has an energy capacity of 134.7 MW through 183 turbines.
Cape Wind is the first proposed offshore wind farm in the U.S. The farm will be
located miles from the nearest shoreline off the coast of Nantucket Sound on Horseshoe
Shoal. Upon completion, Cape Wind will consist of 130 turbines that will be capable of
producing 420 MW. The Cape Cod and Nantucket Island areas have seen electric prices
double in the last five years. The installation of the Cape Cod facility will help to reduce
prices and supply the areas with ¾ of their energy needs. Permitting is expected on the
project until 2007 and the Cape Wind wind farm is predicted to be operational by the end
of 2009 (Cape Wind 2006).
14
1.2 Scope of Research
This research investigates the potential to implement CAES systems in porous
media and compares this option to the use of CAES in a more conventional cavern
setting. Using the ECLIPSE 100©, Black Oil Simulator, from GeoQuest Schlumberger,
the base case cavern model was constructed using data from the first CAES plant, the 290
MW E.N. Kraftwerke plant, located in Huntorf, Germany. This plant utilizes 2 mined
salt caverns for the storage of compressed air. This is designated as the base model
because the caverns represent a 100% porosity scenario. After calculating PVT
properties, such as gas formation volume factor and viscosity for air, necessary for the
ECLIPSE 100© input file, a history match of the pressure response associated with the
compression and expansion schedule of the Huntorf facility could be obtained through
the modeling of various sensitivities in Eclipse. These sensitivities include different
permeabilities, alternate pore volumes, changing production rates, and changing injection
rates. The main objective of these sensitivities is to obtain a pressure match and to ensure
that the specified rates are achieved in the model runs. The quality of the match was
verified with the calculation of different objective functions.
This optimal base model could then be converted to a model representing a
hypothetical reservoir to provide a sound judgment for the practicality of utilizing CAES
in porous media. The porous model was constructed using EZGEN, a grid simulator for
ECLIPSE 100©. The volume of the reservoir is determined by using the Huntorf model
volume as the reservoir pore volume and then calculating bulk volume with a 20%
porosity. The rate schedule decided on in the cavern modeling is carried over to the
reservoir model. Various permeabilities are used to create different grids for reservoir
model sensitivities. A base reservoir dataset can then be created in ECLIPSE 100© using
an EZGEN file with 100 md permeability. Figure 1.6 demonstrates the scenario
simulated for porous media.
15
Wind Power
Reservoir Production
well Injection
well
Figure 1.6 Sketch of CAES for a Wind Farm
The injection well is responsible for injecting the compressed air into the reservoir, while
the production well allows the compressed air to return to the surface. To satisfy the
Huntorf rate requirements, additional modeling of various sensitivities, such as different
permeability and alternate pore volumes are necessary.
In order to provide a real world application of CAES in porous media, different
injection and production rates are examined with a range of porosities and permeabilities
consistent with the Greater Green River Basin (GGRB). This area was chosen for study
because of its location in the Rocky Mountain region and its proximity to the Foote Creek
Rim wind farm described in Section 1.1.3. Additionally, parameters such as porosity and
permeability could be obtained for study in this region. With the combination of wind
energy and CAES, an additional energy source is possible that is both environmentally
friendly and cost competitive with existing energy options.
16
1.3 Research Objectives
The main objectives of this research are to determine if CAES can be used in a
reservoir setting and if ECLIPSE 100© is the appropriate tool to model this storage
technology. In order to accomplish these objectives, various ECLIPSE 100© models were
constructed in various geographic locations and geologic settings. By using known
pressure data from an injection and production schedule at Huntorf and achieving an
adequate pressure match, this model could be modified to a hypothetical reservoir model.
After matching the same injection and production rates as the Huntorf facility, this
idealized reservoir model could be applied to actual reservoirs within the GGRB. An
optimal injection and production schedule could be found based on the formation
properties and calculated power output based on modeled production rates.
Some other research objectives include determining whether or not the GGRB
would be a good location for the implementation of combined wind energy and CAES.
The ability for model comparison between the three main models is the final research
objective.
1.4 Application from Petroleum Industry
The process of CAES has a direct application to the petroleum industry. As
previously stated, hydrocarbon resources are on the decline and sustainable energy
sources need to be researched to ensure a smooth transition in the future. A petroleum
engineer is the ideal candidate to explore CAES. The same hydrocarbon reservoir issues,
such as fluid PVT properties, porosity, permeability, relative permeability, and
saturations, exist in the geologic structures associated with CAES. The same tools can be
used to evaluate these properties and reservoir simulators, such as Eclipse, can be used in
17
both instances. The combination of these abilities demonstrates the practicality of
applying petroleum engineering to the CAES process.
The idea of storage is not new to the petroleum industry; in fact it has been a
staple in the industry for a number of years through the concept of natural gas storage.
The first successful implementation of natural gas storage occurred in 1915, in Weland
County, Ontario, Canada. A depleted natural gas well was converted into a successful
storage facility. The first U.S. application of natural gas storage was located near
Buffalo, NY. The idea increased in popularity shortly after WWII and by 1979 a total of
7.5 Tcf of storage was present in 26 states. This amount of storage could be accounted
for in more than 399 pools (Katz and Tek 1981). The concepts behind natural gas storage
and CAES are quite similar and can be summarized as the storage of excessive energy in
the form of natural gas or air for use at a later date. The objectives of natural gas storage
and large scale energy storage are the same. These include compensating for
shortcomings in the energy market, such as peak demand, price volatility, weather, and
grid stabilization. Natural gas storage also relies heavily on natural gas prices. When
prices are low gas can be stored until prices increase to the desired level (NaturalGas.org
2004).
Natural gas storage occurs in depleted reservoirs, mines, aquifers or salt caverns
(NaturalGas.org 2004). Until 1950, natural gas storage only used depleted reservoirs as
its storage devices. Depleted reservoirs are the method of choice for storage because of
their low cost and the fact that hydrocarbons have already been produced out of the
reservoir. Porosity and permeability are the main factors driving natural gas storage in a
depleted reservoir. Porosity determines how much gas can be stored and permeability
monitors the flow rates of injection and production. If storage is desired in locations, not
close to a depleted reservoir, then an aquifer can be used. This is the least desirable
scenario because the water has to be removed from the aquifer, which can be costly.
Additionally, the geologic properties have to be determined since these are typically
unknown; this adds to the increased cost as well. The salt cavern is ideal for natural gas
18
storage, but as previously stated, these kind of structures are hard to come by
(NaturalGas.org 2004). Figure 1.7 shows the location of the existing natural gas storage
facilities within the U.S.
Figure 1.7 Locations of natural gas storage facilities within the U.S.Source: (National Energy Technology Laboratory 2004)
At the moment, CAES mainly occurs in salt caverns, which is ideal for the storage
of air and natural gas, but the expansion of CAES into depleted reservoirs would open the
door for a number of additional facilities. The idea behind natural gas storage and CAES
is the same. Therefore, CAES could exploit the same geologic structures as natural gas
storage. This research will explore the option of using CAES in a reservoir setting,
similar to that of natural gas storage.
19
CHAPTER 2
LITERATURE REVIEW
The electricity chain is currently thought of as having 5 key steps (Figure 2.1).
However, in order to create a competitive market and compensate for shortcomings such
as peak demand, grid stabilization, and price volatility, energy must be stored. In other
commodities, storage is presented as one of the steps on their commodity product
timelines. Electricity itself cannot be stored for future use, but the energy necessary to
create it can be, by means of fuel energy, potential energy of a stored fluid, mechanical
energy, and chemical energy. Figure 2.2 shows an adjusted electricity chain with the
addition of storage (Makansi 2001).
Figure 2.1 Conceptual electricity chain currently usedSource: (Makansi 2001)
20
Figure 2.2 Conceptual electricity chain with the addition of storageSource: (Makansi 2001)
2.1 Benefits of Large-Scale Energy Storage
The benefits from adding storage to the electricity chain can be seen in every step
of the process. Figure 2.3 conveys these benefits along with the industry challenges
associated with them.
21
Figure 2.3 Benefits of energy storage along the electricity value chainSource: (Makansi 2001)
Storage makes sense from both a consumer and supplier standpoint. With
storage, any excess energy produced from a variety of sources can be stored for later use
instead of being wasted. Similarly, if a period of peak demand is encountered, then
stored energy can be used to compensate. This allows power facilities to be designed for
high demand instead of peak demand, helping to decrease costs. Storing energy for use
during peak demand also allows transmission and distribution systems to operate at their
full capacities. This eliminates the need for newer or upgraded lines. Storing energy is
independent of weather; therefore shortages should be minimized with the addition of a
storage component. Additionally, storing energy for shorter time periods can aid in the
smoothing of small peaks and sags in voltage. If efficiency and cost savings can be
improved on the supplier end, then these improvements can be passed on to the consumer
in the form of decreased costs. The benefits of using storage to meet demand on a diurnal
basis are shown in Figure 2.4 (Makansi 2001).
22
Figure 2.4 Load profile of a large-scale energy storage facilitySource: (Makansi 2001)
Storing energy uses baseload generation more efficiently and eliminates the requirement
of peaking facilities. By storing energy during low demand periods and releasing this
energy during peak demand periods, significant price reductions can also be realized.
The need for storage is definitely apparent, especially for large-scale applications.
Large-scale storage can be thought of as being of the same scale as current energy
production methods. The following subsections will provide an introduction to the
methods available or currently under development for large-scale energy storage.
2.1.1 Fuel Cells
Fuel cells are an example of using stored chemical energy for conversion into
electrical energy. In most cases, a gaseous fuel is used as the energy source. Instead of
burning the fuel, it is reacted with oxygen in the atmosphere. This allows the energy to
be directly converted into electricity, thus increasing its efficiency and eliminating
23
pollutants. The most common and preferred fuel for fuel cells is hydrogen. The
combustion of hydrogen in oxygen only produces water and the energy density of
hydrogen is quite high when compared to other fuels. Hydrogen has an energy density of
38 kWh/kg, while the next highest of available energy options is gasoline, with an energy
density of 14 kWh/kg (Cheung, Cheung, De Silva, Juvonen, Singh and Woo 2006).
The composition of a fuel cell includes an anode and a cathode that are separated
by an electrolyte (Figure 2.5). Hydrogen is passed through the anode and oxygen is
passed through the cathode. This leads to the formation of hydrogen ions and electrons at
the anode. These hydrogen ions travel to the cathode by means of the electrolyte and the
electrons travel to the cathode through an external circuit. Water is formed at the cathode
through the combination of hydrogen and oxygen. The current of the fuel cell is supplied
by the flow of electrons through the external circuit.
Figure 2.5 Composition of a fuel cellSource: (Cheung et al. 2006)
24
Various types of fuel cells exist including alkaline, polymer electrolyte
membrane, molten carbonate, and solid oxide cells. These different types of fuel cells all
operate according to Figure 2.5, but vary in their electrolyte composition, operating
temperature, and efficiency. Additionally, regenerative fuel cells are in the development
stage. Just like traditional fuel cells, regenerative fuel cells use stored chemical energy
for the conversion to electricity. Their difference comes in how the energy is stored.
Instead of using hydrogen and oxygen to store energy, an electrolytic solution is
employed. Fuel cells are very efficient and do not produce any pollutants. Conversion
from gasoline and diesel engines to fuel cells in automobiles is a practical and viable
option (Cheung et al. 2006).
2.1.2 The Flywheel
Flywheels have been used for a number of years in steam boats and windmills as
a means for transferring energy. Simply put, a flywheel is a spinning disc with a hole in
the center for rotation (Research Reports International 2004). Figure 2.6 shows the
components of a typical flywheel.
Figure 2.6 Cross-section of a typical flywheelSource: (Cheung et al. 2006)
25
Initially, flywheels were composed of metal. The idea was to increase the mass of
the flywheel in order to increase the rotary inertia and subsequently store kinetic energy.
Flywheels can be designed to release a large amount of energy in a short period of time
or a small amount of energy in a longer time period. The first flywheels were designed
for long lasting energy storage, but this caused a decrease in rotational speed and an
energy output of just 5% of the total available energy of the spinning flywheel. Rectifiers
were added that increased the usable energy to 75%. This addition added an increase in
price and caused aerodynamic friction since these models operated in air. These
traditional flywheels were composed of steel, a low specific strength material when
compared to composites used in flywheel design (Table 2.1), and could only operate at
low speeds. Steel flywheels can produce 1,650 kW of power, released over a few
seconds (Cheung et al. 2006).
Table 2.1 Properties of various flywheel building materials Source: (Cheung et al. 2006)
Table 2.1 demonstrates that materials such as a carbon fibre composite can allow
for higher speeds due to its low density and high specific strength. For this reason, the
invention of composite flywheels (Figure 2.7) was justified. Composite flywheels are
smaller in size and capable of storing large quantities of energy for a short period of time.
26
A prototype composite flywheel reached speeds of 100,000 rpm with the speed at the tip
exceeding 1,000 m/sec, but most commercial models operate around 68,000 rpm
(Research Reports International 2004). 750 kW of power can be released over 20
seconds, or 100 kW can be released over an hour. One of the downsides of composites
designs is that mechanical bearings cannot be used because of the high rotational speeds.
Therefore magnetic bearings are utilized in composite designs. Magnetic forces are used
to levitate the rotor and eliminate frictional losses from rolling elements and fluids. The
heat produced by the ohmic losses in the flywheel becomes trapped because of the partial
vacuum created by aerodynamic losses. Additionally, advanced control systems are
necessary to operate the levitation system and the bearings have lower specific strengths
than the composite flywheels themselves causing a decrease in maximum flywheel speed
(Cheung et al. 2006).
Figure 2.7 A composite flywheel Source: (Cheung et al. 2006)
27
The current application of flywheels of all designs is to provide high output
voltage assistance to components or machines during a power surge or a shutdown.
Research is being conducted for the use of flywheels in other disciplines. For example,
flywheels are being used in the starting and braking of locomotives and as a battery
replacement in electric vehicles. There is still a lot of work to be done in perfecting the
flywheel design, but the concept is sound and their effectiveness has already been proven
in many applications (Cheung et al. 2006).
2.1.3 Superconducting Magnetic Energy Storage
A superconducting magnetic energy storage (SMES) system is designed to store
and instantaneously discharge large amounts of power. The flow of DC in a
superconducting coil that has undergone cryogenic cooling creates a magnetic field
(Research Reports International 2004). A SMES system utilizes this magnetic field to
store energy (Figure 2.8). Storage is made possible because of the superconducting
material of the coil. This material type enhances storage capacity because in a low
temperature environment, their electric currents experience a minimal amount of
resistance. This need for a low temperature environment creates a less than desirable
situation. The ability to maintain minimal resistance without the temperature restraint is
currently being investigated. The idea of a SMES system first arose courtesy of Ferrier in
1969, but the first system was not built until 1986, a 5 MJ system. For several years now,
SMES systems have been used to improve industrial power quality and to service
applications prone to voltage fluctuations (Cheung et al. 2006).
28
Figure 2.8 SMES system designSource: (Cheung et al. 2006)
Currently, SMES systems are capable of storing up to 10 MW and research
groups have been able to produce systems that can store hundreds of MW. Some
researchers believe that SMES systems have the potential of generating up to 2,000 MW
of power, but this has yet to become a reality. In theory, a 150-500 m radius coil can
handle a load of 5,000 MWh at 1,000 MW. The energy loss associated with SMES
systems is small at about 0.1% per hour; this loss is necessary for cooling of the system.
The minimal loss is due to the fact that the energy is stored directly within the magnetic
field generated by the coil. SMES systems are also environmentally friendly because
superconductivity does not yield a chemical reaction and no toxins are created (Cheung et
al. 2006).
29
2.1.4 Supercapacitors
Capacitors ability to hold DC voltages are one of the main components of electric
circuits. In the past, it was believed that capacitors could only store as much energy as
what is found in a normal battery. However, research has shown that this energy storage
method could be utilized on a much larger scale. Conceptually, capacitors could be used
to store energy for extended periods of time. Capacitors are composed of two conductive
parallel plates that are separated by a dielectric insulator (Figure 2.9). To create an
electric field, the plates hold opposite charges (Research Reports International 2004).
Unlike batteries that store energy chemically, capacitors stores energy in an electric field.
Figure 2.9 Principles behind a capacitorSource: (Cheung et al. 2006)
30
The first supercapacitor was not developed until 1997. Researchers at CSIRO
discovered that a significant amount of charge could be stored if the dielectric layer was
composed of thin film polymers and the electrodes were constructed out of carbon
nanotubes. A normal capacitor has an energy density of 0.5 Wh/kg, but these
supercapacitors can store four times the amount of energy. Supercapacitors are a
practical replacement for the traditional battery. Even though they have a larger energy
density than batteries, supercapacitors are not plagued with many of the disadvantages
associated with batteries. Batteries have a limited number of charge/discharge cycles and
take time to charge and discharge because of the chemical reactions necessary for the
process. The acidity of batteries is also harmful to the environment when the lifetime of
the battery has expired. Supercapacitors have an unlimited number of charge/discharge
cycles. They can discharge in milliseconds and can produce significant amounts of
currents. The lifetime of a supercapacitor is extremely long and no hazardous substances
are produced that could harm the environment. Currently, the main uses of
supercapacitors are within hybrid vehicles and handheld electronic devices (Cheung et al.
2006).
2.1.5 Underground Thermal Energy Storage
Using the subsurface to store energy is an efficient and aesthetically pleasing
option. Instead of constructing a large power plant on the surface, underground thermal
energy storage (UTES) can be used. The whole operation is invisible to the human eye
and even with extensive research, no drawbacks have been found with UTES (Cheung et
al. 2006).
One type of UTES is aquifer thermal energy storage (ATES), in which aquifers
are used for the storage process. Figure 2.10 shows the concept behind ATES.
31
Figure 2.10 Conceptual design of ATESSource: (Paksoy 2005)
With ATES, two wells are used, one for warm water and the other for cold water. During
the winter months, warm water is cooled and transported to the cold well. A heat
exchanger can then collect the energy for heating purposes. In the summer a reversal
occurs and the cold water is utilized for cooling. After the water is heated, it is stored in
the cold well (Paksoy 2005). ATES is environmentally safe since the water always
travels within the system. Additionally, no net loss of water occurs within the
32
subsurface. The only drawback of ATES is geologically based; only areas that are above
aquifers can utilize this storage method. Europe and Asia have implemented ATES the
most. By 1984, China had 492 cold storage wells to help cool down machinery and the
same type of system was constructed in Sweden, but on a smaller scale. Fossil fuel
consumption can be reduced by 80-90% with the use of ATES (Cheung et al. 2006).
Duct thermal energy storage (DTES) is a much more complex operation than
ATES (Figure 2.11).
Figure 2.11 Conceptual design of a DTES systemSource: (Paksoy 2005)
33
With DTES, holes are drilled to accommodate heat exchangers; these holes are
typically 50-200 m (164-656 ft). The efficiency of a DTES system is dependant on a
number of parameters, such as the ground temperature, the operational temperature of the
storage area, the conditions of the groundwater, and the thermal properties of the ground.
All of these parameters vary with location and can have a significant impact on efficiency
(Paksoy 2005). The first application of a large-scale DTES system was in Lulea,
Sweden. 120 holes at 60 m (197 ft) were drilled for the purpose of storing warm thermal
energy of about 70 ºC (158 ºF) to heat the local university. The largest DTES system in
operation today is located in Fort Polk, LA, U.S. It consists of 8,000 holes used to
provide warm and cool thermal energy to residents (Cheung et al. 2006).
Both UTES applications are environmentally friendly. UTES reduces the amount
of electrical energy necessary for heating or cooling. For cooling purposes, UTES
diminishes the use of mechanical or chemical cooling, which produces hazardous
substances that are responsible for damage to the ozone layer and rivers and lakes. A
reduction in cost is also seen because of the more accurate performance of UTES. For
heating applications, UTES replaces conventional heating methods that produce harmful
gases such as carbon dioxide and nitrogen oxide. UTES is a clean system due to the fact
that there is no net water loss. It is a practical alternative to countries that have low fossil
fuel reserves and it is also a viable option for storing solar energy for future use (Cheung
et al. 2006).
2.1.6 Pumped Hydroelectric Energy Storage
Pumped hydro is the oldest energy storage option, being in use since 1929.
Currently, 90 GW of power worldwide can be attributed to pumped hydro; this accounts
for 3% of the global energy capacity (Research Reports International 2004). The idea
behind a pumped hydro plant is relatively simple. Two large reservoirs are located at
34
different elevations. When energy is needed, water is released from the upper reservoir.
The water travels through high-pressure shafts, turbines, and eventually ends up in the
lower reservoir. When the demand for energy is low, water can be pumped back up into
the upper reservoir for future use. The general operation of a pumped hydro facility can
be viewed in Figure 2.12.
