refcl trial: ignition tests - energy · 2017-03-10 · this report contains test results,...

ABN 88 151 368 964

PO Box 7092, Beaumaris VIC 3193

phone +61(0) 412 136 041

email [email protected]

REFCL Trial: Ignition Tests

4 August 2014

mailto:[email protected]

P a g e | 2

© Marxsen Consulting Pty Ltd Monday, 4 August 2014

Disclaimer

This report outlines the results of tests carried out for the Powerline Bushfire Safety Program at a purpose-built facility in Frankston Victoria in the first half of 2014 in accordance with an Agreement between Marxsen Consulting Pty Ltd and the Victorian Department of State Development, Business and Innovation. This report contains test results, observations, analysis, commentary and interpretation.

Subject to the Agreement, no warranty can be offered to third parties for:

• The application of anything in this report for any purpose other than those required by the specific objectives of the test program stated in the body of this report.

• The direct application of anything contained in this report to any situation other than those that were recorded in the tests.

A complete set of test records is available in the public domain or (in the case of very large video files) upon request from the Powerline Bushfire Safety Program. Readers are advised to rely on their own analysis of these records if they wish to use this report for any purpose other than the specific objectives of the test program stated in the body of this report.

Readers should in particular note the following qualifications:

• The information in this report relates to ‘wire on ground’ powerline earth faults only. Readers who wish to use these results to derive conclusions for other types of network earth fault should rely on their own investigations.

• Definitions of worst case fire risk conditions for ignition were derived from limited data and small numbers of tests. No warranty is offered that even worse fire risk condition will not occur in practice.

• All reasonable care has been taken to clearly outline the rationale and evidence for findings, but readers should make their own judgements of the merits such findings before relying on them.

• Where a level of statistical uncertainty is stated, this is a statistical measure which cannot be applied to individual cases or small numbers of cases, but can only be validly applied to cohorts of cases large enough to meet the normal criteria set out in statistical theory.

• Quantification of statistical certainty has not been possible for many findings due to test-to-test variation of factors that influence ignition outcomes. In such cases, readers should form their own judgement of the level of confidence they can place on the findings concerned.

• Definition of typical and worst case ‘wire on ground’ faults has been done with the sole purpose of illustration of one method for potential application of test results to real situations. Readers should rely on their own investigations to define ‘wire on ground’ faults to which they wish to apply the tests results and to define the method they use to do so.

• Many assumptions were used to generate insights, derive findings and interpret results. All reasonable care has been taken to explicitly document these assumptions and explain the rationale in each case, but no warranty can be offered that such documentation is complete or that any implicit or explicit assumptions used are valid.

• Where mathematical theory has been used to derive insights from test results, care has been taken to outline the theory and how it was applied. However, no warranty is offered that the theory employed is valid or correctly applied.

Advice to the reader

Readers who are not familiar with REFCL technology are advised that prior reading of Appendices A and B may provide potentially useful perspective and background to assist a sound understanding of the findings and analysis set out in this report.

P a g e | 3


Contents

1 EXECUTIVE SUMMARY ............................................................................................................................ 7

1.1 THE REFCL TRIAL PROJECT ............................................................................................................................ 7 1.2 TESTS CONFIRM REFCLS CUT FIRE RISK COMPARED TO TRADITIONAL TECHNOLOGY....................................................... 7 1.3 FINDINGS FROM THE TEST PROGRAM ................................................................................................................ 8 1.4 OPTIMUM SETTINGS FOR REFCLS TO MINIMISE FIRE RISK ...................................................................................... 9 1.5 CHALLENGES FACING NETWORK OWNERS IN ADOPTION OF REFCLS ........................................................................ 10 1.6 KEY DESIGN DECISIONS FOR THE REFCL TRIAL .................................................................................................. 10

2 THE PROJECT ......................................................................................................................................... 11

2.1 GENESIS AND OBJECTIVES ............................................................................................................................ 11 2.2 GOVERNANCE .......................................................................................................................................... 11 2.3 PROJECT TIMELINE ..................................................................................................................................... 12 2.4 PROJECT TEAM ......................................................................................................................................... 13

