advanced analytics to capture the full value of demand response and energy flexibility in industrial...

Advanced Analytics to Capture the Full Value of Demand Response and Energy Flexibility in

Industrial SitesESGI 2016

May 2016, Avignon

Sébastien Mouthuy, N-SIDE

Why and how to leverage flexibility from energy-intensive processes ?

2

Increasing complexity of the energy structure of the industries

Increasing need for Demand Response services

Increasing amount of data to be leveraged

Increasing electricity price volatility

0.00

20.00

40.00

60.00

80.00

100.00

120.00

EUR/MWh

S S M T W T F

Advanced Analytics to transform this complexity into opportunities

1. Demand Response: Opportunities and challenges for industrial sites

A. The markets: Where to value my flexibility ?B. The processes: Where to find and how to manage

my flexibility ?

2. Advanced Analytics to make the most out of energy flexibility

How to design an optimal energy flexibility strategy ?

5

Process Aspects:Where to find and how to manage my flexibility ?

MarketsWhere to value my

flexibility ?

ProcessesWhere to find and how to manage my flexibility

?

6

Strong Incentive to optimize planning

with respect to variable electricity

price

The Concept:Flexible process planning based on electricity prices

MarketsWhere to value my

flexibility ?

ProcessesWhere to find and how to manage my flexibility

?

• TOU (time of use)• Spot-based contracts• Balancing price impacted• …

• Oversized processes allowing to shift load

• Multi-product processes with difference of electricity consumptions

Variable electricity price

Electricity-Intensive process with flexibility

7

Years / Months in advance 1 to 7 days in advance Real-time

Day-ahead market

Spot-price based contracts

Intraday and balancing markets

Deviation penalties

Price based

Forward contracts (OTC)

Fixed contracts

Reserve

Reserve participation(e.g. France, Belgium)

Contract with an aggregator

Activation from TSO

Activation from aggregator

Reserve participation(e.g. Germany, Austria)

The market challenge: Where to value my flexibility ?

Direct Market AccessIndirect Market Access

Strategic Optimization

Reserve Optimization

Scheduling Optimization

Real-time Optimization

… on the different key timeframes

8

• Optimal electricity contract

• Optimal investment in flexibility assets

• Optimal choice of flexibility products and volumes

• Optimal power and energy price

• Optimal scheduling of electricity load

• Optimal planning of CHP unit

• Optimal imbalance minimization

• Optimal activation management

The Concept:Cement Industry Example

9

Integrated Cement Plant • 2.5 Million Tons of cement• Electricity Consumption: 300 GWh/Year

Kiln Feed Preparation

Clinker Production

Clinker Grinding

Continuous Process – ON/OFF


10

Clinker Production:• Main Process in the Cement

production• Continuous process• No possibility to stop • Load : 10 MW

Clinker Production

Mostly not flexible



11


Raw Mill:• Overcapacity:

40%• Max Load:

10 MW• ON-OFF principle• No possibility of modulation


12


Flexibility Lever 1: Optimization of the raw mill planning based on the electricity price by leveraging the overcapacity and the up/downstream storage

Raw Mill:• Overcapacity:

40%• Max Load:

10 MW• ON-OFF principle• No possibility of modulation


13


Cement Mill:• Overcapacity: 40 %• 3 Cement Mills: 100T/h• Max Load:

15 MW• ON/OFF principle• No possibility of modulation


14


Flexibility Lever 2: Optimization of the cement mill planning based on the electricity price by leveraging the overcapacity and the up/downstream storage

Cement Mill:• Overcapacity: 40 %• 3 Cement Mills: 100T/h• Max Load:

15 MW• ON/OFF principle• No possibility of modulation

The Concept:Paper Industry Example

15

Paper machine• Power:

15MW• Light overcapacity:

10%• Modulation possibility • Multiple papers • Consumption/ Paper type• Up/downstream stock• No ON/OFF• Paperweight: Light

• Load: 13 MW• 1/3 Production

• Paperweight: Heavy• Load: 17 MW• 1/3 Production

• Paperweight: Medium• Load: 15 MW• 1/3 Production

Continuous Process – Modulation


16

• Paperweight: Light• Load: 13 MW• 1/3 Production




Flexibility Lever 1: Modulate the load depending on electricity price



10%• Modulation possibility • Multiple papers • Consumption/ Paper type• Up/downstream stock• No ON/OFF


17

• Paperweight: Light• Load: 13 MW• 1/3 Production




Flexibility Lever 2: Schedule the different products depending on

electricity prices



10%• Modulation possibility • Multiple papers • Consumption/ Paper type• Up/downstream stock• No ON/OFF

The Concept:Different types of flexible planning…

18

Batch – ON/OFF Process

Monday Tuesday Wednesday Thursday Friday Saterday Sunday05

101520253035404550

Continuous – ON/OFF Process

012345678

Monday Tuesday Wednesday Thursday Friday Saterday Sunday

Continuous – Modulation Process

02468

1012141618



Multi-Product Process


The Concept:… with opportunities in various energy intensive industries

19

Continuous – ON/OFF Process

• Mechanical pulp production• Compressors in Air separation• Extrusion in chemical industry• Electrolysis

Batch – ON/OFF Process

• EAF in steel production• Pulp mixer in non-integrated paper mill

Continuous – Modulation Process

• Paper machine • Compressors in Air separation • Extrusion in chemical industry

Multi-Product Process

• Paper machine Schedule of different products

• EAF schedule of different products

The process challenge: Where to find and how to manage my flexibilities ?