Figure 2.12 Operation of a pumped hydroelectric storage facilitySource: (Cheung et al. 2006)
The greater the vertical distance between the two reservoirs, the higher the head of the
system will be. A higher head means more energy; therefore, reservoirs with a large
vertical separation are desired (Cheung et al. 2006).
One of the best known pumped hydro plants is the Dinorwig plant located in
Wales. Construction began in 1976 and concluded in 1982. Upon completion, Europe
had its largest manmade cavern. The facility consists of six large pump turbines capable
35
of generating 317 MW each. Together the turbines can produce 1800 MW of power from
a water volume of 6 million m3 (212 million ft3) and an operating head of 600 m (1,969
ft). The system response is impressive with any of the turbines being capable of full
power in 10 seconds, if it is already spinning. Even if the system is at a complete
standstill, full power can be obtained in one minute.
Pumped hydro is one of the most effective forms of large-scale energy storage. It
has a storage capacity of over 2,000 MW and can store its energy for over half a year.
After half a year, leakoff and sealing properties become an issue. The response time to
get a system up and running is minimal and its operating cost is quite low due to its
straightforward design. Additionally, no environmental hazards are created from the use
of pumped hydro. This type of storage does have some disadvantages. The main setback
is the lack of suitable geologic formations. Two large reservoirs have to be present with
enough vertical distance to establish a usable head. This type of scenario is uncommon
and usually occurs in mountain settings where connection to a grid is unlikely and
construction is quite cumbersome. The capital cost of building a pumped hydro facility is
quite large due to the construction of dams and very large underground pipes. There are
parts of the world where the potential for pumped hydro exists. However, in the U.S.
most of the areas suitable for this technology have already been developed (Cheung et al.
2006).
2.1.7 Compressed Air Energy Storage
More emphasis will be placed on this energy storage technology since it is the one
chosen for continued research. A CAES system uses air pressure as its means for energy
storage; the air’s energy is readily available for extraction for future power generation.
Figure 2.13 shows the components and cycle of a CAES system. In principle, a CAES
system is a modification of a standard gas turbine generation cycle. In a typical cycle, the
36
turbine is connected directly to an air compressor. Consequently, whenever gas is
combusted in the turbine, about 2/3 of the energy output has to be used for compression.
CAES separates the combustion and generation cycle from the compression cycle. This
allows for air to be compressed using off-peak energy before it is required for energy
usage. This compressed air is stored in a subsurface reservoir, ready for extraction.
Typically, a solution-mined salt cavity, a mined hard rock cavity, or an aquifer is used for
underground storage. Above ground storage in tanks is also an option, but this alternative
is quite costly. Since the air is already compressed, a CAES turbine can generate three
times the amount of electricity as a simple cycle turbine that requires connection to the
air compressor (Ridge Energy Storage & Grid Services L.P. 2005).
Figure 2.13 Operations of a CAES facilitySource: (Ridge Energy Storage & Grid Services L.P. 2005)
37
In a CAES cycle, air is collected by the compressor, compressed, and then
injected into a reservoir via an injection well. Existing CAES technology requires the air
to be cooled during compression to allow for storing near ambient temperature. When
energy is demanded, the compressed air is extracted by means of a production well. The
air must be heated to avoid freezing the system components during air expansion in the
turbine. Currently, this is done through a recuperator and the burning of a fuel such as
natural gas. The air and natural gas are then expanded in the turbine to generate
electricity. This means that a traditional CAES plant is not completely pollution free;
CO2 emissions are still present (Greenblatt, Succar, Denkenberger, Williams and
Socolow 2006). It has been shown that carbon-neutral fuels are an option (Denholm,
Kulcinski and Holloway 2005) and, theoretically, it is feasible to store the heat generated
by compression separately from the air itself (Bullough, Gatzen, Jakiel, Koller, Nowi and
Zunft 2004). This would completely eliminate the need for fuel in a CAES facility. The
compressed air that enters the turbine does take the place of gas that would have been
used in the generation and compression processes, which significantly decreases CO2
emissions.
The principal equipment of a CAES facility can be split into four components:
“(i) the power island, (ii) the compression island, (iii) the underground portion, and (iv)
the balance of the plant (Ridge Energy Storage & Grid Services L.P. 2005). The power
island contains the turbine, the generator, and the recuperator. CAES designs typically
have two turbines, one high pressure (HP) air turbine and a low pressure (LP) gas turbine.
The HP turbine reduces risk by moderating pressure, temperature, and airflow upon
entrance into the LP turbine. These parameters are altered to values that the LP turbine
would experience if a compressor was still attached. The compression island provides
the required air volume to increase the pressure from atmospheric to the desired pressure
in the underground storage reservoir. A typical compression cycle for a CAES plant
contains a train of axial and centrifugal compressors. The compressors are connected to
the underground storage system. The subsurface facilities vary depending on the scope
38
of the project. In order to produce more power, a reservoir with a larger volume is
necessary. Typically, caverns are sized to store up to 50 hours of power and operate
between 950 – 1,250 psig. The balance of the plant contains the remaining equipment
that is pertinent to the operation of a typical power plant. This includes the cooling
tower, the switchgear, the substation, the plant distribution and controls, and the step-up
transformers. The main difference between the balance of a typical power plant and the
balance of a CAES plant is the auxiliary transformer. This component deals with the
power entering the plant from the grid. An auxiliary transformer in a CAES plant is
designed to accommodate 200 MW of power from compression. Also included in the
balance of the plant is the control room, maintenance facilities, fuel metering and control
valves, water treatment and related facilities, such as pumps and tanks (Ridge Energy
Storage & Grid Services L.P. 2005).
CAES has the advantage of being able to operate on a very large scale. It has a
very high storage capacity around 50-300 MW. It has the largest storage capacity of the
energy storage technologies due to its minimal losses. A CAES system is capable of
storing energy for up to one year due to the high quality seal that salt caverns create.
However, after a year’s time, pressure leakoff becomes a concern. CAES also has the
advantage of a fast start up time. If an emergency start is necessary, then 9 minutes is
needed to get everything up and running. Under normal conditions, a start-up time of 12
minutes can be expected. Conventional turbine plants require 20 to 30 minutes for a
normal start-up. Additionally, since all of the storage is beneath the surface, huge
expensive installations are unnecessary and the storage is invisible to the human eye.
The main drawback of CAES is its reliance on geologic structures. Underground
caverns are the only structures currently in use and these are quite rare in occurrence. An
objective of this research is to show that CAES can be conducted in other geologic
settings.
Currently, there are only two CAES plants in operation today; they are the 290
MW Huntorf facility, located in Huntorf, Germany (Figure 2.14) and the 110 MW
39
McIntosh facility, located in McIntosh, AL. E.N. Kraftwerke currently operates the
Huntorf facility; it was opened in 1978 and has been in successful operation since its
commencement. Upon opening, the main purpose of the Huntorf plant is as an
emergency reserve in the case of failure of surrounding power plants. Now, the plant
also serves as a supplemental energy source for the growing number of wind farms in
northern Germany (Crotogino, Mohmeyer and Scharf 2001). The Huntorf facility will be
explained in greater detail in the beginning of Chapter 3.
Figure 2.14 Aerial view of the Huntorf plantSource: (Crotogino et al. 2001)
The McIntosh facility was built in 1991 by Dresser-Rand and is currently owned
by the Alabama Electric Corporation. It made several improvements on the Huntorf
design. One of these improvements incorporated a recuperator (an air-to-air heat
exchanger) to recover the exhaust heat to preheat the cavern air upon entry into the
turbines, thus improving the heat rate and reducing fuel usage by 25%. It uses a roughly
cylindrical salt cavern about 300 m deep and 80 m in diameter (total volume of 5.32
40
million m3). Pressures range from 45 to 74 bar (653 – 1,073 psia) and the plant can
supply power for 26 hours. Start up times range from 9 to 13 minutes. Figure 2.15
provides a comparison of the specifics of the Huntorf facility to those of the McIntosh
facility (Ridge Energy Storage & Grid Services L.P. 2005).
Figure 2.15 Comparison of Huntorf and McIntosh facilitiesSource: (Ridge Energy Storage & Grid Services L.P. 2005)
Other projects include a 25 MW R&D facility, named Sesta, which was in
operation in Italy during the early 1990s, but has since been decommissioned. A wind-
CAES facility is in the works in Iowa. It already has some investors for the project
development and is currently seeking funding for capital purchases. Air will be stored in
41
an aquifer instead of the typical cavern (Holst 2005). Based on projected economics, the
plant will have a CAES energy capacity of 200 MW and a wind energy capacity of 100
MW (Research Reports International 2004). The plant is scheduled to open in April 2010
The largest CAES facility ever to be constructed, as well as, the largest energy storage
system in the U.S. is being developed in Norton, OH. It is a 2,700 MW plant consisting
of 9 turbines; the facility will be able to compress air to 104 bar (1,508 psia) in a man-
made limestone mine approximately 670 m (2,200 ft) beneath the surface. The plant has
an underground capacity of 9.5 million m3 (338 million ft3) (Holst 2005). The Norton
CAES facility is currently in the permitting stage of the project. The plan is to build the
facility in phases over a five year period. Upon completion, the facility will be able to
provide 2,700 MW of energy with the emissions comparable to a 600 MW gas-powered
combustion plant (Research Reports International 2004).
Ridge Energy Storage has two planned projects, one in Markham, Texas and the
other entitled Pierce Junction in Houston, Texas. The Markham plant plans on utilizing
four 135 MW turbines (540 MW total power) and four 100 MW units for compression.
Several caverns are available at the site for storage. Permitting has been obtained for
both the caverns and the air regulations (Ridge Energy Storage & Grid Services L.P.
2005). In spite of these advances, in a May 21, 2006 summary updating electricity
generating facilities for the state of Texas, the Markham CAES project was listed as
cancelled. Pierce Junction has designed for two 135 MW turbines (270 MW total power)
and two 100 MW compression units. One cavern is currently permitted and available for
storage. If an additional cavern was added along with the CAES equipment, the total
power output would be either 405 MW or 540 MW. The above summary has no word on
the Pierce Junction facility (State of Texas 2006).
42
2.2 Advantages of CAES over Other Storage Technologies
Although the aforementioned storage techniques can be effective at a large scale,
only pumped hydroelectric storage and CAES are proven to be cost-effective at the large
temporal scales, consisting of several hours to days (Greenblatt et al. 2006). Figure 2.16
shows the electricity generating capacity of the major storage technologies. This figure
demonstrates that pumped hydroelectric storage and CAES are also the only technologies
capable of storing high amounts of energy for future electricity conversion. Figure 2.17
shows the installed costs of major storage technologies.
Figure 2.16 Power ratings of major storage technologiesSource: (Makansi 2001)
43
Figure 2.17 Cost and performance of major storage technologiesSource: (Research Reports International 2004)
Simply looking at installed costs does not provide an accurate representation of
the associated variable costs or the amount of energy that can be produced with a
technology. Therefore, a balanced annual cost for each technology must be examined.
This annual cost takes into account amortized installed cost of the system, the operating
and maintenance cost, fuel cost, and replacement costs. Figure 2.18 shows the annual
cost for the same technologies as in Figure 2.17. Figure 2.18 can then be divided by the
hours of operation per year to provide further detail.
44
Figure 2.18 Levelized Annual Cost of Bulk Storage OptionsSource: (Australian Greenhouse Office 2005)
In Figure 2.17, CAES is shown to be one of the more costly options. However,
when a balanced annual cost is considered CAES provides the largest amount of storage
at the lowest cost. Figure 2.19 displays the capital costs of major energy storage options.
Based on this figure, CAES has one of the lowest capital costs per unit energy and capital
costs per unit power. CAES also has the advantage of providing a large amount of stored
energy for an extended period of time at low cost. Some of the other storage
technologies can provide one or two of these services, but only CAES can supply all
three.
45
Figure 2.19 Capital cost of major storage technologiesSource: (Research Reports International 2004)
The previous figures demonstrate the practicality of CAES; the only real
challenger to CAES is pumped hydroelectric. However, CAES has the upper hand over
pumped hydroelectric for a couple of reasons. Firstly, public and environmental pressure
is making the building of above ground storage facilities quite difficult, to the point that
future facilities may be prevented from being constructed. Even if pumped hydroelectric
is conducted in the subsurface, CAES is considered to be about 20% more efficient and
around 1/3 of the cost of pumped hydroelectric (Makansi 2001).
In addition to being cost effective, CAES can be utilized throughout the U.S. A
1990s DOE study found that approximately 85% of the land in the U.S. would be
accessible and of a suitable geology for CAES development (Ridge Energy Storage &
46
Grid Services L.P. 2005). Figure 2.20 shows the areas of the U.S. that have the potential
for CAES.
Figure 2.20 Regions of the United States suitable for CAESSource: (Ridge Energy Storage & Grid Services L.P. 2005)
This map shows huge potential for the combination of CAES and wind farms.
Throughout the northwestern and midwestern regions, suitable geology exists for CAES.
These locations are also where a majority of class 3 or higher wind sites exist (Figure
1.1).
47
2.3 Fundamentals of Reservoir Simulation
Models have been a part of human life for thousands of years. Simply put, a
model is “used to obtain a better understanding of the environment and to predict the
behavior of physical phenomena under the constraints of nature’s laws” (Thomas 1982, p.
1). When examining reservoirs, one cannot physically view the reservoir; therefore a
model has to be employed to simulate the reservoir’s behavior. In order to solve
reservoir engineering problems within the oil industry, reservoir simulation has become
the method of choice. “Reservoir simulation is the art of combining physics,
mathematics, reservoir engineering, and computer programming to develop a tool for
predicting hydrocarbon reservoir performance under various operating strategies” (Abou-
Kassem, Farouq Ali and Islam 2004, p.1). Reservoir simulators are primarily based off
of the material balance equation (MBE); Schilthuis first introduced the MBE in 1936
(Schilthuis 1936). He envisioned a reservoir as a sealed tank with uniform properties
throughout. Therefore, the net change in volume is simply the subtraction of the volume
of fluids leaving the tank from the volume of fluids entering the tank. For this reason, it
is also called the tank model. This original form of the MBE has some major deficiencies
that limit its use in reservoir simulators. It does not allow for spatial variation of
reservoir parameters, such as rock and fluid properties. The MBE also does not take the
actual reservoir geometry into consideration and fluid movement within the reservoir is
ignored. To correct for these shortcomings, an alternative form (Equation 2.1) was
introduced (Thomas 1982).
Rate of Fluid In – Rate of Fluid Out = Net Change in Fluid Rate (2.1)
With this model, Darcy’s Law (shown for a single-phase flow in a porous media
in Equation 2.2) can be used and the dynamic behavior of fluid movement can be
correctly applied. If the reservoir is split into a collection of small individual blocks, then
48
fluid properties can be defined differently for each block. This method allows for the
deficiencies of the MBE to be overcome and can be seen in 1-D models. The method can
be extended to 2-D and 3-D simulators, where more detailed rock and fluid property
distribution and flow discretization are possible.
(2.2)
where = fluid velocity, = the permeability tensor, = viscosity, and = the
gradient of the potential function = h , where = fluid density and h = the flow
potential (Thomas 1982).
Reservoir simulators typically fall into three groups: gas reservoir simulators,
black oil simulators, and compositional reservoir simulators. The simulator used for this
investigation, ECLIPSE 100©, is a black oil simulator. With black oil simulators, all
three phase (gas, oil, and water) can be represented in all proportions. The effect of gas
going in and out of solution with oil can also be simulated. More specifically, “ECLIPSE
100 is a fully-implicit, three phase, three dimensional, general purpose black oil simulator
with gas condensate options” (Schlumberger 2004 p. 23). ECLIPSE 100 was written
using FORTRAN77 and can operate on any computer with an ANSI-standard
FORTRAN77 compiler and sufficient memory. Since ECLIPSE 100 uses the fully-
implicit method, stability can be achieved over long time periods. Usually, the fully-
implicit method cannot be applied on a large scale, but this limitation is ramified by
making use of the Nested Factorization. ECLIPSE 100 has the option to simulate one,
two, or all three phases. With two phases, the model is ran as a two component system
which saves time and computer storage. When establishing the gridding of the reservoir,
more conventional block-center geometry can be used, or corner-point geometry can be
chosen (Schlumberger 2004). Block-center geometry makes calculations at the center of
49
each gridblock, while corner-point geometry can conduct calculations at gridblock
corners.
An input file for Eclipse is created in a free format, using a program such as
Notepad, with the appropriate keywords. These keywords define everything necessary
for a successful simulation, such as the rock (reservoir dimensions, structure tops, net-to-
gross ratios, porosities, permeabilities) and fluid properties (relative permeability,
saturations, formation volume factors, viscosities, fluid densities). The first step in
creating an Eclipse input file for simulation involves dividing the reservoir into
gridblocks. Each cell within the reservoir will have its own x, y, and z coordinate. Using
the appropriate keywords, the above rock and fluid properties can be distributed
throughout the gridblocks. Once these physical parameters are established, keywords are
used to enter wells and any of their available flow rates and pressures into the model.
Production or injection rates can be entered into the model, based off of bottom hole
pressure requirements or dependant on water saturation limitations (Schlumberger 2004).
The time in between calculations (time steps) are also entered with the well information.
This allows for the user to establish a production schedule, an injection schedule, or a
schedule with periods of injection and production.
The purpose of reservoir simulators is to match modeled data to known data.
Results from a model run are compared to previously recorded field data for the modeled
wells. If the results do not provide a satisfactory match, then physical parameters, such
as permeability, relative permeability, saturations, porosity, etc. are changed until a match
can be obtained (Thomas 1982). For instance, if the pressure of a production and
injection schedule is to be modeled over a certain time period, then all of the necessary
rock and fluid parameters are entered into an input model. After running the model, the
pressure curve generated by the simulator can be compared to some known recorded
pressure values. If the pressures match, then the model is a good representation of the
data. If the pressure match is not satisfactory, then the model parameters need to be
adjusted until a match can be obtained. Once this “history match” has been achieved, the
50
model can be used to predict future performance. It is not certain that the model will
accurately predict reservoir performance, but the more data available for model input, the
better the representation of the actual reservoir. This means that having a long time
period of data available for the history match will lead to a more reliable model for
prediction.
51
CHAPTER 3
MODEL STUDY OF CAVERN STORAGE
As previously mentioned, the Huntorf facility was chosen for modeling CAES in
a cavern setting. The plant was chosen for modeling because of its use of infinite
permeability and porosity excavated salt caverns and because of its longstanding
operation. Huntorf’s cavern geology provides an optimal base case for comparison with
varying cavern geologies. This chapter will begin with a detailed description of the
Huntorf facility, followed by the inputs that were needed to construct the model and how
these parameters were justified. The dataset for the ECLIPSE 100© modeling is located
in Appendix A and is referred to as Cavern CAES. This dataset is the base case model
for the study; other sensitivities were explored to obtain the desired history match and to
monitor system response. A discussion of the results from the base model, as well as the
results from various sensitivities will be discussed towards the end of the chapter.
3.1 Description of Huntorf Facility
To begin looking at the Huntorf plant, the specifics for the facility are presented
in Table 3.1.
52
Table 3.1 Specifications of the Huntorf CAES facilitySource: Adapted from (Crotogino et al. 2001)
Metric Units English Unitsoutput turbine operation 290 MW (≤ 3hrs)compressor operation 60 MW (≤ 12 hrs) max air mass flow rates turbine operation 417 kg/s 919.4 lb/scompressor operation 108 kg/s 238 lb/sair mass flow ratio in/out 1/4 1/4Max air volumetric flow rates turbine operation 29.411*106 m3/day 1.039*107 Mscf/daycompressor operation 7.617*106 m3/day 1.345*105 Mscf/daynumber of air caverns 2 2single air cavern volumes 140,000 m3 4,944,053 ft3
170,000 m3 6,003,493 ft3
total cavern volume 310,000 m3 10,947,547 ft3
location of caverns - top 650 m 2,133 ft - bottom 800 m 2,625 ftmaximum diameter 60 m 197 ftwell spacing 220 m 722 ftcavern pressures minimum permissible 1 bar 14.5 psiaminimum operational (exceptional) 20 bar 290 psiaminimum operational (regular) 43 bar 624 psiamaximum permissible & operational 70 bar 1,015.3 psiamaximum pressure reduction rate 15 bar/h 218 psia/h
The two caverns used in the plant design are underground salt caverns that have
been excavated for storage. The total volume could have been obtained with just one
cavern, but two were decided upon for the following reasons. Redundancy was desired
during maintenance or cavern shut-down, two caverns allow for easier cavern refilling
when it is necessary to reduce the pressure in one cavern down to atmospheric, and the
start up procedure for the plant requires that a minimum of 13 bar (189 psia) be obtained
in at least one of the caverns. The depth of the caverns was chosen to allow stable
53
storage for several months and to ensure the specified maximum pressure of 100 bar
(1,450 psia).