3 ‘WIRE ON GROUND’ FIRE RISK: COMPARATIVE NER/ASC/GFN ANALYSIS .............................................. 14

3.1 CATEGORIES OF IGNITION AND FAULTS ............................................................................................................ 14 3.1.1 Ignition categories ..................................................................................................................... 14 3.1.2 Earth fault categories ................................................................................................................ 15

3.2 RELATING TEST RESULTS TO REAL NETWORK EARTH FAULTS ................................................................................... 16 3.2.1 Fault current and soil current ..................................................................................................... 16 3.2.2 The key enabling assumption: current proportional to conductor length ..................................... 17 3.2.3 Real network ‘wire on ground’ faults on Code Red days.............................................................. 18 3.2.4 Bounce ignition – worst case real network fault ......................................................................... 18 3.2.5 Ground ignition – worst case real network fault ......................................................................... 19 3.2.6 Relating test results to real network faults ................................................................................. 20

3.3 NETWORK FIRE RISK LANDSCAPE – INHERENT FIRE RISK ........................................................................................ 21 3.4 FIRE RISK EQUATION ................................................................................................................................... 22

3.4.1 Time to detect ........................................................................................................................... 22 3.4.2 Time to act ................................................................................................................................ 23 3.4.3 Time to ignite - new data from test results ................................................................................. 23

3.5 NETWORK FIRE RISK– RESIDUAL RISK WITH EXISTING (NON-REFCL) PROTECTION ....................................................... 25 3.6 NETWORK FIRE RISK– RESIDUAL RISK WITH ASC PROTECTION (REFCL VARIANT)........................................................ 26

3.6.1 ASC response depends on fault current ...................................................................................... 26 3.6.2 Typical ASC protection systems were not tested ......................................................................... 27

3.7 NETWORK FIRE RISK– RESIDUAL RISK WITH GFN PROTECTION (REFCL VARIANT) ....................................................... 27 3.7.1 Ground ignitions with a GFN ...................................................................................................... 28 3.7.2 Bounce ignitions with a GFN ...................................................................................................... 29

3.8 NETWORK FIRE RISK– RESIDUAL RISK WITH IMPROVED GFN (REFCL VARIANT) ......................................................... 29 3.8.1 Increased sensitivity with improved tolerance for network imbalance......................................... 30 3.8.2 Faster onset of residual compensation ....................................................................................... 30 3.8.3 Fast reliable fault-confirmation and faulted-feeder identification ............................................... 30 3.8.4 Automated RCC tuning .............................................................................................................. 31

3.9 CONCLUSION: COMPARATIVE FIRE RISK ANALYSIS ............................................................................................... 31

4 FINDINGS .............................................................................................................................................. 32

4.1 FINDING 1: IN WORST CASE FIRE CONDITIONS, ‘WIRE ON GROUND’ POWERLINE FAULTS ON NETWORKS WITH TRADITIONAL NON-REFCL PROTECTION CREATE INHERENT RISK OF FIRE ......................................................................................... 33

4.1.1 Rationale ................................................................................................................................... 33 4.1.2 Evidence .................................................................................................................................... 33

4.2 FINDING 2: EXISTING NON-REFCL PROTECTION SCHEMES HAVE THE POTENTIAL TO PREVENT SOME FIRES BUT CANNOT ELIMINATE THE MAJORITY OF FIRE RISK FROM ‘WIRE ON GROUND’ FAULTS ..................................................................... 35

4.2.1 Rationale ................................................................................................................................... 35 4.2.2 Evidence .................................................................................................................................... 35

P a g e | 4


4.3 FINDING 3: REFCLS DRAMATICALLY REDUCE ENERGY RELEASE INTO THE ENVIRONMENT FROM ‘WIRE ON GROUND’ POWERLINE FAULTS ......................................................................................................................................................... 39