20

Produce electricity at optimal moment

ElectricityGeneration

Electricity Consumption

Consume electricity at optimal moment

Load Shifting

Load Scheduling

Load SheddingBy-product Optimization

2

1

3 6

CHP Modulation5

Fuel Switching4

1. Demand Response: Opportunities and challenges for industrial sites

A. The market aspect: Where to value my flexibility ?B. The process aspect: Where to find and how to

manage my flexibility ?

2. Advanced Analytics to make the most out of energy flexibility

A mathematical model is key for considering all the factors…

22

Industrial processes• All input and output flows• Ramping up and down

• Min-max capacities• ON-Off procedure• Operating rates

Product Demand• Quantities and delivery dates• Must / May serve

Grid and market interaction• Different electricity

contracts (OTC, spot based)

• Capacity constraints

Storage facilities• Min-max capacities• Storage target

Economics• RM, NG and electricity costs• Revenue from sales• Opportunity costs• Fix and variable operating

costs• Incentive from DR programs

CHP Unit (boilers, turbines, …)• Min-max capacities• Operating rates

• Ramping up and down• ON-OFF procedure• Different steam pressure

levels

• Fuel mix constraints• Maintenance planning

…in an integrated way…

23

Electricity Generation

Electricity Consumption

24

Modelling industrial processes

Electricity consumption

Blend

x %

y %

z %

Bounds

Yield

Steady

ShutdownStartup

Profile Ramping

25

Mixed Integer Programming (MIP)for industrial processes

Modeling an industrial process• Mostly continuous variables to model quantities (e.g. flows between processes)• Some binary variables, to capture the discrete nature of some decisions (e.g. on-off status)• Usually linear(isable) constraints (e.g. piecewise linear representation of the yield of the

machines)

Finding the optimal set of decisions• Difficult problems: non-convex, no polynomial-

time algorithm• In practice, with a good solver, global optimum is

found within a few minutes• Widely used branch-and-bound algorithm:

recursive tree search of binary options

Linear relaxations

Some technical challenges arising from MIP

• Solving time may rise exponentially with the number of binary variables and the addition of coupling constraints

• Ill-conditioning and numerical difficulties are likely to arise with data of poor quality

• Some processes are better represented by non-linear constraints and integer variables (i.e. batch processes).

26

27

Modeling batch processes

pack

pack

pack

pack

pack

pack

urgent shipping

urgent shipping

to stock

Order 2241

Order 2242

Order 2243

Order 2244

Order 2245

Order 2246

to stock

to stockto stock

Only one packing machine

Precedence Complete urgent orders first

MWh

28

Constraint Programming (CP) for batch production

Modeling a production unit of a batch process• Only discrete variables, to model machines performing activities as well as start

and end time of activities• Need for disjunctive constraints: at most one activity scheduled on a given

machine at any time, stock evolution, setup times,…• Complex precedence and transition constraints

Finding the optimal set of decisions• Difficult problems: highly non-linear, no known

polynomial-time algorithm• In practice, with a good solver, very good solution

is found within a few minutes• Widely used propagation algorithms: do not

explore decisions leading to a dead-end

Propagation

29

Some Technical challenges arising from CP

• Algorithmic efficiency strongly depends on how production has been modelled using CP constraints– Several models are possible– Good model may be millions of time faster than bad ones– Requires expertise

• Constraint programming enables to exploit knowledge humans have of the production process– Advantage: leads to more efficient algorithms– Disadvantage: no automatic configuration with no brain efforts– Advange by far worth the effort

30

The resulting integrated optimization problem is solved using decomposition techniques

Production Scheduling

Electricity Management

• Start of batch production• Machine usage• HR planning Compute electricity

demand

Identify mismatchproduction <-> electricity cost • Cogen management

• When to turn on/off• Needs of by-products

• Total energy cost

Optimality

31

The model should take into account multiple markets and time frames



Choose which processes to stop in response to an activation

Day-ahead Market

ReserveMarket

Bid interruptible consumptions on the reserve market

Nominations on DAM Respect balance

32

Real-time activation is uncertain when committing a schedule



Day-ahead Market

ReserveMarket

33

Multi-Stage Stochastic Programming (MSSP)for multi-market optimization

Multi-market modelling• Multiple real-time scenarios should be considered• Non-anticipativity constraints: scheduling decisions should be taken

commonly for all real-time scenarios

Finding the optimal set of decisions• Difficult problems: non-convex, no polynomial-

time algorithm• In practice, with a good solver, global optimum is

found within a few minutes• Widely used decomposition algorithm: solve a

reduced problem and add only the violated constraints

Reduced problems

34

Technical challenges arising from MSSP

• Problem size rises exponentially with the number of scenarios.

• Generating the minimal number of relevant scenarios is crucial

35

Helping industries get the full value of their energy flexibility thanks to advanced analytics

Produce electricity at optimal moment

Optimal Planning of energy production

Optimal Scheduling of energy consumption

Consume energy at optimal moment

Inte

grat

ed O

ptim

izatio

n

Thank you !

advanced analytics to capture the full value of demand response and energy flexibility in industrial...

Technology