An important design in the Huntorf facility was the construction of the cavern
wells. The wells needed to be able to withstand high withdrawal rates of 417 kg/s (919
lb/s), as well as, low pressure losses. These parameters led to the selection a 20”/21”
production string and a 24 ½” casing string. Since a packer was not included in the
design to hold the production string inside the casing, corrosion could occur from moist
air traveling through the annulus. This was deterred by the injection of dry air into the
annulus. Brine also had to be evacuated with a submersible pump because of an
insufficient maximum pressure and an excessive flow rate in the compressor and
unacceptable air velocities in the production string. In order to minimize costs, the
production string was initially composed of structural steel and hung 80 m (263 ft) in the
cavern without support. This was done in order to prevent the entry of salt dust into the
turbines. However, after a few months of operation, serious corrosion problems occurred
with the materialization of a large amount of rust in the filter upstream of the gas turbine.
After much consideration, a fiberglass reinforced plastic (FRP) steel string was selected
as a replacement. The FRP string experienced 20 years of problem free operation, but is
now experiencing corrosion problems itself. Investigations for pipe replacements are
being conducted. Unlike the production string, the casing cannot be replaced. Therefore
extra care was taken when designed this portion of the plant. Cleaning of the string has
been carried out, but no surface corrosion or pitting has been observed (Crotogino et al.
2001).
The CAES cycle of the Huntorf plant is the same as in Figure 2.12, except that the
Huntorf facility does not utilize a recuperator leading to a poorer heat rate (Crotogino
2006). A photograph of its equipment is detailed in Figure 3.1. Overall, the operation of
the Huntorf facility has been very successful. The plant has accumulated over 7,000
starts in its 28 years of operation. It has shown 90% availability and a 99% starting
reliability (EA Technology 2004).
54
Figure 3.1 Components of the Huntorf facilitySource: (Gonzalez, O Gallachoir and McKeogh 2004)
3.2 Cavern CAES Inputs
For the modeling of Huntorf, the parameters contained in Table 3.1 were
represented as closely as possible in the model. Some values required adjusting, but the
model is a realistic representation of the facility. The goal of the Cavern CAES model is
to match pressure for a daily operation schedule of the Huntorf facility. Pressures were
obtained for the match by reading hourly values from the pressure response to a daily
power production curve based on energy demand, shown in Figure 3.2.
55
1 bar = 14.5 psiaFigure 3.2 Daily Power Production and Associated Pressure Response
Source: (Crotogino et al. 2001)
This was the only pressure response that could be obtained for Huntorf. A
personal communication with F. Crotogino, who has been involved with Huntorf since its
opening and is the author of the paper used for data collection (Crotogino et al. 2001),
determined that additional pressure data has not been collected because the plant is
operating problem free. Reading pressures of a graph leads to some error associated with
the pressure values. Various sensitivities were examined to show model reliability and
obtain the best pressure match. To begin with a base case model was used that
maintained the original Huntorf parameters as closely as possible. Table 3.2 shows a
comparison of the actual Huntorf parameters to those used in the Cavern CAES base
model.
56
Table 3.2 Comparison of actual Huntorf data to Cavern CAES base model dataSource: Adapted from (Crotogino et al. 2001)
Huntorf Parameters Model Parameters Metric Units English Units Metric Units English Unitsoutput turbine operation 290 MW (≤ 3hrs) 290 MW (≤ 3hrs) compressor operation 60 MW (≤ 12 hrs) 60 MW (≤ 12 hrs) max air mass flow rates turbine operation 417 kg/s 919.4 lb/s 417 kg/s 919.4 lb/scompressor operation 108 kg/s 238 lb/s 108 kg/s 238 lb/sair mass flow ratio in/out 1/4 1/4 1/4 1/4Max air volumetric flow rates turbine operation 29.411*106 m3/day 1.039*107 Mscf/day 29.411*106 m3/day 1.039*107 Mscf/daycompressor operation 7.617*106 m3/day 1.345*105 Mscf/day 7.617*106 m3/day 1.345*105 Mscf/daynumber of air caverns 2 2 2 2single air cavern volumes 140,000 m3 4,944,053 ft3 150,000 m3 5,116,800 ft3
170,000 m3 6,003,493 ft3 150,000 m3 5,116,800 ft3
total cavern volume 310,000 m3 10,947,547 ft3 300,000 m3 10,233,600 ft3
location of caverns - top 650 m 2,133 ft 650 m 2,133 ft - bottom 800 m 2,625 ft 800 m 2,625 ftmaximum diameter 60 m 197 ft 60 m 197 ftwell spacing 220 m 722 ft 220 m 722 ftcavern pressures minimum permissible 1 bar 14.5 psia 1 bar 14.5 psiaminimum operational (exceptional) 20 bar 290 psia 20 bar 290 psiaminimum operational (regular) 43 bar 624 psia 43 bar 624 psiamaximum permissible & operational 70 bar 1,015.3 psia 70 bar 1,015.3 psiamaximum pressure reduction rate 15 bar/h 218 psia/h 15 bar/h 218 psia/h
The actual cavern shapes (Figure 3.3) could not be duplicated because of the lack
of necessary information and model practicality. Therefore the dimensions of the Cavern
CAES base model are based on the layouts shown in Figure 3.4 and Figure 3.5. Each
cavern is represented as a cuboid, according to the dimensions established in Crotogino et
al. (2001). The two caverns were split into equal volumes for the scope of this study.
Since the maximum diameter of the caverns is 60 m (197 ft), this value had to be
57
converted into x and y coordinates for model input. It was decided that each cavern
would have an x dimension of 40 m (130 ft) based on the given dimension data. The
provided thickness of 150 m (492 ft) was used for the z dimension. The y dimension was
calculated as 25 m (80 ft), using the total model cavern volume of 300,000 m3
(10,233,600 ft3).
Figure 3.3 Actual dimensions of the Huntorf salt cavernsSource: (Crotogino et al. 2001)
58
Figure 3.4 Cross-sectional view of the Huntorf Facility
Figure 3.5 Plan view of the Huntorf Facility
59
These figures were also the basis for the gridding used in the model. The 25 m
(80 ft) y dimensions are split into 5 gridblocks, 5 m (16 ft) in height. The total distance
in the x direction is 300 m (980 ft); this includes both caverns and the 220 m (720 ft)
spacing between them. The model uses a total of 20 gridblocks in the x direction; for the
areas where the caverns are located, the gridding consists of 5 gridblocks per cavern, 8 m
(26 ft) each. The area between the caverns consists of 10 gridblocks, all 22 m (72 ft)
long. For the cavern thickness, 15 layers exist, all 10 m (33 ft) thick. This gridding
scheme was selected in order to obtain the desired level of information without
unnecessary processing. Emphasis was placed on the actual cavern locations and not on
the area between the caverns.
The top of the caverns is 650 m (2,133 ft) and is constant along each cavern and
in the spacing between the caverns. The actual porosity of the caverns is 100%, but
because of the limitations of ECLIPSE 100©, a porosity of 99% had to be used. The
porosity between the caverns is 0% since this area does not impact the pressure response
of CAES. Likewise, the net to gross ratio of the model is one where the caverns are
located and zero in the space between them. The actual permeability for the excavated
caverns is infinite, but once again due the limitations of ECLIPSE 100© some sort of
number is required. For the Cavern CAES base model, a permeability of 10,000 md is
assumed in all directions; a very high permeability was selected to assure that this
parameter did not have an effect on the model response.
Relative permeability of the model is the simplest form of relative permeability
curves since the brine has been removed from the caverns. Although some small amount
of residual brine does exist in the caverns, the vast majority of the structures are filled
with air. Therefore, no residual water or gas saturation exists and relative permeability
for water or gas is the same as the corresponding saturation (Figure 3.6).
60
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Water Saturation (fraction)
Rel
ativ
e Pe
rmea
bilit
y (fr
actio
n)
krw
krg
Figure 3.6 Relative permeability curves for water and air used in ECLIPSE 100© input
An approximation was used for the PVT properties for the water phase since an
analysis of any produced water from the field was unavailable. The water formation
volume factor was calculated according to,
(3.1)
where Bw = water formation volume factor, = change in volume during the pressure
reduction, and = change in volume due to the reduction in temperature (McCain Jr.
1990). and are both found using Figures 16-6 and Figures 16-7, respectively
of McCain Jr. (1990). Assuming a cavern temperature of around 35 ºC (95 ºF) and an
average cavern pressure of 69 bar (1,000 psia), a of 0.008 and a of -0.0005
61
were determined. Entering these values into Equation 3.1, yielded a Bw of 1.0075
reservoir bbl/stb. The coefficient of isothermal compressibility was also needed for
model input. An average compressibility of 3.1*10-6 psi-1 was determined from Figure
16-12 of McCain Jr. (1990), using the same cavern temperature and pressure. The cavern
temperatures and pressures were determined from the initial values along the cavern line
in Figure 3.7. The initial values were used since these profiles represent the response
when emptying the caverns.
Figure 3.7 Pressures, temperatures, and air flow when emptying the caverns
62
The PVT properties for air required more intensive calculations. In order to
calculate the air formation volume factor (Bg) necessary for model input, the following
equation was used:
(3.2)
where VR = cubic foot of reservoir volume, Vsc = standard cubic foot of gas, z = gas-
deviation factor, T = temperature in Rankine, p = pressure in psia and 5.03676 converts
Bg from rcuft/scf to Rbbl/Mscf; this is the unit that is necessary in ECLIPSE 100© when
running models in “field” units. Before this equation could be used, the gas-deviation
factor had to be calculated. Calculations of the gas deviation factor require the use of
critical pressure and temperature. Since the gas is air, no pseudocritical calculations were
necessary. According to Table 5.2 in Towler (2006), the critical pressure and
temperature for air are 546.9 psia and -221.4 ºF, respectively. Based on pressures
ranging from 14.7 to 4,014.7 psia, reduced pressures and temperatures were calculated
with Equations 3.3 and 3.4:
(3.3)
(3.4)
where pc = critical pressure and Tc = critical temperature.
With these values known, the Dranchuk and Abou-Kassem equation of state
(Towler 2006) could be employed to find the gas-deviation factor (z-factor). This
relationship has an average absolute error of 0.486% and a standard deviation of 0.00747
63
when the reduced temperatures and pressures are the following: 0.2 < < 30, 1.0 < <
3.0 and < 1.0, 0.7 < < 1.0. The Dranchuk and Abou-Kassem equation of state is
(3.5)
where and the constants A1 through A11 are the following: A1 =
0.3265; A2 = -1.0700; A3 = -0.5339; A4 = 0.01569; A5 = -0.05165; A6 = 0.5475; A7 = -
0.7361; A8 = 0.1844; A9 = 0.1056; A10 = 0.6134; and A11 = 0.7210 (Towler 2006). This is
an iterative equation that was solved with the various reduced temperatures and pressures
for the pressure range given above. This hydrocarbon relationship was validated for use
with air by examining values typically seen with air. Excel was used to calculate the gas-
deviation factors and the results are shown in Table 3.3.
The and values shown in Table 3.3 are within the ranges specified for better
accuracy when applying the Dranchuk and Abou-Kassem equation of state. Therefore,
this equation provides more accurate gas-deviation factors.
64
Table 3.3 Gas-deviation factor values for various pressures
Pressure Temp Tr Pr ρr Z factorpsia Rankine 14.7 520 2.1824 0.0269 0.0033 0.9993
264.7 520 2.1824 0.4840 0.0606 0.9881514.7 540 2.2663 0.9411 0.1142 0.9822764.7 550 2.3083 1.3982 0.1671 0.9786
1,014.7 560 2.3503 1.8554 0.2180 0.97791,264.7 570 2.3922 2.3125 0.2664 0.97991,514.7 580 2.4342 2.7696 0.3121 0.98421,764.7 590 2.4762 3.2267 0.3552 0.99052,014.7 600 2.5182 3.6839 0.3956 0.99852,264.7 610 2.5601 4.1410 0.4333 1.00792,514.7 620 2.6021 4.5981 0.4685 1.01832,764.7 630 2.6441 5.0552 0.5014 1.02963,014.7 640 2.6860 5.5123 0.5320 1.04163,264.7 650 2.7280 5.9695 0.5605 1.05413,514.7 660 2.7700 6.4266 0.5871 1.06703,764.7 670 2.8119 6.8837 0.6119 1.08014,014.7 680 2.8539 7.3408 0.6352 1.0934
In order to obtain the temperature profile, values were read from Figure 3.7 that
corresponded with the appropriate pressure values. This gives an approximation of the
temperature response with a change in pressure; Figure 3.7 was the only temperature
profile available in the supplied data (Crotogino 2006). With the gas-deviation factor
calculated, Equation 3.2 was used to find Bg at the same range of pressures and
corresponding temperatures (Table 3.4).
Table 3.4 Gas Formation Volume Factors for various pressures
65
Pressure Temp Z factor Bgpsia Rankine Rbbl/Mscf14.7 520 0.9993 178.0439
264.7 520 0.9881 9.7766514.7 540 0.9822 5.1901764.7 550 0.9786 3.5449
1,014.7 560 0.9779 2.71831,264.7 570 0.9799 2.22451,514.7 580 0.9842 1.89821,764.7 590 0.9905 1.66802,014.7 600 0.9985 1.49782,264.7 610 1.0079 1.36732,514.7 620 1.0183 1.26452,764.7 630 1.0296 1.18173,014.7 640 1.0416 1.11383,264.7 650 1.0541 1.05713,514.7 660 1.0670 1.00923,764.7 670 1.0801 0.96824,014.7 680 1.0934 0.9328
Another necessary PVT parameter is the viscosity as a function of pressure. In
order to calculate viscosities, the Lee et al. analytical method (Towler 2006) was used
within an Excel spreadsheet. This method requires pressure, temperature, z factor, and
molecular weight (the molecular weight of air is 28.964 lb-mole). The Lee et al.
equations are designed for use with specified units; these units are listed with the
equations below:
(3.6)
66
where , ,
, and and where µg = gas viscosity, cp; ρ = gas
density, g/cm3; p = pressure, psia; T = temperature, ºR; and Mg = gas molecular weight.
The Lee et al. analytical method has an average standard deviation of 2.7%, with a
maximum standard deviation of 9%. The distribution of variables used in developing the
method is as follows: 100 psia < 8,000 psia, 100 < T(F) <340, 0.90 < CO2 (mol%) < 3.20
and 0.0 < N2 (mol%) < 4.80. Table 3.5 displays the results from the Lee et al. equations.
Table 3.5 Lee et al. viscosity calculations
Pressure Temp Z factor ρ K1 X Y µpsia Rankine g/cc cp14.7 520 0.9993 0.0012 0.0092 5.6858 1.2628 0.0093264.7 520 0.9881 0.0223 0.0092 5.6858 1.2628 0.0097514.7 540 0.9822 0.0420 0.0096 5.6156 1.2769 0.0106764.7 550 0.9786 0.0615 0.0098 5.5824 1.2835 0.0115
1,014.7 560 0.9779 0.0802 0.0100 5.5504 1.2899 0.01241,264.7 570 0.9799 0.0980 0.0102 5.5195 1.2961 0.01341,514.7 580 0.9842 0.1148 0.0104 5.4896 1.3021 0.01441,764.7 590 0.9905 0.1307 0.0106 5.4608 1.3078 0.01552,014.7 600 0.9985 0.1455 0.0108 5.4330 1.3134 0.01662,264.7 610 1.0079 0.1594 0.0110 5.4060 1.3188 0.01772,514.7 620 1.0183 0.1724 0.0112 5.3800 1.3240 0.01892,764.7 630 1.0296 0.1844 0.0114 5.3547 1.3291 0.02003,014.7 640 1.0416 0.1957 0.0115 5.3303 1.3339 0.02113,264.7 650 1.0541 0.2062 0.0117 5.3066 1.3387 0.02233,514.7 660 1.0670 0.2160 0.0119 5.2836 1.3433 0.02343,764.7 670 1.0801 0.2251 0.0121 5.2613 1.3477 0.02454,014.7 680 1.0934 0.2337 0.0123 5.2396 1.3521 0.0256The viscosity values complete the inputs for the air PVT section in ECLIPSE
100. The rock compressibility of the base model is 3*10-6 psia-1 and average densities of
water and air are 1,000 kg/m3 (62.4 lb/ft3) and 1.297 kg/m3 (0.081 lb/ft3), respectively.
67
Measurements of salt contamination were conducted at the Huntorf facility for two
withdrawl cycles of 365 kg/s (805 lb/s). The results of each test showed salt content to be
less than 1 mg (salt) / kg (air). Therefore, fresh water density values described above are
used for the model input. Table 3.6 shows a diurnal schedule for the compression and
expansion of air.
Table 3.6 Daily schedule for the Huntorf facility
Flow Type Time (hrs)nothing 0nothing 1injection 2injection 3injection 4nothing 5nothing 6injection 7injection 8injection 9injection 10nothing 11
production 12production 13
nothing 14nothing 14.5injection 15.5injection 16.5injection 17
production 18production 19production 20injection 21injection 22nothing 23nothing 24
Table 3.6 was created by looking at the compression and expansion schedule given in
Figure 3.2. The Cavern CAES base model makes use of four wells, consisting of a
production and injection well for each cavern; the production wells are labeled as P-1 and
P-2 and the injection wells are I-1 and I-2 for caverns 1 and 2.
68
The production and injection rates for each well could not be entered into
ECLIPSE 100© as mass flow rates; therefore an air density had to be selected to convert
the mass flow rates into volumetric flow rates. A temperature of 15 ºC (59 ºF) and an
atmospheric pressure were chosen as the near surface conditions, yielding an air density
of 1.225 kg/m3 (0.0765 lbm/ft3). This density differs from the previously stated density
because the density of 1.225 kg/m3 (0.0765 lbm/ft3) represents near surface conditions
and the previous density of 1.297 kg/m3 (0.081 lbm/ft3) is an average density. Using the
density and mass flow rates of 417 kg/s (919.4 lb/s) and 108 kg/s (238 lb/s) for
production and injection, respectively, volumetric rates were obtained. For input into
ECLIPSE 100©, the rates were converted into units of Mscf/day; these rates were then
divided by two to compensate for an injection and production well per cavern. The total
production rate for the system is 1,038,649 Mscf/day and 519,324.5 Mscf/day for each
cavern production well. The total injection rate is 269,003 Mscf/day and 134,501.5
Mscf/day for each cavern injection well. The model is comprised mostly of one hour
time steps in which the system is injecting, producing, or remaining static. This set up
allows for a direct comparison to actual Huntorf data.
Table 3.7 Conversion factors for oilfield units to metric units
Oilfield Unit Conversion Factor Metric Unitbbl x 1.589873 E-01 m3
ft x 3.048 mgal x 3.785413 m3
lbm x 4.535924 E-01 kgpsia x 1.751268 E+02 Pa
deg F x (deg. F-32)/1.8 deg. CMscf/day x 2.832 E+02 m3/day
Since ECLIPSE 100© was set up to perform model runs in oilfield units, the
remainder of this chapter will only show oilfield units. Table 3.7 shows the conversion
factors for oilfield units to metric units for the applicable parameters.
69
3.3 Cavern CAES Sensitivities and Results
With all of the necessary parameters entered into an ECLIPSE 100© dataset
format, the base case model was run. The results from the base model were compared
with the actual Huntorf pressure data. A simple change to permeability in the model also
allowed for the first model sensitivity of different permeabilities. The initial pressure
match is displayed below in Figure 3.8.
0
100
200
300
400
500
600
700
800
900
1000
0 5 10 15 20 25 30
Time (hrs)
Pres
sure
(psi
a)
Actual Pressure
Modeled Pressure w/ 1,000 md perm
Modeled Pressure w/ 10,000 md perm
Modeled Pressure w/ 20,000 md perm
Figure 3.8 Comparison of actual and modeled data for initial model runsFigure 3.8 shows that during the initial stages of injection and remaining static, all
of the model runs are close to the actual values. However, once production begins during
the 12th hour, all of the values begin to deviate from the actual pressures. The same
general shape of the models mimics the actual pressure curve, but with the onset of
production, the modeled pressures fall below the actual values. This can especially be
70
seen in the base model with the 10,000 md permeability and in the 20,000 md
permeability model. The 1,000 md permeability model has a little better pressure
response; however, a 1,000 md permeability is not an accurate representation of cavern
geology. The pressure curve of the 1,000 md permeability model does show that
permeability has an effect on the model with low values. With the higher permeability
models, the pressure curves are almost exactly the same. This is encouraging because it
demonstrates that once the permeability gets to a large enough value, it does not affect
model performance, as should be the case with a cavern environment. However, since
this is a cavern, the 1,000 md permeability should provide similar results, which is not the
case in Figure 3.8.