4.3.1 Rationale ................................................................................................................................... 39 4.3.2 Evidence .................................................................................................................................... 40 4.3.3 Bounce ignition ......................................................................................................................... 41 4.3.4 Ground ignition ......................................................................................................................... 43

4.4 FINDING 4: REFCLS CAN DETECT AND RESPOND TO ‘WIRE ON GROUND’ POWERLINE FAULTS THAT TRADITIONAL NON-REFCL NETWORK PROTECTION DOES NOT ‘SEE’ ................................................................................................................ 46

4.4.1 Rationale ................................................................................................................................... 46 4.4.2 Evidence .................................................................................................................................... 47 4.4.3 Normal sensitivity setting .......................................................................................................... 47 4.4.4 Heightened sensitivity setting .................................................................................................... 47 4.4.5 Opportunities to increase fault detection sensitivity ................................................................... 49 4.4.6 Factors that affect fault detection sensitivity.............................................................................. 49

4.5 FINDING 5: REFCLS CAN SIGNIFICANTLY REDUCE FIRE RISK FOR A WIDE RANGE OF ‘WIRE ON GROUND’ POWERLINE FAULTS. . 51 4.5.1 Rationale ................................................................................................................................... 51 4.5.2 Evidence .................................................................................................................................... 51 4.5.3 Reduction of bounce ignition risk ............................................................................................... 54 4.5.4 Reduction in ground ignition risk ................................................................................................ 55 4.5.5 The threshold current for ignition............................................................................................... 57 4.5.6 Faulted-phase earthing (FPE) – an untested REFCL type ............................................................. 57

4.6 FINDING 6: THERE ARE SOME ‘WIRE ON GROUND’ POWERLINE EARTH FAULTS WHERE TODAY’S REFCL PRODUCTS MAY NOT PREVENT IGNITION .......................................................................................................................................... 58

4.6.1 Rationale ................................................................................................................................... 58 4.6.2 Evidence .................................................................................................................................... 58 4.6.3 Bounce ignition with the REFCL in service ................................................................................... 58 4.6.4 Ground ignition with the REFCL in service ................................................................................... 62

4.7 FINDING 7: THOUGH BOTH REFCL VARIANTS REDUCE FIRE RISK, GFNS OFFER SUPERIOR FIRE RISK REDUCTION BENEFITS COMPARED TO ASCS ....................................................................................................................................... 67

4.7.1 Rationale ................................................................................................................................... 67 4.7.2 Evidence .................................................................................................................................... 68 4.7.3 Limitations of ASC ignition tests ................................................................................................. 70 4.7.4 Typical ASC protection systems were not tested ......................................................................... 71

4.8 FINDING 8: REFCL DESIGNS CAN BE IMPROVED TO FURTHER REDUCE FIRE RISK ......................................................... 72 4.8.1 Rationale ................................................................................................................................... 72 4.8.2 Evidence .................................................................................................................................... 72 4.8.3 Increased sensitivity with improved tolerance for standing network imbalance .......................... 73 4.8.4 Faster residual compensation .................................................................................................... 75 4.8.5 Automated RCC tuning .............................................................................................................. 75 4.8.6 Faster, more reliable, fault-confirmation and faulted-feeder identification ................................. 76

4.9 FINDING 9: REFCLS OFFER BENEFITS TO PUBLIC SAFETY ...................................................................................... 77 4.9.1 Rationale ................................................................................................................................... 77 4.9.2 Evidence .................................................................................................................................... 77 4.9.3 Public safety benefits of a GFN................................................................................................... 79 4.9.4 Public safety benefits of an ASC ................................................................................................. 80 4.9.5 Overseas experience .................................................................................................................. 81

4.10 FINDING 10: REFCLS OFFER BENEFITS TO SUPPLY RELIABILITY ............................................................................. 82 4.10.1 Rationale .............................................................................................................................. 82 4.10.2 Solidly earthed and NER-based network protection philosophy .............................................. 82 4.10.3 REFCL-based network protection philosophy .......................................................................... 82 4.10.4 Evidence ................................................................................................................................ 83