In addition to the pressure match, the specified rates of 1,038,649 Mscf/day
(519,324.5 Mscf/day per cavern) for production and 269,003 Mscf/day (134,501.5
Mscf/day per cavern) for injection needed to be replicated in the appropriate time step
according to the schedule of Table 3.6. All of the model runs with the different
permeabilities achieved the same injection rate profile as desired; however, none of the
runs could simulate the specified production rates. In order to achieve these rates and
hopefully get a more accurate pressure response, model runs with different pore volumes
were conducted. Using a permeability of 10,000 md, pore volume was adjusted to 50-
90%, 110%, 150 %, and 200% of the original pore volume values of 300,000 m3
(1,0233,600 ft3). These cavern bulk volumes explained in Section 3.2 are also the pore
volumes because of a 100% porosity in a cavern environment. Figure 3.9 shows the
pressure response for varying pore volumes.
71
0
100
200
300
400
500
600
700
800
900
1000
0 5 10 15 20 25 30
Time (hrs)
Pres
sure
(psi
a)Original Pore VolumeActual Values90% Pore Volume80% Pore Volume110% Pore Volume150% Pore Volume200% Pore Volume
Figure 3.9 Pressure Match with Varying Pore Volumes and 10,000 md Permeability
Figure 3.9 displays promising results for the lower pore volumes at early time
periods, but these lower pore volumes have the worst match at later time periods. The
opposite trend is true for the larger pore volumes; at early time periods the match is poor,
but at late time periods the match becomes closer. The 50% and 60% values are not
included because the later trends of the prior decreasing pore volumes demonstrate the
undesired response. A change in pore volume created better matches, but these occurred
with different pore volume percentages based on the selected time period. The injection
rate with different pore volumes was as desired (134,501.5 for both wells P-1 and P-2) as
shown in Figure 3.10. Figure 3.11 and Figure 3.12 display the hard to achieve production
rates for wells P-1 and P-2, respectively.
72
0
20000
40000
60000
80000
100000
120000
140000
160000
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 13.0 14.0 14.5 15.5 16.5 17.0 18.0 19.0 20.0 21.0 22.0 23.0 24.0
Time (hrs)
Msc
f/d
Figure 3.10 The injection rate for actual data and modeled data for different pore volumes
0
100000
200000
300000
400000
500000
600000
0 5 10 15 20 25 30
Time (hrs)
Msc
f/d
Actual RateOriginal Pore Volume90% Pore Volume80% Pore Volume70% Pore Volume60% Pore Volume50% Pore Volume
Figure 3.11 The production rate for decreased pore volumes for well P-1
73
0
100000
200000
300000
400000
500000
600000
0 5 10 15 20 25 30
Time (hrs)
Msc
f/dActual RatesOriginal Pore Volumes90% Pore Volume80% Pore Volume70% Pore Volume60% Pore Volume50% Pore Volume
Figure 3.12 The production rate for decreased pore volumes for well P-2
Only decreased pore volumes were used to achieve the production rates because
an increase in pore volume leads to a lower pressure and thus a lower production rate.
The results from this figure are inconclusive. With well P-1, a 50% pore volume could
initially achieve 519,324.5 Mscf/day, but was then one of the worst performers. A 90%
pore volume started poorly, but was better towards the end of the schedule. Based on the
results from the pressure match and the production rates, a change in pore volume is not
an appropriate method for achieving the desired pressures and rates.
The next sensitivity to be modeled required a personal communication with F.
Crotogino. The ability to continuously achieve such high production rights seemed
questionable. F. Crotogino confirmed that these rates are maximum achievable rates and
typically the plant did not operate with such high production rates. For this reason, the
next sensitivity to be modeled was a dataset with lower production rates. Different
production rates were modeled until the best pressure match was found with a
74
corresponding ability to achieve the specified rate during the given time period. The
rates that met the above criteria were 500,000 Mscf/d (250,000 Mscf/d per production
well) for the initial production period during the 12th and 13th hours and 600,000 Mscf/d
(300,000 Mscf/d per production well) for the final production period during the 18th – 20th
hours. These production rates were achieved in model runs with 1000, 10,000, and
20,000 md permeability. The pressure match with the lower production rates and the
above permeabilities is shown in Figure 3.13.
0
100
200
300
400
500
600
700
800
900
1000
0 5 10 15 20 25 30
Time (hrs)
Pres
sure
(psi
a)
Actual Pressure
Modeled Pressure w/ 1,000 md perm
Modeled Pressure w/ 10,000 md perm
Modeled Pressure w/ 20,000 md perm
Figure 3.13 The pressure match obtained with lower production rates and the original pore volumes
The pressure match obtained with lower production rates is quite an improvement
over the other scenarios and, more importantly, the inputted production rates could be
achieved during modeling. To verify the quality of the pressure match, Figure 3.14 and
75
Figure 3.15 provide a change in pressure (actual – modeled) relationship and a percent
difference ((actual – modeled) / actual) response. Since the pressure response between
the different permeabilities is minimal and 10,000 md permeability was selected for the
cavern permeability, only the 10,000 md permeability is examined in Figure 3.14 and
Figure 3.15.
-5
0
5
10
15
20
25
30
35
0 5 10 15 20 25
Time (hrs)
Pres
sure
(psi
a)
Figure 3.14 Change in pressure between modeled and actual values for lower production rates and 10,000 md permeability
76
-1
-0.5
0
0.5
1
1.5
2
2.5
3
3.5
4
0 5 10 15 20 25
Time (hrs)
Perc
ent D
iffer
ence
Figure 3.15 Percent difference of actual and modeled pressure values for lower production rates and 10,000 md permeability
Overall, the change in pressure and percent difference are quite small; the only
major discrepancy occurs during the second injection period between the 7th and 9th hours
of the schedule. The modeled pressure follows the same trend, but cannot obtain as high
as values as the actual pressures. For this reason, a final sensitivity with changing
injection and productions rates was examined. Different injection and production rates
were used until the best pressure match could be obtained while still injecting and
producing the specified rates. After a number of model runs, the rates in Table 3.8 were
selected.
77
Table 3.8 Schedule of injection and production rates with the best pressure match
Single Cavern Flow Rate
Total Flow Rate
Flow Type
Mscf/d Mscf/dnothing 0 0nothing 0 0injection 160,000 320,000injection 160,000 320,000injection 160,000 320,000nothing 0 0nothing 0 0injection 160,000 320,000injection 160,000 320,000injection 160,000 320,000injection 160,000 320,000nothing 0 0
production 350,000 700,000production 350,000 700,000
nothing 0 0nothing 0 0injection 160,000 320,000injection 160,000 320,000injection 160,000 320,000
production 300,000 600,000production 300,000 600,000production 300,000 600,000injection 134,501.5 269,003injection 134,501.5 269,003nothing 0 0nothing 0 0
By increasing the rate for each cavern injection well from 134,501.5 Mscf/d to
160,000 Mscf/d for all injection periods except the final period, the production rate also
had to be increased from the values established in the previous sensitivity. Using these
new rates, an ECLIPSE 100© model run was conducted; the pressure match with this
sensitivity is shown in Figure 3.16.
78
0
100
200
300
400
500
600
700
800
900
1000
0 5 10 15 20 25 30
Time (hrs)
Pres
sure
(psi
a)Actual Pressure
Modeled Pressure w/ 1,000 md perm
Modeled Pressure w/ 10,000 md perm
Modeled Pressure w/ 20,000 md perm
Figure 3.16 Pressure match with changing production and injection rates and different permeabilities
With an increase in injection rates and a decrease in production rates, a better pressure
match could be realized. Figure 3.17 and Figure 3.18 further exemplify this point.
Figure 3.17 shows a change in actual and modeled pressures and Figure 3.18 displays the
percent difference between actual and modeled pressures; both graphs depict 10,000 md
permeability. The spike in change in pressure and percent difference that is seen around
the 10th hour has been eradicated by the increase in injection rates; the change also
produced lower overall values for these parameters.
79
-10
-5
0
5
10
15
20
0 5 10 15 20 25
Time (hrs)
Pres
sure
(psi
a)
Figure 3.17 Change in actual and modeled pressure for changing injection and production rates with 10,000 md permeability
-1
-0.5
0
0.5
1
1.5
2
2.5
0 5 10 15 20 25
Time (hrs)
Perc
ent D
iffer
ence
Figure 3.18 Percent difference in actual and modeled pressure for changing injection and production rates with 10,000 md permeability
80
3.4 Objective Functions
In order to compare the pressure match obtained with the different sensitivities
several objective functions were calculated. An objective function is “a function
associated with an optimization problem which determines how good a solution is”
(Black 2004). The lower the objective function, the better the solution. Objective
functions (OF) 1 – 6 are displayed below in Equations 3.7 – 3.12.
(3.7)
(3.8)
(3.9)
(3.10)
(3.11)
(3.12)
These objective functions were calculated for the Cavern CAES base model and all of the
sensitivities described in Section 3.3. Following these calculations, OF5 and OF6 were
chosen for comparative purposes because of their integration of change in pressure and
the standard deviation of the actual pressures. By including both of these parameters, a
more informative objective function is created. Table 3.9 shows OF5 and OF6 for all of
the scenarios.
Table 3.9 Objective functions for Cavern CAES base model and the described sensitivities
81
Case Description OF5 OF6
Base Original Pore Volumes, Original Rates, and 10,000 md permeability 96.9727 9.8475Sensitivity
A Original Pore Volumes, Original Rates, and different permeabilities A.1 1,000 md permeability 37.5114 6.1247A.2 20,000 md permeability 96.9777 9.8477
Sensitivity B Alternate Pore Volumes, Original Rates, and 10,000 md permeability
B.1 90% of Original Pore Volume 110.5139 10.5126B.2 80% of Original Pore Volume 130.6064 11.4283B.3 70% of Original Pore Volume 161.5500 12.7102B.4 110% of Original Pore Volume 87.5634 9.3575B.5 120% of Original Pore Volume 80.8715 8.9929B.6 150% of Original Pore Volume 69.8266 8.3562B.7 200% of Original Pore Volume 63.8099 7.9881
Sensitivity C
Original Pore Volumes, Different Production Rates, and 10,000 md permeability 2.4219 1.5562
Sensitivity D
Original Pore Volumes, Different Production Rates, and 1,000 md permeability 2.4068 1.5514
Sensitivity E
Original Pore Volumes, Different Production Rates, and 20,000 md permeability 2.4220 1.5563
Sensitivity F
Original Pore Volumes, Different Injection and Production Rates, and 10,000 md permeability 0.4053 0.6366
Sensitivity G
Original Pore Volumes, Different Injection and Production Rates, and 1,000 md permeability 0.3996 0.6321
Sensitivity H
Original Pore Volumes, Different Injection and Production Rates, and 20,000 md permeability 0.4055 0.6368
The base case and sensitivities A and B all have the largest objective functions,
thus verifying the previously stated results. These objective functions prove that a
change in pore volume does not improve the quality of the pressure match. The larger
pore volumes do decrease the objective functions. However, Figure 3.9 shows that
during later time periods the match with these larger volumes becomes poorer and poorer.
The objective functions for the sensitivities with the original pore volumes and a change
in rate confirm a more direct correspondence. Sensitivities C, D, and E all have OF5
values around 2.4 and OF6 values around 1.55. This similarity is desired because it
shows that pressure is not a function of permeability in a cavern setting. With a change
82
in injection and production rates, even smaller objective functions are realized.
Sensitivities F, G, and H have OF5 values around 0.40 and OF6 values around 0.63. The
indirect relationship between pressure responses with a change in permeability can also
be seen with these sensitivities.
3.5 Discussion of Cavern CAES Models
Based solely on the objective functions and the desire to use a permeability of
10,000 md, Sensitivity F would be the obvious choice. However, a personal
communication with Crotogino acknowledged that there is no evidence to justify the
increase in the injection rate. Crotogino stated that the rates provided in Crotogino et al.
(2001) are maximum obtainable rates. Although this sensitivity provides the best
pressure match, an increase in injection rates cannot be justified. Crotogino did confirm
that a decrease in production rates is reasonable because the turbine can not consistently
operate at its maximum rate. In fact, the maximum rate of 1,038,649 Mscf/day can rarely
be obtained in Huntorf’s daily operation. For this reason, Sensitivity C is the best
representation of the diurnal schedule and pressure response for the Huntorf facility.
With OF5 and OF6 values of 2.4219 and 1.5562, respectively, Sensitivity C is still a very
reasonable choice.
The quality of the acquired pressure match and the ability to inject and produce
the desired rates validates the use of ECLIPSE 100© in the modeling of CAES. A widely
used reservoir simulator in the oil industry can also be applied in a storage sense. This
realization could help diminish the gap between traditional energy sources and a
sustainable energy option for the future.
83
CHAPTER 4
VERIFICATION OF MODEL USE FOR RESERVOIR STORAGE
In order to explore the possibility of using CAES in porous media, an ECLIPSE
100© dataset was constructed depicting a shallow reservoir. This reservoir could be a
depleted hydrocarbon reservoir or an excavated aquifer. The composition of the reservoir
is a sandstone with 20% porosity or higher. Sandstone was chosen because of its
availability and the high porosities and permeabilities that can be seen with this structure.
This chapter will detail the creation of an EZGEN grid for input into ECLIPSE 100©, the
ECLIPSE 100© dataset itself, and the results from numerous runs of the model. The
reservoir model uses a 20% porosity and includes runs with varying permeabilities. Just
as with the Huntorf model, there is no residual water saturation in the reservoir. Results
form the varying permeability models are compared and discussed.
4.1 EZGEN grid input and Model Setup
The reservoir ECLIPSE 100© model uses the same rates as Sensitivity C in the
cavern CAES model, as well as, two production wells, P-1 and P-2, and two injection
wells, I-1 and I-2, just as in the Cavern CAES model. A reservoir model with the same
injection and production rates as Sensitivity C and the same number of wells provides a
sound comparison for the use of CAES in a cavern setting versus a reservoir setting. By
keeping the rates and wells unchanged, it can be determined if the same CAES process
84
that is successful in Huntorf could be effectively transferred to a porous medium. Table
4.1 shows the rates for each well that was used in Sensitivity C and will be used for the
reservoir model.
Table 4.1 Injection and production rates for each well used in Sensitivity C and for Reservoir CAES
Injection Rates Production Rates Time I1 Rate I2 Rate P1 Rate P2 Ratehrs Mscf/d Mscf/d Mscf/d Mscf/d0.0 0 0 0 01.0 0 0 0 02.0 134,501.5 134,501.5 0 03.0 134,501.5 134,501.5 0 04.0 134,501.5 134,501.5 0 05.0 0 0 0 06.0 0 0 0 07.0 134,501.5 134,501.5 0 08.0 134,501.5 134,501.5 0 09.0 134,501.5 134,501.5 0 0
10.0 134,501.5 134,501.5 0 011.0 0 0 0 012.0 0 0 250,000 250,00013.0 0 0 250,000 250,00014.0 0 0 0 014.5 0 0 0 015.5 134,501.5 134,501.5 0 016.5 134,501.5 134,501.5 0 017.0 134,501.5 134,501.5 0 018.0 0 0 300,000 300,00019.0 0 0 300,000 300,00020.0 0 0 300,000 300,00021.0 134,501.5 134,501.5 0 022.0 134,501.5 134,501.5 0 023.0 0 0 0 024.0 0 0 0 0
Before building any of the reservoir models in ECLIPSE 100©, a typical reservoir
structure had to be devised. EZGEN was used to construct a dipping anticline, ideal for
hydrocarbon reservoirs, and to avoid the use of a layer-cake model. EZGEN develops
85
grids, distributes reservoir parameters throughout the grid blocks, calculates
petrophysical parameters, and constructs relative permeability tables (Fanchi 2002). For
this reservoir model, the same petrophysical parameters and relative permeability tables
were used as in the Cavern CAES models.
In order to make a comparable reservoir model and cavern model, the volumes of
the two models are directly related. The volume of the Huntorf salt caverns is both the
pore volume and the bulk volume because of its 100% porosity. Therefore when a
reservoir structure with 20% porosity is considered, the Huntorf volume is the effective
pore volume and the bulk volume can be calculated with Equation 4.1,
(4.1)
where = porosity, VP = pore volume, and VB = bulk volume. Using the Huntorf pore
volume of 300,000 m3 (10,594,400 ft3) and a 20% porosity, a reservoir bulk volume of
1,500,000 m3 (52,972,000 ft3) was calculated. Since EZGEN and ECLIPSE 100© utilize
oilfield units (English units), the remainder of this chapter will present information in
these units, refer to Table 3.7 for the conversion factors. A gross thickness of 125 ft and
a net thickness of 100 ft are assumed for the reservoir model with equal distances of 720
ft in the x and y directions that yield an approximate reservoir bulk volume defined
above. The anticlinal crest is 1,500 ft from surface and the deepest point of the reservoir
is 1,775 ft. These parameters allow the anticline to have a dip of 12º. With these values
established, a number of control points were set. A total of 25 control points were used;
each control point contains an x, y, and z dimension, a gross thickness, a net thickness, a
permeability, and a porosity. Based on these 25 control points, EZGEN could then create
a reservoir structure and an ECLIPSE 100© input model for the initial keywords. Using
3DView, the reservoir can be viewed three-dimensionally; Figure 4.1 shows the anticlinal
reservoir.
86
Figure 4.1 Reservoir structure created with EZGEN for use in ECLIPSE 100© base reservoir model
For the initial reservoir model, a 100 md permeability was used. Different
EZGEN files were written for 1,000 md and 10,000 md permeability. The EZGEN
datasets for the 100 md, 1,000 md, and 10,000 md models can be found in Appendix B.1,
B.2, and B.3, respectively. The order of parameters given above for each control point is
the same order for the control points in the Appendices. The output from the EZGEN
datasets includes the gridding of the reservoir and the distribution of reservoir properties.
The EZGEN outputs for the 100 md, 1,000 md, and 10,000 md models can be found in
Appendix B.4, B.5, and B.6, respectively. These output files can be directly inputted into
an ECLIPSE 100© dataset. As previously mentioned, the injection and compression
87
schedule is the same as the one used in Sensitivity C. The water and gas PVT properties
and relative permeabilities are also the same for the reservoir model. This assumes that
any fluids previously occupying the reservoir have been removed. The addition of the
EZGEN output file alters the porosity, permeability, grid dimensions and the distribution
of the petrophysical properties throughout the grid. The original ECLIPSE 100© input
dataset for the three different permeability models is located in Appendix B.7. There is
only one ECLIPSE 100© dataset for all three permeability models. This is because the
only difference between the models is the EZGEN output file. This unique output file is
inserted into the ECLIPSE 100© dataset with an include statement.
4.2 Reservoir CAES Sensitivities and Results
With the 100 md EZGEN file inputted into the reservoir model, an ECLIPSE
100© run was performed. Just as in the cavern modeling, all of the injected air was
realized in the initial reservoir model run; therefore, the injection rate schedule is the
same as in Figure 3.8. The same was not true for the production rates. These rates could
not be achieved during the initial model run with the original pore volumes and 100 md
permeability. Therefore, additional model runs with an increase in the pore volume were
necessary. The pressure response and the rate schedule for Well P-1 and Well P-2 for the
original and increased pore volumes can be seen in Figure 4.2, Figure 4.3, and Figure 4.4,
respectively.
88
0
100
200
300
400
500
600
700
800
900
1000
0 5 10 15 20 25 30
Time (hrs)
Pres
sure
(psi
a)Original Pore Volume
5x Pore Volume
100x Pore Volume
500x Pore Volume
1,000x Pore Volume
10,000x Pore Volume
Figure 4.2 Pressure Response for model runs with varying pore volume and 100 md permeability in a reservoir setting
0
50000
100000
150000
200000
250000
300000
350000
0 5 10 15 20 25 30
Time (hrs)
Msc
f/d
Actual RatesOriginal Pore Volume5x Pore Volume100x Pore Volume500x Pore Volume1,000x Pore Volume10,000x Pore Volume100,000x Pore Volume
Figure 4.3 Production rates with varying pore volumes and 100 md permeability for Well P-1 in a reservoir setting
89
0
50000
100000
150000
200000
250000
300000
350000
0 5 10 15 20 25 30
Time (hrs)
Msc
f/dActual RatesOriginal Pore Volumes5x Pore Volume100x Pore Volume500x Pore Volume1,000x Pore Volume10,000 Pore Volume100,000x Pore Volume
Figure 4.4 Production rates with varying pore volumes and 100 md permeability for Well P-2 in a reservoir setting
In Figure 4.2, a change in pressure can be seen with the original pore volume and
the smaller increases in pore volume. However, as the increase in the pore volume
intensifies, the pressure begins to remain almost constant throughout the series of
compression and expansion. This is due to the fact that a pressure response can not be
expected with such a large pore volume. These large increases in pore volume were
necessary to try and obtain the desired production rates. However, even with an increase
of 100,000 times the original pore volume, the production rates could not be achieved.