5 OPTIMAL OPERATIONAL SETTINGS FOR REFCL FIRE RISK BENEFITS ....................................................... 85

5.1 FULLY EXPLOIT MODERN TECHNOLOGY IN NON-REFCL PROTECTION SYSTEMS ........................................................... 86 5.2 MONITOR TRANSIENT FAULTS IN REFCL NETWORKS TO ASSESS RELEVANCE TO FIRE RISK .............................................. 86 5.3 FOR SUSTAINED FAULTS IN REFCL NETWORKS TRIP THE FAULTED FEEDER AND DO NOT RECLOSE .................................... 87 5.4 TEMPORARILY INCREASE REFCL FAULT DETECTION SENSITIVITY ON HIGH FIRE RISK DAYS .............................................. 87

P a g e | 5


5.5 PROMOTE CONTINUED DEVELOPMENT OF THE GFN FAULT CONFIRMATION TEST ....................................................... 88 5.6 CONFIRM NETWORK ‘HARDENING’ PRIOR TO EACH FIRE SEASON ............................................................................ 89 5.7 CALIBRATE RCCS REGULARLY AND BEFORE EACH FIRE SEASON ............................................................................... 89 5.8 PROVE FIRE PERFORMANCE BY REAL TESTS ....................................................................................................... 90

6 REFCL IMPLEMENTATION CHALLENGES ................................................................................................. 91

6.1 CHALLENGE 1: LEARN BY DOING- CULTURE CHANGE FOR NETWORK OWNERS AND SUPPLIERS ........................................ 92 6.2 CHALLENGE 2: HARDEN NETWORKS TO REDUCE RISK OF CROSS-COUNTRY FAULTS ....................................................... 93 6.3 CHALLENGE 3: UPGRADE NETWORKS TO REFCL-COMPATIBLE EQUIPMENT............................................................... 94 6.4 CHALLENGE 4: MINIMISE NETWORK IMBALANCE ................................................................................................ 95 6.5 CHALLENGE 5: FAULT LOCATION .................................................................................................................... 96

7 REFCL TRIAL DESIGN .............................................................................................................................. 97

7.1 SELECTION OF FIRE CAUSE AND FAULT GEOMETRY .............................................................................................. 97 7.1.1 Fire cause analysis: 2010 national survey ................................................................................... 97 7.1.2 Fire cause analysis: Victorian data 2004-2009 ............................................................................ 97 7.1.3 Fire cause analysis: Victorian data 2011-2014 ............................................................................ 98 7.1.4 The 2011 Powerline Bushfire Safety Taskforce arc-ignition research ........................................... 99 7.1.5 Selection of powerline fire cause for the REFCL Trial ................................................................... 99 7.1.6 Selection of fault geometry for the REFCL Trial ......................................................................... 100

7.2 DEFINITION OF UNDERLYING PRINCIPLES ........................................................................................................ 101 7.3 SELECTION OF CONDUCTOR TYPE, CONDUCTOR IMPACT SPEED AND BOUNCE HEIGHT ................................................. 101 7.4 SELECTION OF SOIL, FUEL AND SOIL/FUEL BED GEOMETRY .................................................................................. 103

7.4.1 Selection of soil........................................................................................................................ 103 7.4.2 Selection of fuel ....................................................................................................................... 104 7.4.3 Soil/fuel bed design ................................................................................................................. 104

7.5 SELECTION OF WORST CASE FIRE WEATHER CONDITIONS AND FUEL MOISTURE CONTENT............................................. 104 7.6 DESIGN OF TEST FACILITY ........................................................................................................................... 105

7.6.1 Site selection ........................................................................................................................... 105 7.6.2 Site concept and construction .................................................................................................. 108 7.6.3 Safety architecture .................................................................................................................. 110 7.6.4 High voltage supply ................................................................................................................. 111 7.6.5 Test rig .................................................................................................................................... 113 7.6.6 Simulation of additional conductor length in ignition tests ....................................................... 115 7.6.7 Control and protection systems................................................................................................ 117 7.6.8 Data acquisition systems ......................................................................................................... 118

7.7 TEST PROCEDURE .................................................................................................................................... 119 7.7.1 Test run procedure .................................................................................................................. 120

7.8 TEST SETTINGS ........................................................................................................................................ 120

8 APPENDICES ........................................................................................................................................ 122

8.1 APPENDIX A: HISTORY OF DISTRIBUTION NETWORK EARTHING PRACTICES .............................................................. 123 8.2 APPENDIX B: OPERATION OF REFCLS .......................................................................................................... 128 8.3 APPENDIX C: TEST RECORDS....................................................................................................................... 132

8.3.1 Valid ignition tests ................................................................................................................... 132 8.3.2 Bolted fault tests ..................................................................................................................... 137 8.3.3 Setup and invalid tests ............................................................................................................. 139

8.4 APPENDIX D: IGNITION FROM FULGURITE FORMATION ...................................................................................... 143 8.4.1 Fulgurite formation in NER tests .............................................................................................. 143 8.4.2 Test 208: anomalous fulgurite formation in a GFN test ............................................................. 145 8.4.3 Test 217 – fulgurite formation due to a cross-country fault ...................................................... 147

8.5 APPENDIX E: HRL TECHNOLOGY REFCL TRIAL REPORT ..................................................................................... 149

P a g e | 6


List of abbreviations

Acronym Explanation

22kV 22,000 volts – the ‘wire to wire’ voltage on most of Victoria’s electricity networks

50Hz 50 cycles per second – the frequency of electricity distributed in Victoria

ACR Automatic Circuit Recloser – a remote control pole-mounted high voltage switch

amp, A Amperes – the unit used to measure flow of electric current

ASC Arc Suppression Coil – a REFCL component used in all resonant earthing schemes

CPR Cardio-pulmonary resuscitation – first aid for cases of low voltage electrocution

DBRG The Distribution Business Reference Group – network owner senior executives

DSDBI The Victorian Department of State Development, Business and Innovation

ESV Energy Safe Victoria – Victoria’s energy safety regulator

FPE Faulted Phase Earthing – a type of REFCL used mainly in Ireland

GFN Ground Fault Neutraliser – a REFCL product manufactured by Swedish Neutral AB

IEF Instantaneous Earth Fault – a protection scheme that operates for high fault currents

ITDEF Inverse Time Delayed Earth Fault – a medium fault current protection scheme

MAIFI A standard index to measure the frequency of momentary supply interruptions

NER Neutral Earthing Resistor – a non-REFCL network earthing approach used in Victoria

ohm The unit of measurement of electrical resistance (ratio of voltage to current)

PBST, PBSP Powerline Bushfire Safety Taskforce (finished in 2011) and Program (current)

RCC Residual Current Compensator – a component of the GFN product

REFCL Rapid Earth Fault Current Limiter – a technology that quickly limits earth fault current

SAIDI A standard index to measure average duration of sustained supply interruptions

SAIFI A standard index to measure average frequency of sustained supply interruptions

SEF Sensitive Earth Fault protection scheme for low fault currents

TWG The REFCL Trial Technical Working Group

VESI The Victorian Electricity Supply Industry

P a g e | 7


1 Executive summary Rigorous tests on a real electricity distribution network have confirmed that Rapid Earth Fault Current Limiter (REFCL) technology can reduce the fire risk associated with bare-wire overhead powerlines. Tests confirmed that when a live high voltage conductor falls to the ground under worst case fire weather conditions such as those experienced on Black Saturday 2009, a REFCL can reduce the conductor-soil arcing in many circumstances to levels below that required to start a fire.

1.1 The REFCL trial project In 2011, the Powerline Bushfire Safety Taskforce (PBST) recommended that, subject to further trials on a real network to confirm their effectiveness in reducing fire risk, REFCLs be installed on the Victorian distribution network. The PBST estimated REFCLs could cut total powerline fires in Victoria by around 50 per cent.