During the initial production period, both wells under the 100,000 times the original pore
volume constraint could produce 250,000 Mscf/day, but neither well could obtain the
300,000 Mscf/day production rate during the final production stage. Figure 4.5 illustrates
the percent difference of the different pore volumes for the 100 md model.
90
0
10
20
30
40
50
60
70
80
0 5 10 15 20 25 30
Time (hrs)
Perc
ent D
iffer
ence
Original Pore Volumes5x Pore Volume100x Pore Volume500x Pore Volume1,000x Pore Volume10,000 Pore Volume100,000x Pore Volume
Figure 4.5 Percent difference of the total production rate for 100 md permeability with varying pore volumes
Figure 4.5 confirms the fact that an increase in pore volume does help the
production rates become closer to the Huntorf rates. However, even with the 100,000
times increase in pore volume, the second period of production has a 17% difference.
This lack of ability to attain the specified production rates even with an unrealistically
large pore volume, leads to the conclusion that the injection and production schedule used
at Huntorf can not be reproduced in a reservoir with 100 md permeability. Therefore
model runs with larger permeability were deemed necessary.
The 1,000 md EZGEN input file was then entered into ECLIPSE 100©, using the
original pore volumes, and an initial model run was conducted. The results obtained
from this model run are much better than the results produced with the 100 md model.
As with the previous models, the 1,000 md model had no problems matching the
injection rate schedule. However, with the 1,000 md model, the production rate schedule
91
was almost matched perfectly. With the original pore volumes, Well P-2 obtained the
desired production rates and Well P-1 could match the rate schedule except for the last
hour of production. This is because Well P-2 is updip of Well P-1 in the ECLIPSE 100©
input dataset. The pore volume had to be increased three times in order for Well P-1 to
match the pressure rate of 300,000 Mscf/day in the last hour of production. Figure 4.6
displays the pressure response of the 1,000 md model and Figure 4.7 show the production
rates for the 1000 md model with varying pore volumes for Well P-1. Well P-2 obtained
the production rates with the original pore volumes, so a graph of its rates is unnecessary.
0
100
200
300
400
500
600
700
800
900
1000
0.0 5.0 10.0 15.0 20.0 25.0 30.0
Time (hrs)
Pres
sure
(psi
a)
Original Pore Volume
2x Pore Volume
3x Pore Volume
Figure 4.6 Pressure response for the 1,000 md model with varying pore volumes in a reservoir setting
92
0
50000
100000
150000
200000
250000
300000
350000
0 5 10 15 20 25 30
Time (hrs)
Msc
f/dActual Rate
Original Pore Volume
2x Pore Volume
3x Pore Volume
Figure 4.7 Production rates of model runs with 1,000 md permeability and varying pore volumes for Well P-1 in a reservoir setting
The pressure response with the original pore volumes of the 1,000 md
permeability model is similar to the one produced with the 100 md model. However, the
original pore volume response of the 1,000 md model is not as pronounced as it is with
the 100 md model. Since only an increase of three times the pore volume was necessary
to obtain the production rates, some pressure response can still be seen when the desired
rates are achieved. This amount of increase in pore volume is much more reasonable
than the very large increase in the 100 md permeability model, which couldn’t even
reproduce the desired production schedule. Even with the original pore volumes of the
1,000 md permeability model, all of the rates are matched except for the last production
hour of Well P-2. This is a much more promising rate response than the initial 100 md
model. However, one final increase in permeability was needed to produce an exact
93
match in the actual Huntorf rate response and the modeled rate response with the original
pore volume.
For the next simulation, the 10,000 md EZGEN input file was inputted into the
ECLIPSE 100© model and an initial model run was performed. As expected, the results
from this increase in permeability were even better than the 1,000 md model. The
injection rates were once again obtained and unlike the previous reservoir models, a
match in the production rates was produced without an increase in pore volume. Figure
4.8 contains the pressure response for the 10,000 md model.
0
100
200
300
400
500
600
700
0 5 10 15 20 25
Time (hrs)
Pres
sure
(psi
)
Figure 4.8 Pressure response with original pore volume and 10,000 md permeability in a reservoir setting
94
The pressure response of the 10,000 md model is very similar to that of the 1,000 md
model and therefore a less significant response is seen in this model when compared to
the 100 md model.
4.3 Reservoir CAES Model Comparison
Various graphs have been constructed to offer a comparison between the model
runs with different permeability. Figure 4.9 illustrates the pressure response of the three
permeabilities with the original pore volume.
0
100
200
300
400
500
600
700
800
900
1000
0.0 5.0 10.0 15.0 20.0 25.0 30.0
Time (hrs)
Pres
sure
(psi
a)
100 md perm
1,000 md perm
10,000 md perm
Figure 4.9 Pressure response of the three permeabilities with the original pore volumes in a reservoir setting
95
The 100 md permeability has the same response as the other two permeabilities
up until the 12th hour of the production and injection schedule. This is the beginning of
the first section of production. The lower permeability model does not experience as
much pressure loss as the higher permeability models due to the fact that more production
can occur with the higher permeability models. This higher production caused a higher
pressure loss than with the lower production of the 100 md model. This point is further
exemplified with the comparison of production rates of the different permeabilities with
their original pore volumes. Figure 4.10 and Figure 4.11 show this comparison for Well
P-1 and Well P-2.
0
50000
100000
150000
200000
250000
300000
350000
0 5 10 15 20 25 30
Time (hrs)
Msc
f/d
Actual Rate100 md perm1,000 md perm10,000 md perm
Figure 4.10 Production rate comparison for Well P-1 for the three permeabilities with the original pore volumes in a reservoir setting
96
0
50000
100000
150000
200000
250000
300000
350000
0 5 10 15 20 25 30
Time (hrs)
Msc
f/dActual Rates
100 md perm
1,000 md perm
10,000 md perm
Figure 4.11 Production rate comparison for Well P-2 for the three permeabilities with the original pore volumes in a reservoir setting
Because of its low permeability the 100 md model is not even close to achieving
the specified rates. However, if the permeability is increased by a factor of ten, then the
rates are much more obtainable. This small rate of the 100 md model helps demonstrate
why the pressure loss is not as significant as it is with the 1,000 md and 10,000 md
models. For further comparison, percent difference was calculated for the 100 md and
1,000 md models (the 10,000 md model has 0% difference), which is shown in Figure
4.12.
97
0
10
20
30
40
50
60
70
80
0 5 10 15 20 25 30
Time (hrs)
Perc
ent D
iffer
ence
100 md perm
1,000 md perm
Figure 4.12 Percent difference of the total production rate for 100 md and 1,000 md permeability with the original pore volume
The 1,000 md model has 0% difference during the first phase of production and
the first two hours of the last production stage. Its last hour of production couldn’t quite
obtain 300,000 Mscf/day and therefore has a 8% difference. The percent difference of
the 100 md model is quite different. Its percent difference values range from 43% during
the first hour of production to 75% during the last hour of production.
The same kind of analysis was conducted for the remaining corresponding pore
volumes. Since the 10,000 md model achieved the production rates with its original pore
volume, it’s excluded from the analysis. The 1,000 md model obtained the production
rates with three times the original pore volumes. Therefore the remaining analyses will
include the 100 md and 1,000 md model with two and three times the original pore
volume. Figure 4.13 contains the pressure response for two times the original pore
volume.
98
0
100
200
300
400
500
600
700
800
900
1000
0 5 10 15 20 25 30
Time (hrs)
Pres
sure
(psi
a)100 md perm
1,000 md perm
Figure 4.13 Pressure response of the 100 md and 1,000 md models with two times the original pore volume in a reservoir setting
Increasing the pore volume in both of the models diminishes the effect that the
production and injection cycle has on the pressure response. Since pressure and volume
have an inverse relationship, this response is expected. The 100 md model pressure
response still has higher overall pressure values than the 1,000 md model because not as
much of the air could be produced out of the system with the 100 md model. Figure 4.14
and Figure 4.15 display the production rates for two times the pore volume for Well P-1
and Well P-2, respectively.
99
0
50000
100000
150000
200000
250000
300000
350000
0 5 10 15 20 25 30
Time (hrs)
Msc
f/dActual Rate
100 md perm
1,000 md perm
Figure 4.14 Production rates for Well P-1 for 100 md and 1,000 md permeabilities with two times the original pore volume in a reservoir setting
0
50000
100000
150000
200000
250000
300000
350000
0 5 10 15 20 25 30
Time (hrs)
Msc
f/d
Actual Rates
100 md perm
1,000 md perm
Figure 4.15 Production rates for Well P-2 for 100 md and 1000 md permeabilities with two times the original pore volume in a reservoir setting
100
The above figures demonstrate that for Well P-1, the 1,000 md model achieves all
of the production rates, except during the last hour of production. The 100 md model
does not show much improvement from the modeling of the original pore volumes. For
Well P-2, the 1,000 md model matches all of the rates, while the 100 md model is still far
from the desired rates. Figure 4.16 shows the percent difference of the total production
rate for two times the pore volume.
0
10
20
30
40
50
60
70
80
0 5 10 15 20 25 30
Time (hrs)
Msc
f/d
100 md perm
1,000 md perm
Figure 4.16 Percent difference of the total production rate for 100 md and 1,000 md permeability with two times the original pore volume
An increase in the pore volume for the 1,000 md model decreases its nonzero
percent difference from 8% to 1%. The minimum and maximum percent difference for
the 100 md model decreases from 43% and 75 % to 40% and 72%. The 1,000 md model
shows better improvement because an increase in pore volume has a greater effect on
101
higher permeability reservoirs. A higher permeability allows for easier flow; with a
lower permeability reservoir an increase in pore volume helps, but is not enough to
greatly improve flow efficiency.
A final analysis was performed with three times the pore volume for the 100 md
and 1,000 md models. Figure 4.17 displays the pressure comparison of the two models.
0
100
200
300
400
500
600
700
800
900
1000
0 5 10 15 20 25 30
Time (hrs)
Pres
sure
(psi
a)
100 md perm
1,000 md perm
Figure 4.17 Pressure response of the 100 md and 1,000 md models with three times the original pore volume in a reservoir setting
With three times the original pore volume, the pressure response of the two models
begins to level. Additionally, the pressure of the models is very similar in the early
stages of injection, but once production begins, the 1,000 md model starts exhibiting
lower pressures than the 100 md model. Once again, this can be attributed to the inability
of the 100 md model to produce the desired rates. The production rates of the two
102
models are depicted in Figure 4.18 and Figure 4.19 for Well P-1 and Well P-2,
respectively.
0
50000
100000
150000
200000
250000
300000
350000
0 5 10 15 20 25 30
Time (hrs)
Msc
f/d
Actual Rate
100 md perm
1,000 md perm
Figure 4.18 Production rates for Well P-1 for 100 md and 1,000 md permeabilities with three times the original pore volume in a reservoir setting
At three times the original pore volume and 1,000 md permeability, the
production rate objectives have been satisfied in both Well P-1 and Well P-2. For the
100 md scenario, some improvement can be seen over the two times the original pore
volume rates, but a significant difference still exists between the Huntorf rates and the
100 md rates.
103
0
50000
100000
150000
200000
250000
300000
350000
0 5 10 15 20 25 30
Time (hrs)
Msc
f/dActual Rates
100 md perm
1,000 md perm
Figure 4.19 Production rates for Well P-2 for 100 md and 1,000 md permeabilities with three times the original pore volume in a reservoir setting
For the 100 md permeability model, the minimum and maximum percent difference
decreased to 39% and 70%. Since the 1,000 md model achieved the production rates at
three times the original pore volume, no additional comparisons can be made between the
1,000 md and 100 md models. Table 4.2 summarizes the percent difference findings
from the above analysis.
Table 4.2 Summary of maximum percent difference between actual production rates and modeled production rates for the three permeability models in a reservoir setting
Permeability Model Production Rate Max % DiffOriginal Pore Volume 2x Pore Volume 3x Pore Volume
100 md 75 72 701,000 md 8 1 0
10,000 md 0 0 0
104
4.4 Discussion of Reservoir CAES Models
A comparison of the cavern models to the reservoir models will first be examined
for repeatability confirmation, followed by a comparison of the different reservoir
models.
4.4.1 Comparison of Cavern Models and Reservoir Models
Comparing the pressure response of the reservoir models (Figure 4.9) to the
pressure response of the cavern models (Figure 3.13) reveals the same general pressure
trend between the corresponding permeabilities. The high and low pressure peaks occur
at the same point in time for the reservoir models and the cavern models. The overall
pressures are lower with the reservoir models; this is because the depths of the reservoir
models are less than those of the cavern models. ECLIPSE 100© was set up to record an
average reservoir pressure based on the recorded pressure at each gridblock. The average
pressure at each gridblock will be higher with the Cavern models because of the
increased ambient pore pressures, matrix pressures, and overburden pressures of a deeper
structure. In spite of the differences in depth, the expected pressure increase with
injection and pressure decrease with production can be seen in both models. This similar
pressure response validates the use of ECLIPSE 100© to model CAES in a cavern and
reservoir setting.
As far as the production and injection rates of the cavern and reservoir models,
the specified injection rate could be achieved in all of the models. With the rates of
Sensitivity C, these decreased production rates were emulated in all two corresponding
permeabilities (1,000 md and 10,000 md). Using these same rates in the reservoir model,
only the 10,000 md model could reproduce the rates with the original pore volume. In
order to get the desired production rates into the 1,000 md model, the pore volume had to
105
be increased three times. By adding porosity and a decreased permeability, the reservoir
could not handle the production rates with its original pore volume. This shows that if
lower permeabilities are to be used, lower rates need to be implemented. The formation
properties inhibit these higher rates from being produced from the reservoir.
4.4.2 Comparison of Reservoir Models
Using the same injection and production schedule and rates as Schedule C in the
Huntorf model, the 10,000 md model would be the best choice for implementing CAES
in a reservoir setting. However, this type of permeability is extremely rare, especially in
the continental U.S. For this reason, the 1,000 md model is the model of choice. Simply
using the original pore volume, the 1,000 md model could achieve all of the Huntorf
production rates with Well P-2 and the first period of production rates with Well P-1.
Within the last production set of three hours, only the last hour of production could not be
achieved with Well P-1. This last hour of production could be matched by increasing the
pore volume to three times the original pore volume. An increase in permeability
combined with a small increase in pore volume is a more reasonable and realistic
scenario than using a smaller permeability with an extremely large pore volume, or an
unrealistically large permeability.
A 1,000 md reservoir will effectively support CAES. However, reservoirs with
this large of a permeability are still hard to come by, especially in the Rocky Mountain
region, where a large amount of wind energy is possible and where aging gas fields could
contribute depleted reservoirs for CAES use. A 100 md model would be a better
representation of reservoirs in the Rockies, but based on the above results, the injection
and production rates need to be adjusted to accommodate for the decreased permeability.
The rates and representative volume of the Huntorf facility are not applicable to a 100 md
106
permeability reservoir. Therefore, these parameters need to be adjusted to more realistic
values in order to accommodate CAES in tighter reservoirs.
107
CHAPTER 5
CAES SIMULATION OF THE GREATER GREEN RIVER BASIN
Based on the conclusions reached in Chapter 4, if CAES is to be implemented in a
lower permeability environment, then different rates need to be used in the compression
and expansion schedule. This scenario was explored by modeling a practical, real-life
application of CAES. A combination of wind energy and CAES was modeled for the
Greater Green River Basin (GGRB) to determine an optimal schedule of injection and
production based on different geologic settings. Modeling of CAES focused on the
Frontier formation within the GGRB. The GGRB has been producing oil and gas out of a
number of reservoirs for over 80 years (DeJarnett, Lim and Calogero 2003). Some of the
fields in the basin are aging and seeing decreased production. After the profitable
production period in these fields have been realized, these depleted reservoirs could serve
as storage for CAES. This area of the country also sees high-sustained wind that would
be ideal for wind farms (Figure 1.1). There are already existing wind farms and the
potential exists for others. Foote Creek Rim, described in Section 1.1.3, is located within
the boundaries of the GGRB, as well as, the Medicine Bow Wind Project Site. The
Medicine Bow site has been in operation since the late 1970s. The main purpose of the
site was to analyze wind data to determine future wind farm construction. Today, the site
includes 10 wind turbines of various size and power ratings. The data collected from the
turbines has determined that the Medicine Bow area is a prime location for wind energy
(Table 5.1) A 40 unit, 100 MW wind farm is in the planning phase, after an
108
environmental assessment determined the wind farm would have no adverse effect on the
surroundings (Platte River Power Authority 2006).
Table 5.1 Predicted energy values based on average wind speeds for Historical Data (1987 – 1992) and 2004 Data and Actual Energy values collected on wind turbines
Historical Data (1987-1992) 2004 Data Performance Summary
MonthAvg Wind
SpeedPredicted
Energy
Avg Wind
SpeedPredicted
EnergyActual Energy
% of Hist. Predicted
% of 2004
Predictedmph MWh mph MWh MWh
Jan 27.8 2,551.7 23.2 2,255.9 2,010.1 78.8 89.1Feb 25.0 1,990.4 16.7 1,279.7 1,336.7 67.2 104.5Mar 22.9 2,023.5 20.8 1,917.5 1,917.7 94.8 100.0Apr 17.3 1,264.6 15.5 1,211.4 1,188.9 94.0 98.1May 16.5 1,148.6 19.2 1,794.0 1,758.3 153.1 98.0Jun 16.1 1,073.3 15.7 1,196.0 1,137.8 106.0 95.1Jul 13.0 671.4 14.9 755.5 820.4 122.2 108.6Aug 14.0 825.7 13.1 920.6 821.5 99.5 89.2Sep 13.2 725.3 14.3 1,101.2 1,036.1 142.9 94.1Oct 19.5 1,624.0 17.0 1,505.9 1,492.6 91.9 99.1Nov 25.1 2,226.4 16.8 1,308.7 1,387.4 62.3 106.0Dec 22.3 1,917.2 25.5 2,474.2 2,483.0 129.5 100.4
Totals 19.4 18,042.0 17.7 17,721.7 17,390.5 96.4 98.1
Modeling of the GGRB uses the same Huntorf injection and production schedule,
given in Table 3.6. This is used because it represents compression and expansion based
on daily energy demands. Since one of the goals of combined wind and CAES is to
compensate for peak demand, basing compression and injection off of energy needs is
logical. Different injection and production rates will be modeled to determine the
optimal rates for a given volumetric and geologic setting.
Using these different injection and production rates, the power necessary to run
the compressor and the expected power output of the turbine have been analyzed. A
wind schedule based on the Medicine Bow Wind Project site was compared to the power
109
input necessary to run the compressor for different injection rates. This comparison
showed the available excess energy that could be directly transferred to the utility grid.
Using the different production rates, energy output from the turbine was determined for
the hours in which the turbine was operating. This analysis will show how much energy
is available to consumers from the combination of wind energy and CAES.
This chapter will begin with geologic information on the GGRB and the fields of
study, followed by the necessary inputs for ECLIPSE 100© modeling, and then focus on
the results from these models. A discussion of the various models and their implications
will conclude the modeling portion of the chapter. The last section of the chapter will
concentrate on the power output possible with various production rates. The units in this
chapter will be in oilfield units, refer to Table 3.7 for conversion factors.
5.1 Geology of the Greater Green River Basin
The GGRB is located in the southwestern portion of Wyoming, extending into
portions of Colorado and Utah. Figure 5.1 highlights the basin.
110
Figure 5.1 Map of the Greater Green River Basin with reserve information on existing oil and gas fields Source: (Kirschbaum and Roberts 2005)
The GGRB is composed of a series of depressions, with separation coming from
various uplifts and ridges (Gibson 1997). Its area encompasses 19,700 square miles. The
basin boundaries are determined by major thrust faults or similar structures, including the
Wyoming-Utah-Idaho Overthrust Belt on the western edge of the basin, the Rawlins
Uplift and Park Range Uplift on the basin’s eastern side, the Uinta Mountains and the
Axial Basin Arch on the southern boundary, and the Wind River Mountains on the
northernmost part of the basin (DeJarnett et al. 2003). The GGRB consists of various
111
smaller basins, such as the Green River, Great Divide, Hoback, Sand Wash, and
Washakie Basins. Figure 5.2 highlights the major structures of the GGRB.
Figure 5.2 Major structures within the Greater Green River BasinSource: (Kirschbaum and Roberts 2005)
112
Various uplifts, the Rock Springs Uplift being the most prominent, as well as, four
significant anticlines, the Cherokee Ridge, Moxa Arch, Pinedale Anticline, and
Wamsutter Arch exist within the GGRB (Kirschbaum and Roberts 2005). Each uplift
transports crystalline Precambrian basement rocks to the surface and the anticlines
contain deep-seated highs that transform the basement rocks, along with the sedimentary
rocks above.
The majority of the reservoirs within the GGRB are sandstones, most of them of
Cretaceous age (Gibson 1997). Since modeling of combined wind and CAES was
conducted within the Frontier formation, its reservoir properties will be further examined.
The lithology of the Frontier is primarily composed of sandstone, siltstone, and
shale, with small amounts of coal and conglomerate (DeJarnett et al. 2003). Its
depositional environment is considered to be shoreface/deltaic, estuarine, and fluvial.
Five members of the Frontier can be identified on an outcrop in the western portion of the
GGRB. These five members include the Chalk Creek, Coalville, Allen Hollow, Oyster
Ridge, and Dry Hollow, in ascending order. The Frontier is segregated into the First
through Fourth Frontier in the subsurface, with the first and second Frontier acting as the
main reservoirs. In the western portion of the GGRB, the Frontier can be over 1000 ft
thick; towards the northeastern portion of the basin, thicknesses of 600-1,000 ft thick can
be expected. Depths of the Frontier can be around 2,000 ft to upwards of 20,000 ft from
the surface (Figure 5.3).
113
Figure 5.3 Subsurface depth of the Frontier formation within the Greater Green River BasinSource: (Kirschbaum and Roberts 2005)
The porosity and permeability of the Frontier are quite variable depending on the
location within the GGRB. Along the Moxa arch, the tidal, fluvial, and shoreface
sandstones all have similar porosities, averaging 9.3% to 11 %, with porosities reaching
maximums around 17%. Permeability in this area can reach up to 50 md, but some
114
measurements can be as low as 2 md. In areas east of the Moxa arch, the sandstone
porosity ranges from 2.4% to 28% and the permeabilities can be as low as 0.0008 md or
as high as 500 md (Kirschbaum and Roberts 2005).
The regional dip of the GGRB is 10-15 degrees on the west flank of the Rock
Springs Uplift (Figure 5.2) and 5-8 degrees on the uplift’s east flank (Flores and Bader
1999). According to a study conducted by the U.S. Geological Survey, the geothermal
gradient of the GGRB is highly variable. Gradients in the basin range from 1.2ºF/100 ft
to 2.2ºF/100 ft. Areas east of the Rock Springs Uplift typically exhibit the highest
gradients, while the lowest gradients can be found towards the southern boundary of the
Uinta Mountains (Finn 2005). A typical net to gross ratio in the GGRB ranges from 25%
to 40% (Oil and Gas Investor 2005). The net to gross ratio is described as the percentage
of the thickness of oil-bearing or gas-bearing rock to the thickness of entire reservoir
interval.
5.2 Model Inputs
In order to model CAES in the GGRB, a number of different models were
constructed for a range of porosities, permeabilities, injection and production rates, and
location within the basin. Two different basin locations were selected for modeling.
Locations where the Frontier is at shallower depths were chosen for modeling. The
reasonings behind this are as follows. As previously stated, depths of the Frontier in the
GGRB can range from 2,000 ft to greater than 20,000 ft. However, injecting air down to
depths of 20,000 ft is unreasonable for reasons such as high pressures and temperatures
and air compressibility issues. Additionally, the Frontier can become overpressured
around depths of 8,000 to 12,000 ft; the pressure gradients can exceed 0.9 psia/ft close to
these depths and throughout the remaining portion of the formation. Well design in these
regions is difficult and quite costly. Also at these greater depths, the porosities and
115
permeabilities begin to decline, permeabilities are typically less than 0.1 md. Based on
these stipulations, data from the Tip Top oil and gas field (Figure 5.4) and the Baxter
Basin South field (Figure 5.4) were used for the model locations.
Figure 5.4 Location of Tip Top field and Baxter Basin South within the Greater Green River BasinSource: (Kirschbaum and Roberts 2005)
116
The Tip Top field is located along the northern portion of the Moxa arch, and data from
the entire Moxa arch area was used for the first model location. The Baxter Basin South
model focuses on the central portion of the GGRB, and the high porosities and
permeabilities that can be attributed to this area.
The first model location, Tip Top, was discovered in 1928 and has produced
689,639 MMscf of gas and 8,572 Mbbl of oil. It still has numerous producing wells,
thanks to advanced stimulation and fracturing techniques (Wyoming Oil and Gas
Commission 2006). However, Tip Top is still an aging field with the potential for future
CAES development in depleted reservoirs. For a particular well in the Tip Top field, the
top of the First Frontier is located at 6,689 ft, the Second Frontier top is 7,122 ft, and the
Third Frontier is 7,613 ft. For modeling of this section of the Frontier, it was decided to
only consider the First Frontier due to permeability barriers that would be encountered if
the whole interval was considered. It would be unrealistic to assume that air could be
injected throughout the entire thickness of the Frontier. Therefore an average top of the
First Frontier was assumed to be 7,000 ft and an average thickness of 400 ft was used. A
net to gross ratio of 35% was also assumed leaving 140 ft of net thickness. A net to gross
ratio was used because the depleted reservoirs optimal for CAES will be within the net
reservoir zone. In order to get the remaining reservoir dimensions, the dip information
given in the previous section was applied. Using a dip of 12º, a length of 1,880 ft was
calculated. Applying this value of 1,880 ft to both the x and the y dimensions and using
the 400 ft total thickness, a bulk reservoir volume of 1.4138*109 ft3 was calculated.
Based on the porosity and permeability information given above for the Moxa arch area,
two different sets of porosity and permeabilites were used to construct the two models for
the Tip Top/Moxa arch region. Moxa 1, the best case scenario and Moxa 2, a less
desirable scenario are described in Table 5.2.
For the second model location, the Central GGRB/Baxter Basin South has been
modeled. The Baxter Basin South has been in production since 1922 with cumulative oil
and gas production of 570 bbl and 178,686 MMscf, respectively. As the numbers
117
indicate, Baxter is mostly a gas field. A review of some of the well files indicated that
the field produces minimal amounts of water. This creates an optimal scenario for using
CAES in a depleted reservoir. The well files also contained formation tops of the
Frontier formation, ranging from around 2,200 ft to 2,800 ft, with thicknesses around 180
ft (Wyoming Oil and Gas Commission 2006). An average top of 2,500 ft was selected
for modeling. The shallow depth of the Frontier in the Baxter field is also ideal for
CAES. This would help to decrease well costs and even allow for multiple wells, so
higher injection and production rates could be achieved. Baxter is also located close to
the Foote Creek Rim wind farm. Upon hydrocarbon depletion, this field would make an
excellent candidate for combined wind energy and CAES, from both a geologic and
geographic standpoint.
For the modeling of Baxter, two different geologic models were constructed.
However, just as was the case with the Moxa models, the bulk volume of each model is
the same. Since the Baxter South Basin is to the west of the Rock Springs Uplift, a dip of
10º-15º is still evident in this area. Using an average dip of 12º and the specified
thickness of 180 ft, a lateral extent of the reservoir could be approximated; this was
calculated as 850 ft. Assuming equal x and y dimensions, a bulk volume of 1.3005*108
ft3 was determined. Using this volume information, two models were constructed.
Baxter 1 is the best case scenario and Baxter 2 is a less desired scenario. The specifics of
these models are presented in Table 5.2. A net to gross ratio of 40% was applied to both
of the Baxter 1 and Baxter 2 models, yielding a net thickness of 72 ft.
Table 5.2 Model inputs for the Moxa and Baxter models
Model Porosity Permeability Pore Volume% md ft3
Moxa 1 17 50 240.339E+06Moxa 2 10 10 141.376E+06Baxter 1 28 500 36.414E+06Baxter 2 5 100 6.503E+06
118
Each model required a different EZGEN file. Since both the Moxa arch and Rock
Springs Uplift are anticlines, an anticlinal reservoir with the appropriate volumetric and
geologic properties, defined above, was constructed for each model. All of the EZGEN
input files contain 25 control points to create the structure and distribute parameters. The
EZGEN input files for the Moxa 1, Moxa 2, Baxter 1, and Baxter 2, models can be found
in Appendix C.1, C.2, C.3, and C.4, respectively. Figure 5.5 and Figure 5.6 show the
reservoir structure for the Moxa models and the Baxter models, respectively.
Figure 5.5 Reservoir structure created with EZGEN for use in Moxa 1 and Moxa 2 reservoir models
119
Figure 5.6 Reservoir structure created with EZGEN for use in Baxter 1 and Baxter 2 reservoir models
The EZGEN output files that were entered into ECLIPSE 100©, these files for the
Moxa 1, Moxa 2, Baxter 1, and Baxter 2 models can be found in Appendix C.5, C.6, C.7,
and C.8, respectively. With the construction of the grid and distribution of parameters in
ECLIPSE 100© with EZGEN, the appropriate injection and production rates could now
be determined for the Huntorf daily demand schedule. According to Benge and Dew
(2006), injection of 100 MMscf/day of a 65% H2S and 35% CO2 is possible within the
Madison formation of the GGRB. The Madison is a carbonate, consisting of anhydrite
and dolomite zones, all within a limestone formation. Permeability in the Madison
ranges from 0.1 md to 11 md. These high injection rates can be handled by the Madison
because of the secondary permeability that is made possible by the dolomitization
process. These high rates are also possible in areas that are naturally fractured, or could
120
be artificially fractured (Benge and Dew 2004). According to DeJarnett et al. (2003), the
Frontier formation is typically naturally fractured (DeJarnett et al. 2003). Based on the
injection rate capable within the GGRB, a maximum injection rate of 100 MMscf/day
will be modeled. Since the Huntorf injection and production schedule in Table 3.6 is
based on daily energy demand, this same schedule will be used in the GGRB modeling.
Huntorf also uses a four-to-one ratio of production rates to injection rates. This
production to injection ratio will be modeled, as well as, a two-to-one ratio for each
geographical model. Injection ratios of 100, 50, and 1 MMscf/day were initially modeled
for each model. However, the geologic properties of some models inhibited one or more
of these rates from being realized. Based on the results from these simulations and the
power analysis for the CAES compressor and turbine, optimal injection and production
rates were selected for each model.
Some of the production rates being used in the models could be hard to obtain
because of equipment limitations. Using these high production rates provides an
indication of what the formation is capable of producing. If these high rates are to be
achieved in an actual depleted reservoir, then a large production string would have to be
used, or multiple wells could be employed. The latter of the two options would be a
realistic scenario, given the number of existing wells in the two fields of study. As long
as the wells are tapping the same reservoir, then multiple wells could be used to obtain
these high production rates.
The PVT and saturation properties were kept the same as in the Cavern CAES and
Reservoir CAES models. If residual water can be pumped from the reservoir before
CAES operations begin and leakage from aquifers is not present, then this assumption
should be adequate. An ECLIPSE 100© example dataset for the GGRB modeling can be
found in Appendix C.9.
121
5.3 Model Results
Results from various runs of the models described in Table 5.1 have been
conducted and will be explained in detail in the following subsections.
5.3.1 Moxa 1 Model Results
The first model to be ran in ECLIPSE 100© was the Moxa 1 model. Table 5.3
shows the different injection and production rates that were modeled for the Moxa 1
model based on the results from the previous rate response.
Table 5.3 Moxa 1 modeled injection and production rates
Inj Rate Prod Rate Time SpanMMscf/day MMscf/day Days
100 400 11 75 period, rest 100 400 11 75 period, rest 100 400 31 75 period, rest 100 200 33 75 period, rest 100 200 3
50 200 350 100 31 4 31 2 3
For the initial model run, an injection rate of 100 MMscf/day and a production
rate of 400 MMscf/day were used. The reservoir could reproduce the large production
rate, but during the second injection period, the model could not obtain the 100
122
MMscf/day rate. Figure 5.7 shows the modeled injection rates and Figure 5.8 shows the
modeled production rates.
0
20000
40000
60000
80000
100000
120000
0 5 10 15 20 25 30
Time (hrs)
Msc
f/day
Actual Inj RatesInj Rates for 400 MMscf/day
Figure 5.7 Moxa 1 injection rates for the 100 MMscf/day injection and 400 MMscf/day production model for one day
123
0
50000
100000
150000
200000
250000
300000
350000
400000
450000
0 5 10 15 20 25 30
Time (hrs)
Msc
f/dActual Rates
Modeled Rates
Figure 5.8 Moxa 1 production rates for the 100 MMscf/day injection and 400 MMscf/day production model for one day
In order to determine if future injection rates were being meet, the model was
extended over a period of three days, using the same daily injection and production
schedule. The results from this model run showed that only the second period of
injection on the first day was not being achieved. Therefore, this second injection period
was decreased from the 100 MMscf/day to 75 MMscf/day. The model was again run
with the one period of new injection rates. With this the change, all the injection rates
were achieved and there were no adverse effect on the production rates. Figure 5.9
shows the actual injection rates, the modeled injection rates, and the modeled injection
rates with the one 75 MMscf/day injection period; these are all modeled over a three-day
span.
124
0
20000
40000
60000
80000
100000
120000
0 10 20 30 40 50 60 70 80
Time (hrs)
Msc
f/day
Actual Inj RatesInj Rates for 400 MMscf/dayInj Rates for 400 MMscf/day w/ 1 75MMscf/day inj period
Figure 5.9 Moxa 1 injection rates for the 100 MMscf/day injection and 400 MMscf/day production model for three days
The pressure response was the same for the single daily schedule as it was for the
first day of the three-day schedule. Therefore, Figure 5.10 shows the pressure response
for the three-day schedule with the original injection schedule of 100 MMscf/day and
with the modified injection schedule of one injection period of 75 MMscf/day and the
remaining injection periods staying constant at 100 MMscf/day. To provide more detail
of the actual pressures seen in the model, Figure 5.11 provides a zoomed in view of the
pressure response.
125
0
500
1000
1500
2000
2500
3000
3500
4000
0 10 20 30 40 50 60 70 80
Time (hrs)
Pres
sure
(psi
a)
3 days
3 days w/ 1 lower inj period
Figure 5.10 Moxa 1 pressure response for the 100 MMscf/day injection and 400 MMscf/day production model for three days
3500
3550
3600
3650
3700
3750
3800
0 10 20 30 40 50 60 70 80
Time (hrs)
Pres
sure
(psi
a)
1 day
3 days
3 days w/ 1 lower inj period
Figure 5.11 Zoomed in Moxa 1 pressure response for the 100 MMscf/day injection and 400 MMscf/day production model for three days
126
Figure 5.11 shows that one lower injection period has a minimal effect on the
pressure response. However, the overall pressure response indicates that the production
is too high because the pressure is continually dropping. The highs and lows expected
with the injection and production schedule are still present, but the pressure isn’t able to
rebound to as high of a pressure as it demonstrated in the same time step of the previous
day. This indicates that lower production rates need to be modeled. Therefore, a two-to-
one ratio of production and injection rates was modeled.
For the initial model, with the 100 MMscf/day of injection and 200 MMscf/day of
production, a three-day time frame was considered using the same production and
injection schedule with the one decreased injection period. As before, the reservoir had
no issues handling the production rates. However, a decrease in production rate caused
the second injection period in days 2 and 3 to not reach 100 MMscf/day (Figure 5.12).
0
20000
40000
60000
80000
100000
120000
0 10 20 30 40 50 60 70 80
Time (hrs)
Msc
f/day
Actual Inj Rates w/ 1 75 MMscf/day inj period
Inj Rates for 200 MMscf/day prod w/ 1 75MMscf/day inj period
Figure 5.12 Moxa 1 injection rates for the 100 MMscf/day injection with 1 injection period of 75 MMscf/day and 200 MMscf/day production model for three days
127
Based on the above results, another model run was conducted with all of the daily
second injection periods to 75 MMscf/day. With this change, all the specified injection
rates could be injected into the reservoir. Figure 5.13 provides a comparison of the
injection schedule that was initially applied to the 100 MMscf/day injection and 200
MMscf/day production model to the schedule that was necessary to obtain all of the
specified injection rates.
0
20000
40000
60000
80000
100000
120000
0 10 20 30 40 50 60 70 80
Time (hrs)
Msc
f/day
Previous Inj Rate schedule w/ 1 75 MMscf/day inj period
New Inj Rate schedule w/ 3 75 MMscf/day inj period
Figure 5.13 Moxa 1 injection rates for the actual initial schedule of 100 MMscf/day injection with 1 injection period of 75 MMscf/day and 200 MMscf/day production for three days compared to the
new schedule with 3 injection periods of 75 MMscf/day for three days
The pressure of the Moxa 1 model with the one lower injection period and the
pressure with three lower injection periods are displayed normally in Figure 5.14 and
with greater detail in Figure 5.15.
128
0
500
1000
1500
2000
2500
3000
3500
4000
0 10 20 30 40 50 60 70 80
Time (hrs)
Pres
sure
(psi
a)3 days w/ 1 lower inj period
3 days w/ 3 lower inj periods
Figure 5.14 Moxa 1 pressure response for the 100 MMscf/day injection and 200 MMscf/day production model for three days with 2 different injection schedules
3500
3550
3600
3650
3700
3750
3800
3850
3900
3950
4000
0 10 20 30 40 50 60 70 80
Time (hrs)
Pres
sure
(psi
a)
3 days w/ 1 lower inj period
3 days w/ 3 lower inj periods
Figure 5.15 Zoomed in Moxa 1 pressure response for the 100 MMscf/day injection and 200 MMscf/day production model for three days with 2 different injection schedules
129
Figure 5.15 shows that an adjustment of the second set of injection rates on each
daily rate schedule definitely has an effect on the pressure response. With the one lower
injection period, the pressure has a steadily decreasing trend. However, with the three
lower injection periods, the overall pressure change is constant. This relationship is
desired because all of the intended production rates are obtained and the overall pressure
change is constant throughout the three days of the daily production and injection
schedule.
The Moxa 1 model was then modified to inject a 50 MMscf/day rate. A four-to-
one and two-to-one production to injection ratio again served as the model runs. Figure
5.16 shows the pressure response for the production to injection ratios for a 50
MMscf/day injection rate and Figure 5.17 provides a zoomed in view of Figure 5.16.
0
500
1000
1500
2000
2500
3000
3500
4000
0 10 20 30 40 50 60 70 80
Time (hrs)
Pres
sure
(psi
a)
3 days w/ 200 MMscf/day prduction
3 days w/ 100 MMscf/day production
Figure 5.16 Moxa 1 pressure response for the 50 MMscf/day injection and 200 and 100 MMscf/day production model for three days
130
3500
3550
3600
3650
3700
3750
3800
0 10 20 30 40 50 60 70 80
Time (hrs)
Pres
sure
(psi
a)3 days w/ 200 MMscf/day production
3 days w/ 100 MMscf/day production
Figure 5.17 Zoomed in Moxa 1 pressure response for the 50 MMscf/day injection and 200 and 100 MMscf/day production model for three days
With these model stipulations, all of the injection and production rates could be achieved
without having to alter any of the injection rates. However, the overall change in
pressure decreased to a greater degree than it did with the previous models. The two-to-
one production to injection ratio could also inject and produce the specified rates without
any alterations, but the effect on the pressure was the same as with the four-to-one ratio.
Both the four-to-one and the two-to-one models show a minimal pressure change,
but the two-to-one model keeps a relatively constant pressure throughout the injection
and production schedule. The four-to-one model has the same highs and lows as the two-
to-one model, but the overall pressure change is decreasing over the three-day period.
The overall change in pressure of the four-to-one model is 28 psia, while the overall
change in pressure for the two-to-one model is 8.55 psia. By examining Figure 5.17, it
can be seen that the highs and lows are not significantly different between the two
131
models. Therefore, the majority of the difference between the overall change in pressure
between the two models can be attributed to the steadily declining pressures of the four-
to-one model.
From the previous two models, it can be seen that a two-to-one injection to
production ratio allows the change in pressure to remain relatively constant. The final
modification to the Moxa 1 model involved decreasing the injection rate to 1 MMscf/day.
A modeling of this rate showed the reservoir response to smaller rates that might be
necessary because of equipment limitations. With this adjustment, all of the injection and
production rates were realized for both the 4 MMscf/day (four-to-one ratio) and the 1
MMscf/day (two-to-one ratio) production rates. The normal pressure response is similar
to the above figures; Figure 5.18 shows a zoomed in view of the pressure response for the
4 MMscf/day and 1 MMscf/day production rates.
3500
3550
3600
3650
3700
3750
3800
0 10 20 30 40 50 60 70 80
Time (hrs)
Pres
sure
(psi
a)
3 days w/ 4 MMscf/day prduction
3 days w/ 2 MMscf/day production
Figure 5.18 Zoomed in Moxa 1 pressure response for the 1 MMscf/day injection and 4 and 2 MMscf/day production model for three days
132
The pressure change within the reservoir was minimal due to the reservoir size and the
low rates. Even with a maximized view of the pressure response, hardly any change in
pressure can be seen. This is due to the fact that these low of rates have a minimal
response on the reservoir because of the reservoir’s size. The overall change in pressure
is 0.563 psia for the four-to-one ratio model and 0.171 psia for the two-to-one ratio
model.
5.3.2 Moxa 2 Model Results
In order to cover different facets of the Moxa arch and Tip Top field area, the
Moxa 2 EZGEN dataset was constructed with lower possible values of porosity and
permeability. After input into the ECLIPSE 100© dataset, an initial model run could be
conducted. As with the Moxa 1 model, the Moxa 2 modeling began with an injection rate
of 100 MMscf/day and a 400 MMscf/day. The various injection and production rates that
were simulated with the Moxa 2 model are shown in Table 5.4.
Table 5.4 Moxa 2 modeled injection and production rates
Inj Rate Prod Rate Time SpanMMscf/day MMscf/day Days
100 400 120 80 320 40 31 4 31 2 3
The results obtained with the initial rates of 100 MMscf/day of injection and 400
MMscf/day of production were quite different than what was seen in Moxa 1. With the
133
Moxa 2 model, neither the injection or production rates could be obtained. Additionally,
the highs and lows of the pressure response were not as pronounced as in the initial
model runs with Moxa 1. Figure 5.19 and Figure 5.20 show the injection and production
rates, respectively and Figure 5.21 shows the detailed pressure response of the 100
MMscf/day injection and 400 MMscf/day production rates model.
0
20000
40000
60000
80000
100000
120000
0 5 10 15 20 25 30
Time (hrs)
Msc
f/day
Actual Inj RatesInj Rates for 400 MMscf/day
Figure 5.19 Moxa 2 injection rates for the 100 MMscf/day injection and 400 MMscf/day production model for one day
134
0
50000
100000
150000
200000
250000
300000
350000
400000
450000
0 5 10 15 20 25 30
Time (hrs)
Msc
f/d
Actual Rates
Production Rate of 400 MMscf/day
Figure 5.20 Moxa 2 production rates for the 100 MMscf/day injection and 400 MMscf/day production model for one day
3500
3550
3600
3650
3700
3750
3800
0 5 10 15 20 25 30
Time (hrs)
Pres
sure
(psi
a)
Figure 5.21 Zoomed in Moxa 2 pressure response for the 100 MMscf/day injection and 400 MMscf/day production model for one day
135
Figure 5.19 and Figure 5.20 indicate that the lower porosity and permeability of
Moxa 2 are not able to handle the high injection and production rates. By examining the
injection and production rates that could be achieved through the modeling, it looked as
though a 25 MMscf/day injection and 100 MMscf/day production rate could be handled
by the reservoir. However, the model still had trouble injecting the 25 MMscf/day. The
100 MMscf/day production rate could be obtained, but a decrease to 25 MMscf/day
injection rate was still not enough. Therefore, a rate schedule of 20 MMscf/day injection
and 80 MMscf/day production rates was modeled. The Moxa 2 model was able to inject
and produce the specified rates of this schedule, so the two-to-one production to injection
ratio was also modeled. All of the rates were reproduced in this model as well. The
zoomed in pressure response for the four-to-one and two-to-one production to injection
ratios for an injection rate of 20 MMscf/day is given in Figure 5.22.
3500
3550
3600
3650
3700
3750
3800
0 10 20 30 40 50 60 70 80
Time (hrs)
Pres
sure
(psi
a)
3 days w/ 80 MMscf/day production
3 days w/ 40 MMscf/day production
Figure 5.22 Zoomed in Moxa 2 pressure response for the 20 MMscf/day injection and 80 and 40 MMscf/day production model for three days
136
Figure 5.22 shows that with a 20 MMscf/day injection rate, the pressure begins to
level off; the highs and lows are not as prominent. The above figure also emulates a
result found with the Moxa 1 model. With a two-to-one production to injection ratio, the
overall change in pressure is not as large as with the four-to-one ratio. In this case, the
pressure change was 19 psia for the four-to-one ratio model and 5.8 psia for the two-to-
one ratio model.
As with the Moxa 1 model, lower rates were modeled with the Moxa 2 model to
determine how the reservoir will respond. As expected, all of the injected rates of 1
MMscf/day and the production rates of 4 MMscf/day for the four-to-one model and 2
MMscf/day for the two-to-one model were obtained in the model runs. The pressure
response for these two simulations is given in Figure 5.23.
3500
3550
3600
3650
3700
3750
3800
3850
3900
3950
4000
0 10 20 30 40 50 60 70 80
Time (hrs)
Pres
sure
(psi
a)
3 days w/ 4 MMscf/day prduction
3 days w/ 2 MMscf/day production
Figure 5.23 Zoomed in Moxa 2 pressure response for the 1 MMscf/day injection and 4 and 2 MMscf/day production model for three days
137
The results of the Moxa 2 model are similar to the results seen with the Moxa 1
model for these low rates. Because of the size of the reservoir, the injection and
production do not have much of an effect on the reservoir. Therefore the change in
pressure is minimal.
5.3.3 Baxter 1 Model Results
Using the EZGEN dataset constructed for the Baxter 1 model, an initial model run
was performed in ECLIPSE 100© with the injection rates of 100 MMscf/day and the
production rates of 400 MMscf/day. The injection and production rates modeled for the
Baxter 2 model is shown below in Table 5.5.
Table 5.5 Baxter 1 modeled injection and production rates
Inj Rate Prod Rate Time SpanMMscf/day MMscf/day Days
100 400 3100 200 350 200 350 100 31 4 31 2 3
All of the Baxter model runs used a three-day time span for the injection and
production schedule. Due to the high porosity and permeability associated with the
Baxter 1 model, all of the specified injection and production rates could be replicated in
the model. Another model run was conducted with a two-to-one ratio of production to
injection rates. The pressure response for the four-to-one ratio and the two-to-one ratio is
depicted in Figure 5.24.
138
0
100
200
300
400
500
600
700
800
900
1000
0 10 20 30 40 50 60 70 80
Time (hrs)
Pres
sure
(psi
a)3 days w/ 400 MMscf/day prod rate
3 days w/ 200 MMscf/day prod rate
Figure 5.24 Baxter 1 pressure response for the 100 MMscf/day injection and 400 and 200 MMscf/day production model for three days
The same result is again seen with the Baxter 1 model as with the previous
models. The four-to-one ratio model has a greater overall pressure drop than the two-to-
one ratio. With this model, however, the slight pressure increase that was seen with the
two-to-one ratio model is more pronounced. The change in pressure is 136 psia with the
four-to-one model and 97 psia for the two-to-one model.
Since the Baxter 1 model had no problem injecting 100 MMscf/day and
producing 400 MMscf/day, the reservoir could obviously handle the lower injection rates
of 50 MMscf/day and 1 MMscf/day and the corresponding production to injection ratios
of four-to-one and two-to-one. The pressure response for the 50 MMscf/day models and
the 1 MMscf/day models are given in Figure 5.25 and Figure 5.26, respectively.
139
0
100
200
300
400
500
600
700
800
900
1000
0 10 20 30 40 50 60 70 80
Time (hrs)
Pres
sure
(psi
a)3 days w/ 200 MMscf/day production
3 days w/ 100 MMscf/day production
Figure 5.25 Baxter 1 pressure response for the 50 MMscf/day injection and 200 and 100 MMscf/day production model for three days
0
100
200
300
400
500
600
700
800
900
1000
0 10 20 30 40 50 60 70 80
Time (hrs)
Pres
sure
(psi
a)
3 days w/ 4 MMscf/day production
3 days w/ 2 MMscf/day production
Figure 5.26 Baxter 1 pressure response for the 1 MMscf/day injection and 4 and 2 MMscf/day production model for three days
140
The 50 MMscf/day exhibits the same behavior as the 100 MMscf/day model, but
the highs and lows are not as pronounced as they are in the 100 MMscf/day model. The
low rate model shows little pressure response because of the reservoir size and properties.
In order to determine what the reservoir was capable of handling with these low rates, a
large production to injection ratio model was run. With a 1 MMscf/day injection rate, a
20 to 1 ratio was achieved for production to injection rates without any issues obtaining
the specified rates. With this large of a ratio, the pressure never rebounds and a steady
decline in pressure is observed. Also when this high of a ratio is considered, the fact that
air is not being replaced as it is extracted from the reservoir needs to be taken into
account.
5.3.4 Baxter 2 Model Results
The lower porosity and permeability of the Baxter 2 model had a significant effect
on production rates and consequently, on the pressure response. Once again, the initial
model run used injection rates of 100 MMscf/day and production rates of 400
MMscf/day. All of the modeled injection and production rates for Baxter 2 are located in
Table 5.6.
Table 5.6 Baxter 2 modeled injection and production rates
Inj Rate Prod Rate Time SpanMMscf/day MMscf/day Days
100 400 325 100 325 50 31 4 31 2 3
141
This reservoir model had no problem injecting the 100 MMscf/day, but could only
produce around ¼ of the production rates. Figure 5.27 shows the production rates for the
100 MMscf/day injection rates.
0
50000
100000
150000
200000
250000
300000
350000
400000
450000
0 10 20 30 40 50 60 70 80
Time (hrs)
Msc
f/d
Actual RatesProduction Rate of 400 MMscf/day
Figure 5.27 Baxter 2 production rates for the 100 MMscf/day injection and 400 MMscf/day production model for three days
The production is increasing from one day of production to the next, but even by
the third day, the modeled production rates are still half of the desired rates, at best. The
pressure response emulates this lack of production with continually increasing values.
With all of the injection rates entering the reservoir, but not all of the production rates,
the pressure has to increase. The reservoir volume is at its threshold from the increased
142
supply of air, but without the available production to relieve the pressure, the pressure
just keeps increasing. Figure 5.28 shows this pressure response.
0
200
400
600
800
1000
1200
1400
0 10 20 30 40 50 60 70 80
Time (hrs)
Pres
sure
(psi
a)
Figure 5.28 Baxter 2 pressure response for the 100 MMscf/day injection and 400 MMscf/day production model for three days
The highs and lows expected with the compression and expansion of air are still
present, but with not as much production to relieve the pressure, a steadily increasing
pressure occurs. In order to try and find more optimal injection and production rates, a
model with 25 MMscf/day of injection and 100 MMscf/day of production was run. Even
with this decrease in rates, the reservoir was still not capable of producing 100
MMscf/day of air. Figure 5.29 displays the production rate results from the model run.
143
0
20000
40000
60000
80000
100000
120000
0 10 20 30 40 50 60 70 80
Time (hrs)
Msc
f/dActual RatesProduction Rate of 400 MMscf/day
Figure 5.29 Baxter 2 production rates for the 25 MMscf/day injection and 100 MMscf/day production model for three days
The production rates with this model are an improvement over the previous model
run of 100 MMscf/day injection and 400 MMscf/day production, but these rates are still
too high to achieve the desired production rate. The pressure response associated with
the injection rate of 25 MMscf/day and the production rate 100 MMscf/day is displayed
in Figure 5.30.
144
0
100
200
300
400
500
600
700
800
900
1000
0 10 20 30 40 50 60 70 80
Time (hrs)
Pres
sure
(psi
a)
Figure 5.30 Baxter 2 pressure response for the 25 MMscf/day injection and 100 MMscf/day production model for three days
The pressure response with these rates for the Baxter 2 model is similar to the
response seen with the previous models. This is because more of the air is being
produced out of the reservoir, allowing for a greater pressure drop to occur. Since the
rates were very close to being achieved for the four-to-one ratio, a two-to-one ratio model
was run. By decreasing the production rates to 50 MMscf/day, all of the rates could be
obtained. The pressure response of the two-to-one ratio model for the 25 MMscf/day
injection rate is located in Figure 5.31.
145
0
100
200
300
400
500
600
700
800
900
1000
0 10 20 30 40 50 60 70 80
Time (hrs)
Pres
sure
(psi
a)
Figure 5.31 Baxter 2 pressure response for the 25 MMscf/day injection and 50 MMscf/day production model for three days
Lower injection rates were modeled with a four-to-one ratio, but with the low
porosity of the Baxter 2 model, a four-to-one ratio could not be achieved. Therefore, the
final model with the Baxter 2 model was the low rate model of 1 MMscf/day of injection
and a four-to-one and two-to-one ratio of production. Like the previous models, all of the
injection and production rates with these low rates were realized and the change in
pressure (Figure 5.32) was minimal.
146
0
100
200
300
400
500
600
700
800
900
1000
0 10 20 30 40 50 60 70 80
Time (hrs)
Pres
sure
(psi
a)3 days w/ 4 MMscf/day prduction
3 days w/ 2 MMscf/day production
Figure 5.32 Baxter 2 pressure response for the 1 MMscf/day injection and 4 and 1 MMscf/day production model for three days
5.4 Power Analysis
A power analysis will be provided for power output that the turbine can generate
based on the production rates obtained above. Efficiencies have not been included in the
calculations since these vary between different pieces of machinery. Therefore, when
efficiencies are considered the power output supplied by the turbine will be lower than
indicated.
In order to determine the power that can be supplied based on the modeled
production rates, a series of calculations were performed. These calculations were
independent of depth and reservoir properties since the calculations are made at the inlet
147
and outlet of the turbine, which are both at the surface. Therefore, the power output for
the Moxa and Baxter models for the various production rates is the same.
To begin the series of calculations, the production rates were converted to
volumes of air based on the length of the given production period. There were two daily
production periods, one lasting for two hours and the other for three hours. Once the
produced air volume from each production period was found, then the mass of air was
calculated using the air surface density of 1.2395 kg/m3 (0.0774 lbm/ft3). According to
Cheung et al. (2006) the power output for a CAES turbine is
(5.5)
where Qs = power output, m = mass of air, cp = specific heat of air = 1.05 kJ/kg K at
surface, and = change in temperature. The term uses the ambient air temperature
of 283.149 K (50 F) and the average temperature of the air after it has been heated for
turbine entry of 1,050 K (1,430 F) (Cheung et al. 2006). With the power calculated, the
energy, in MWh, could also be found by dividing the power term by the power
generation period of either two or three hours. The results from these calculations for all
of the production rates for the Moxa models and the Baxter models are included in Table
5.7.
Table 5.7 Generated power and from the various production rates of the Moxa and Baxter models
148
Production Rate Air Volume Air Mass
Generated Power
Total Daily Power
MMscf/day m3 kg MW MW2 hr 3 hr 2 hr 3 hr 2 hr 3 hr Total
400 943894.9 1415842.41169987.
01754980.
5 261.7 392.5 654.2200 471947.5 707921.2 584993.5 877490.3 130.8 196.3 327.1100 235973.7 353960.6 292496.8 438745.1 65.4 98.1 163.680 188779.0 283168.5 233997.4 350996.1 52.3 78.5 130.850 117986.9 176980.3 146248.4 219372.6 32.7 49.1 81.840 94389.5 141584.2 116998.7 175498.1 26.2 39.3 65.44 9438.9 14158.4 11699.9 17549.8 2.6 3.9 6.52 4719.5 7079.2 5849.9 8774.9 1.3 2.0 3.3
The power created from modeled production rates is quite large for the higher
production rates. If the facilities can support such large rates through wide production
strings or multiple wells, then a significant amount of stored energy can be made
available to meet energy demands. In order to make a comparison to conventional
energy sources, Table 5.8 was constructed.
Table 5. 8 Comparison of CAES daily power output to amount of natural gas and coal necessary to achieve the same power over the five-hour production period
CAES Daily PowerNatural Gas using 820
BTU/scfCoal using 26
million BTU/ton# of Homes
Provided w/ PowerMW BTU/hr scf MMscf ton
654.2 223.23E+07 1361.1E+04 13.6 429.3 198,469327.1 111.61E+07 680.57E+04 6.8 214.6 99,234163.6 55.807E+07 340.28E+04 3.4 107.3 49,617130.8 44.645E+07 272.23E+04 2.7 85.9 39,69481.8 27.903E+07 170.14E+04 1.7 53.7 24,80965.4 22.323E+07 136.11E+04 1.4 42.9 19,8476.5 2.2323E+07 13.611E+04 0.1 4.3 1,9853.3 1.1161E+07 6.8057E+04 0.1 2.1 992
149
For each five-hour power-generating period of CAES, a conversion was made to BTU/hr.
This could then be divided by the 820 BTU/scf for natural gas and the 26 million
BTU/ton of coal (Forest Products Laboratory 2004). Multiplying these values by the
five-hour production period gives the equivalent amount of the more conventional energy
sources. In order to include a more realistic power application, the number of homes that
each MW value can supply power for is provided. With the highest power output, almost
200,000 homes can be provided with electricity.
5.4 Discussion of GGRB Models and Power Implications
The overall response to the various injection and production rates of the GGRB
models was encouraging for implementing CAES in actual reservoirs. The high porosity
and permeability of the Baxter 1 model allowed for the achievement of the highest
modeled injection and production rates. If the appropriate surface facilities can be used,
then 654.2 MW of daily power can be produced with the addition of a CAES plant.
When this is combined with an average wind power around 1,450 MW, a significant
energy source is available. Having 654.2 MW accessible when the wind is not blowing,
or during a period of peak demand, is a great solution to the problem of wind being an
intermittent energy source. For the Baxter 1 model, the optimal injection and production
rates are the 100 MMscf/day of injection and 400 MMscf/day of production. If
information can be found that justifies a higher injection rate in the region, then the
Baxter model could use even higher rates. However, based on the current information
available, the injection and production rates above are the optimal rates for this high
porosity and permeability model. Additionally, if equipment restraints limit the injection
and production rates, then lower rates will still provide an adequate power output (Table
5.10) to compensate for the shortcomings of wind energy.
150
The Moxa 1 model could handle the large production rate of 400 MMscf/day
associated with the initial model run, but could not reproduce the 100 MMscf/day
injection rate during the second injection period of the daily schedule. Once this second
injection period was reduced to 75 MMscf/day, the model could handle all of the
injection and production rates. The power output that can be produced with the Moxa 1
model is 654.2 MW, the same as the power output with the Baxter 1 model. Lower rates
can be used if equipment limitations exist, but the optimal injection and production
schedule for the porosity and permeability of the Moxa 1 model is 100 MMscf/day
injection rates, with 75 MMscf/day injection rates during the second daily period of
injection, and a 400 MMscf/day production rate.
With the lower porosity and permeability values of the Moxa 2 and Baxter 2
models, the high injection and production rates of 100 MMscf/day and 400 MMscf/day,
respectively, could not be achieved. With the Moxa 2 model, the injection rate had to be
decreased to 20 MMscf/day with a corresponding production rate of 80 MMscf/day. This
yields a daily power output of 130.8 MW, which is still a large enough output to justify
the combination of CAES with wind energy in the region. The Baxter 2 model has a
lower porosity, but a higher permeability than the Moxa 2 model. With this combination,
it can achieve a slightly higher injection rate than the Moxa 2 model, but a lower
production rate. The optimal schedule obtained with the Baxter 2 model is an injection
rate of 25 MMscf/day and a production rate of 50 MMscf/day. Even with lower injection
rates, this model required a two-to-one production ratio. With this production rate, the
Baxter 2 model is capable of supplying a daily power output of 81.8 MW. Once again,
both of these models could use lower injection and production rates to compensate for
wind intermittency, if the facilities could not accommodate the higher rates.
Based on the above results, if the reservoir unit has a low permeability then the
injection rates are harder to achieve than the production rates. Overall, both of the Baxter
models have higher permeabilities than the Moxa models; the lowest permeability with
the Baxter models is 100 md (Baxter 2), which is higher than the highest permeability of
151
50 md, with the Moxa 1 model. Consequently, the Baxter models had no problems
injecting the 100 MMscf/day injection rates. Both of the Moxa models had difficulties
injecting the 100 MMscf/day rates. The Baxter 1 model had a very high porosity and
permeability and therefore had no issues obtaining any of the rates. The Baxter 2 model
has low porosity and a higher permeability, which has more of an effect on the
production rates. Since Moxa 2 has both low porosity and low permeability, the
production rates are affected, as well as, the injection rates. Therefore, according to these
four models, permeability is the driving force for injecting and producing the desired
rates. This is because permeability is the measure of a formation’s ability to transport
fluid and porosity is the amount of storage space available. A formation with a higher
permeability allows fluid to flow more easily than a formation with a lower permeability.
This is why the Baxter models could inject air into the formation better than the Moxa
models. Permeability also affects production; the higher the permeability the better the
fluid can be produced out of a formation. The production issue associated with the
Baxter 2 model can be attributed to the combination of a 5% porosity and a 100 md
permeability.
The optimal rate schedules described above all incorporate the four-to-one
production to injection ratios. These ratios were selected because they optimize the
power production that can be obtained. The two-to-one ratios maintain the pressure in
the reservoir better, but do not provide as much power as the four-to-one ratios. If
pressure loss in the reservoir ever becomes an issue, then the production to injection ratio
can be reduced for a few days, until the desired pressures are restored. At this time, the
four-to-one ratio can be resumed to optimize power production. Additionally, if not as
much energy from storage is necessary during a certain time period, then the production
to injection ratio can be reduced, or both rates could be minimized. By knowing the
maximum injection and production rates that these different reservoir models can handle,
a combination of rates can be used to achieve the desired power output from storage of
compressed air.
152
CHAPTER 6
CONCLUSIONS
This final chapter will present the major results and conclusions from the
modeling of CAES in different geologic settings. All three model sets will be examined
and some similarities and differences will be discussed. The important conclusions of the
study will then be the focus. The chapter will look at recommendations for future work
in the area of implementing CAES in reservoir environments and a final discussion will
conclude the study.
6.1 Major Results
The modeling of CAES began with the construction of the Cavern CAES model
to simulate the Huntorf CAES facility. The goal of this model was to obtain a pressure
match using the same production and injection schedule and rates that are used at
Huntorf. With this initial model setup, the pressure match was unsatisfactory. Changes
in pore volume were simulated to try to obtain a match, but these attempts were
unsuccessful. After a personal communication with F. Crotogino, it was realized that the
production rates are maximum obtainable rates. Therefore, the production rates were
decreased for the next model run. With this adjustment, an acceptable pressure match
could be realized. This model with the decreased production rates is labeled as
Sensitivity C and was selected as the best representation of the Huntorf facility. Different
153
permeabilities were modeled with the Cavern CAES model to show that permeability was
not having an effect on the pressure response in a cavern environment; the result was as
desired.
Successful modeling of the Huntorf facility with ECLIPSE 100© shows that
petroleum engineering tools can be used to examine sustainable energy options, such as
CAES. Achieving a reasonable pressure match with the recorded Huntorf pressures,
verifies the model’s creditability for simulating similar processes. Therefore, the model
was then modified to simulate a pore volume equal to the Huntorf cavern volume and the
same rates and schedule as Sensitivity C, but in a reservoir. EZGEN files were
constructed for 100 md, 1,000 md, and 10,000 md permeability and model runs were
conducted in ECLIPSE 100©. The 100 md model could not achieve the Huntorf
production rates, even with an increase up to 100,000 times the original pore volume.
The rates that are used at Huntorf are too high for this lower permeability and porosity
environment. The 1,000 md model gave a better response. It only took three times the
original pore volume, to obtain the rates of Sensitivity C in the Cavern CAES model.
Before the increase, there was only one hour of production in the second production
period of Well P-1 that could not achieve the rate. As expected, if the permeability was
increased to 10,000 md, the reservoir had no problem obtaining the desired production
rates.
Obviously, a 100 md reservoir with 20% porosity is not reasonable for using the
same rates and effective pore volume that can be obtained in a cavern environment. If a
1,000 md reservoir can be used, then the Huntorf rates are reasonable since only one hour
of production with one well could not be reproduced. Even though the 10,000 md model
could produce all of the specified rates, a reservoir with a permeability of 10,000 md is
very rare. Therefore, if the same Sensitivity C Huntorf rates are to be used, then the
1,000 md model is the model of choice. With a working model of CAES in a cavern
setting and a reservoir setting using the same injection and production rates, a more
154
practical application was necessary to look at actual reservoirs where CAES could be
implemented.
Based on reservoir characteristics of different areas of the GGRB, four reservoir
models could be constructed. The Moxa 1 and Baxter 1 models are the best case
scenarios for reservoir properties in their areas and the Moxa 2 and Baxter 2 models, are
less ideal candidates in terms of reservoir properties. Based on available injection data in
the GGRB, an initial injection rate of 100 MMscf/day was used in all of the models, with
a four-to-one production to injection ratio. After this model was run, an optimal rate
schedule was found based on the initial results and subsequent model runs. The Moxa 1
model could produce all of the production rates, but couldn’t handle the injection rates.
The second injection period was causing the problem, so this rate was reduced to 75
MMscf/day for the three-day time span. With this change, an optimal injection and
production schedule was realized for the Moxa 1 model; the production rate of 400
MMscf/day was left unchanged. The high porosity and permeability of the Baxter 1
model allowed it to obtain all of the initial injection and production rates of 100
MMscf/day and 400 MMscf/day, respectively. Since additional data verifying higher
rates could not be found within the GGRB, the optimal injection and production rates are
these initial rates. For the lower porosities and permeabilities of the Moxa 2 and Baxter 2
models, the initial production and injection schedule could not be obtained. The Moxa 2
model required an injection rate of 20 MMscf/day and a production rate of 80
MMscf/day. The Baxter 2 model, with a lower porosity, but a higher permeability, could
inject rates of 25 MMscf/day, but only produce rates of 100 MMscf/day.
By comparing the models of the GGRB, it can be seen that the models with higher
permeabilities could inject and produce higher rates than the models with lower
permeabilities. This is because permeability determines how effective a fluid will flow
through a formation and porosity determines how much storage space is available. For
example, both Baxter models could inject the desired rates, but the lower permeability
Moxa models had problems injecting the initial rates. Excluding the Baxter 2 model, the
155
higher permeability models could also produce better. The combination of porosity and
permeability with the Baxter 2 model led to its inconsistency.
A power analysis of the CAES components necessary for the GGRB models,
determined that the turbine power output based on modeled production rates is quite large
for the higher production rates and still beneficial for lower production rate values.
6.2 Model Comparisons
Some definite similarities can be found between the three model sets. The shape
of the pressure response is comparable for all of the models. This is because even though
all of the models do not use the same injection and production rates, the same rate
schedule is used throughout. Since periods of injection raise the pressure in the structure
and periods of production lower the pressure, the same high and low pressure regions are
emulated throughout the models. The pressure values themselves are different, but the
overall trend is the same.
As far as some of the differences between the models, in the Cavern CAES
models, a modification to pore volume gave inconclusive results, but with the 100 md
and 1,000 md Reservoir CAES models, an increase in pore volume helped to increase the
production rates. With the reservoir models, a larger pore volume meant more space in
which the injected air could migrate. This in turn gives the reservoir a better chance of
producing all of the desired production rates. Since some of the air will be trapped in
pore throats upon injection and cannot be produced, an increase in pore volume increases
the amount of pore space, allowing for more air to be produced. For the Cavern CAES
models, this issue of pore space is not a concern. When trying to achieve the initial
pressure match by changing the pore volume, a smaller pore volume had a better pressure
response during the initial periods of injection and a larger pore volume had the better
response during the later periods of production and injection. Since there is not one pore
156
volume that can improve in both periods, the production rate is better initially with the
smaller pore volumes, but then digresses over time. The opposite is true with the larger
pore volumes.
6.3 Model Conclusions
The major conclusions of modeling CAES in different geologic settings will be
presented in terms of importance. The most significant conclusions gained from the
study will be at the beginning of the list and the conclusions with lesser importance will
be towards the end of the list.
1. CAES could be used in actual reservoirs, depending on the geologic properties.
a. All four of the GGRB models demonstrated the ability to make use of
CAES. The rates that could be used with the lower porosity and
permeability Moxa 2 and Baxter 2 models were not as significant as the
higher porosity and permeability Moxa 1 and Baxter 1 models, but still
high enough to justify the implementation of CAES. However, the
potential for CAES can be maximized if reservoirs similar to the Moxa 1
and Baxter 1 models can be used.
b. Overall, the Baxter 2 model had the best results; this was the only model
able to inject and produce all of the 100 MMscf/day and 400 MMscf/day
respective rates. If studies can be conducted that show formations within
the GGRB can handle injection rates higher than 100 MMscf/day, then an
even greater power output could be realized with this model.
157
2. ECLIPSE 100© is an effective tool for modeling combined wind energy and
CAES.
a. Three different CAES models confirmed that ECLIPSE 100© can model
the CAES process independent of geologic setting. A successful pressure
match with the Cavern CAES model could be modified to represent a
hypothetical reservoir with the same injection and production rates,
schedule, and wells. Once successful model runs could be obtained with
this hypothetical reservoir, the use of a similar model in an actual reservoir
could be justified. Going through this series of steps confirms the validity
of using ECLIPSE 100© for the modeling of CAES.
3. The GGRB has good potential for a combined wind energy and CAES facility.
a. Not only are the results from modeling CAES in the GGRB basin
encouraging, but the abundant supply of wind energy makes the GGRB a
prime location for a combined wind energy and CAES facility. Foote
Creek Rim and the Medicine Bow Wind Project site are currently
providing wind power, and according to wind speeds at the Medicine Bow
Wind Project site and Figure 1.1, Class 4 to 6 wind speeds are present.
These speeds are more than adequate for making an economical wind
farm.
6.4 Recommendations for Future Work
The results from this study are a promising first step for implementing CAES in
porous media and the GGRB is a prime facility for continued research in this field. It has
abundant wind energy and some areas have the high porosities and permeabilities
necessary for successful CAES. However, a more intensive reservoir study needs to be
158
conducted to get a better understanding of the potential for combined wind and CAES in
this region. The list below shows the areas that should be considered for future research,
in terms of importance.
1. A single reservoir should be selected for study and all of the available reservoir
properties should be obtained.
a. A detailed reservoir characterization should be conducted. This should
include an examination of rock properties (porosity, permeability,
saturations, etc.), permeability barriers, such as shale, within the reservoir
unit, reservoir dimensions, natural fracturing, failure mechanisms, such as
stresses and strain of the reservoir rock, and leakoff from the reservoir,
more specifically, how well the reservoir is sealed. Failure mechanisms
are important to determine how high of pressures the formation can
handle, and in order for the pressurized air to stay within the reservoir,
seals need to be present around the entire reservoir.
b. A detailed reservoir simulation can then be conducted with the
information gained from the reservoir characterization. If the reservoir
has some residual water saturation, then the Honarpour or similar
correlations should be used in EZGEN to account for this situation. Once
all of the initial model inputs for the reservoir are entered into ECLIPSE
100©, different model runs can be conducted for a variety of scenarios.
c. A single well that is in charge of both injection and production could be
modeled as a means for cost reduction and possibly increased efficiency.
d. If the reservoir was in a depleted gas reservoir with multiple wells tapping
the reservoir, then injection and production can be modeled using these
existing wellbores. The well design and corrosion issues need to be
considered before implementing this option.
159
e. The use of horizontal wells to obtain higher rates should be modeled as
well, since they generally increase a well’s productivity index.
f. If an aquifer is attached to the reservoir this also needs to be addressed in
the study.
2. A more intensive analysis should be done with the combination of CAES and
wind energy.
a. If possible, wind data from the area where the reservoir is located should
be gathered and a more specific analysis should be made with the
compressor and turbine calculations. These calculations should consider
any energy losses that were found and the efficiencies of the compressor
and turbine to be used.
3. A cost analysis should be conducted for opening a combined wind and CAES
facility.
a. The cost of installing wind turbines, as well as, the cost of constructing a
CAES plant should be analyzed. An estimation of the cost per kW
produced, once the facility is up and running, should also be included.
6.5 Final Discussion
The results from the three models analyzed in this research, show that ECLIPSE
100© is an appropriate tool for modeling CAES in both cavern and reservoir
environments. First, a successful history match of the pressure response associated with
a daily injection and production schedule was obtained using known data in a cavern
environment. This model could then be modified to represent a reservoir environment.
After verifying the same production and injection schedule and rates with the reservoir
160
model, a model could be created for a practical application in the GGRB. The results
from these model runs are very encouraging for combining the power of wind energy
with the reliability of CAES. All of the models could produce rates that would be
beneficial for additional power production during intermittent wind energy or peak
demand periods. With the appropriate geographical location and reservoir properties,
such as those evident in the GGRB, this type of energy source can become a profitable
commodity that can help meet the energy needs of future generations.
161
NOMENCLATURE
CAES – Compressed air energy storageNREL – National Renewable Energy LabHAWT – Horizontal axis wind turbineGGRB – Greater Green River BasinSMES – Superconducting magnetic energy storageUTES – Underground thermal energy storageATES – Aquifer thermal energy storageDTES – Duct thermal energy storageHP – High pressureLP – Low pressureMBE – Material balance equationv = Fluid velocity
= The permeability tensor = Fluid viscosity
= The gradient of the potential functionh = the flow potential
FRP – Fiberglass reinforced plasticBw – Water formation volume factor
= Change in volume during the pressure reduction = Change in volume due to the reduction in temperature
Bg – Gas formation volume factorVR = Cubic foot of reservoir volumeVsc = Standard cubic foot of gasz = Gas-deviation factorT = Temperaturep = Pressurepr = Reduced pressureTr = Reduced temperaturepc = Critical pressureTc = Critical temperatureg = Gas viscosityMg = Gas molecular weight = PorosityVP = Pore volumeVB = Bulk volume
162
REFERENCES
Abou-Kassem, J. H., S. M. Farouq Ali, et al. (2006). Petroleum Reservoir Simulation. Houston, Gulf Publishing Company.
American Wind Energy Association. (1999). "Wyoming Wind Project Begins Powering Pacific Northwest." Retrieved August 2, 2006, from http://www.awea.org/news/wpa12.html.
Australian Greenhouse Office (2005). Advanced Electricity Storage Technologies Programme: Energy Storage Technologies: a review paper. Department of the Environment and Heritage, Commonwealth of Australia.
Benge, G. and E. G. Dew (2004). Meeting the Challenges of Design and Execution of Two High Rate Acid Gas Injection Wells. SPE/IADC Drilling Conference. Amsterdam, Netherlands.
Black, P. E. (2004). "Algorithms and Theory of Computation Handbook." "objective function", in Dictionary of Algorithms and Data Structures Retrieved November 13, 2006, from http://www.nist.gov/dads/HTML/objective.html
Blumer, D. J. (2006). Properties of Produced Water. SPE Petroleum Engineering Handbook. J. Fanchi, SPE.
Bullough, C., C. Gatzen, et al. (2004). Advanced Adiabatic Compressed Air Energy Storage for the Integration of Wind Energy. European Wind Energy Conference London, UK.
Bureau of Land Management. (2006). "Wyoming Wind Energy Project." Retrieved August 2, 2006, from http://www.wy.blm.gov/rfo/wind.htm.
Cape Wind. (2006). "Project at a Glance." Retrieved August 3, 2006, from http://www.capewind.org/article24.htm.
Cheung, K. Y. C., S. T. H. Cheung, et al. (2006). Large-Scale Energy Storage Systems. London, Imperial College.
Crotogino, F. (2006). Clarification of Huntorf Operations. J. Neumiller.
163
Crotogino, F., K.-U. Mohmeyer, et al. (2001). Huntorf CAES: More than 20 Years of Successful Operation. Orlando, Florida, U.S.A.
Danish Wind Industry Association. (2004). "Wind Energy Guided Tour." Retrieved June 11, 2006, from http://www.windpower.org.
DeJarnett, B. B., F. H. Lim, et al. (2003). Greater Green River Basin Production Improvement Project. U.S. Department of Energy's Federal Energy Technology Center. Fort Worth, TX.
Denholm, P., G. L. Kulcinski, et al. (2005). "Emissions and Energy Efficiency Assessment of Baseload Wind Energy Systems." Environmental Science Technology 39(6): 1903-1911.
EA Technology (2004). Review of Electrical Energy Storage Technologies and Systems and of Their Potential for the UK, DTI Technology Programme.
Fanchi, J. R. (2002). EZGEN - Generation of Flow Model Input.
Finn, T. M. (2005). Geothermal Gradient Map of the Southwestern Wyoming Province, Southwestern Wyoming, Northwestern Colorado, and Northeastern Utah. Petroleum Systems and Geologic Assessment of Oil and Gas in the Southwestern Wyoming Province, Wyoming, Colorado, and Utah. Denver, U.S. Geological Survey.
Flores, R. M. and L. R. Bader (1999). Fort Union Coal in the Greater Green River Basin, East Flank of the Rock Springs Uplift, Wyoming: A Synthesis. U.S. Geological Survey Professional Paper 1625-A, U.S. Geological Survey.
Forest Products Laboratory. (2004). "TechLine: Fuel Value Calculator." Retrieved November 10, 2006, from http://www.fpl.fs.fed.us/documnts/techline/fuel-value-calculator.pdf.
Gibson, R. I. (1997). "Greater Green River Basin." Retrieved October 16, 2006, from http://www.gravmag.com/grnriv.htm.
Gipe, P. (1995). Wind Energy Comes of Age. New York, John Wiley & Sons, Inc.
Gonzalez, A., B. O Gallachoir, et al. (2004). Study of Electricity Storage Technologies and Their Potential to Address Wind Energy Intermittency in Ireland. Rockmount Capital
164
Partners. Cork, Ireland, Sustainable Energy Research Group, Department of Civil and Environmental Engineering, University College Cork.
Greenblatt, J. B., S. Succar, et al. (2006). Baseload wind energy: Modeling the competition between gas turbines and compressed air energy storage for supplemental generation. Energy Policy.
Holst, K. (2005). The Iowa Stored Energy Plant: A Project Review and Update.
Katz, D.L., and M.R. Tek (1981). “Overview on Underground Storage of Natural Gas.” Journal of Petroleum Technology: 943-951.
Kirschbaum, M. A. and L. N. R. Roberts (2005). Geologic Assessment of Undiscovered Oil and Gas Resources in the Mowry Composite Total Petroleum System, Southwestern Wyoming Province, Wyoming, Colorado, and Utah. Petroleum Systems and Geologic Assessment of Oil and Gas in the Southwestern Wyoming Province, Wyoming, Colorado, and Utah. Denver, U.S. Geological Survey.
Makansi, J. (2001). "Energy Storage: The Sixth Dimension of the Electricity Production and Delivery Value Chain." Retrieved June 10, 2006, from http://www.energystoragecouncil.org/1%20-%20Jason%20Makansi-ESC.pdf.
Makansi, J. (2001). "Energy Storage: The sixth-and-missing link in the electricity value chain." Global Energy Business July/August 2001.
Manwell, J. F., J. G. McGowan, et al. (2002). Wind Energy Explained: Theory, Design, and Application. West Sussex, England, John Wiley & Sons Ltd.
McCain Jr., W. D. (1990). The Properties of Petroleum Fluids. Tulsa, OK, PennWell Publishing Company.
Miskimins, J. L. (2000). Characterization of Present-Day Stress States Near Faults, North LaBarge Field, Sublette County, Wyoming. Petroleum Engineering. Golden, Colorado School of Mines. Master of Science.
National Energy Technology Laboratory. (2004). "Transmission, Distribution, & Refining: Natural Gas Storage." Retrieved October 20, 2006, from http://www.netl.doe.gov/technologies/oil-gas/TDR/Storage/Storage.html.
National Renewable Energy Laboratory. (2000). "Wind Resource." Retrieved May 30, 2006, from http://www.nrel.gov/wind/wind_map.html.
165
NaturalGas.org. (2004). "Storage of Natural Gas." Retrieved October 13, 2006, from http://www.naturalgas.org/naturalgas/storage.asp.
NEG Micon North America. (2004). "Foote Creek 3." Retrieved August 2, 2006, from http://www.awea.org/projects/summaries/FooteCreek3.pdf.
Oil and Gas Investor (2005). "Going Gangbusters." Tight Gas: A Supplement of Oil and Gas Investor.
Paksoy, H. O. (2005). Underground Thermal Energy Storage - A Choice for Sustainable Future. Adana, Turkey, Cukurova University.
Patel, M. R. (2006). Wind and Solar Power Systems. Kings Point, New York, U.S.A, Taylor & Francis.
Platte River Power Authority. (2006). "Monthly Wind Speed and Performance Data 2004." Retrieved August 2, 2006, from http://www.prpa.org/energysources/windspeedperform04.htm.
Research Reports International (2004). Energy Storage Technologies For Electric Power Applications.
Ridge Energy Storage & Grid Services L.P. (2005). The Economic Impact of CAES on Wind in TX, OK, and NM, Texas State Energy Conservation Office.
Schilthuis, R. J. (1936). "Active Oil and Reservoir Energy." AIME 118: 33-37.
Schlumberger (2004). ECLIPSE Reference Manual.
Schlumberger (2004). ECLIPSE Technical Description.
Sonntag, R. E., C. Borgnakke, et al. (1998). Fundamentals of Thermodynamics. New York, John Wiley & Sons, Inc.
State of Texas (2006). New Electric Generating Plants in Texas Since 1995.
Thomas, G. W. (1982). Principles of Hydrocarbon Reservoir Simulation. Boston, International Human Resources Development Corporation.
166
Towler, B. F. (2006). Gas Properties. SPE Petroleum Engineering Handbook. J. Fanchi and Lake, SPE.
Wyoming Oil and Gas Commission. (2006). "Baxter Basin South Unit 23 Well File." Retrieved October 30, 2006, from http://wogcc.state.wy.us/Wellapino.cfm?napino=375562.
Wyoming Oil and Gas Commission. (2006). "Tip Top Unit T57X-27G Well Files." Retrieved October 25, 2006, from http://wogcc.state.wy.us/Wellapino.cfm?napino=3520807
167