In response to the PBST recommendation, the Powerline Bushfire Safety Program (PBSP) established the REFCL Trial project to rigorously test REFCL technology on a real network with the following objectives:

• Determine whether REFCL technology is effective in reducing fire starts from electric arcs in powerline faults on a real multi-wire 22kV network; and,

• Determine the optimum operational settings for REFCLs to reduce fire starts initiated by electric arcs in powerline faults.

The trial aimed to quantify the powerline fire risk reduction benefits of REFCL technology in high fire risk areas of Victoria under worst case fire risk conditions. This includes the relative benefits of different REFCL variants, specifically the generic Arc Suppression Coil (ASC) and the proprietary Ground Fault Neutraliser (GFN). Other comparatively rare REFCL variants were not tested.

Following 12 months of planning and preparation, during the first half of 2014 a field test facility was designed and built near Frankston Victoria and a comprehensive research program of 259 tests, including 118 ignition tests under rigorously controlled conditions, was carried out on the only Australian public electricity distribution network protected by a REFCL.

1.2 Tests confirm REFCLs cut fire risk compared to traditional technology The test program confirmed that under worst case fire conditions, ‘wire on ground’ powerline earth faults on Victorian rural networks with traditional network protection systems create an inherent risk of fires and that this risk is markedly reduced in a network protected by a REFCL. The test program revealed there are two ways a ‘wire on ground’ fault can start a fire in worst case fire conditions – very fast ‘bounce ignition’ and slower ‘ground ignition’ – and provided a REFCL is present, both are much less likely to occur.

The research program also demonstrated that these two potential ignition risks might be further reduced or entirely eliminated through design improvements to REFCL technology, which until now has never been optimised to deliver minimum fire risk.

Comparative analysis of fire risk performance based on test results led to the following conclusions:

1. Even some of today’s non-REFCL network protection systems can deliver potentially valuable reductions in fire risk from powerline ‘wire on ground’ earth faults.

2. The different network protection schemes tested in the research deliver different levels of fire risk reduction. In order of highest to lowest fire risk, they are listed below.

P a g e | 8


Network protection scheme Fire risk from ‘wire on ground’ powerline faults

Non-REFCL (traditional existing) Highest

REFCL variant: ASC

REFCL variant: GFN (today’s product)

REFCL: GFN (with improvements) Lowest

3. Whilst tests showed today’s GFN product delivers lower fire risk than a broadly equivalent ASC-based scheme, the test program did not succeed in quantifying the difference between these two variants of REFCL technology.

4. An improved GFN design appears capable of virtually eliminating fire risk from ‘wire on ground’ faults in worst case fire conditions.

A key assumption of uniform conductor-soil current spread along the length of a fallen conductor was used in this comparative analysis. Further, specific worst case ‘wire on ground’ faults were selected to apply test results to real network faults in order to describe overall network fire risk. Network owners can use local knowledge to vary this assumption and this selection to suit their circumstances. However, while such variations might lead to adjustment of their estimate of the overall level of network fire risk, they would be unlikely to change the ranking shown above.

1.3 Findings from the test program The specific findings of the research program are:

1. In worst case fire conditions, ‘wire on ground’ powerline faults on networks with traditional non-REFCL protection create an inherent risk of fire. There is a very low threshold level of current into soil above which ignition is close to 100 per cent probable in today’s non-REFCL networks.

2. Existing non-REFCL Sensitive Earth Fault (SEF) protection schemes have the potential to prevent some fires. However, they cannot eliminate the majority of fire risk from ‘wire on ground’ powerline faults. Other over-current earth fault protection schemes have limited if any, potential to cut fire risk.

3. REFCLs dramatically reduce energy release into the environment from ‘wire on ground’ powerline faults. They collapse the voltage on the fallen conductor to reduce fault current, reduce arc power and bring about faster arc self-extinction. Tests with and without a REFCL vividly demonstrate this dramatic reduction of arc energy.

4. REFCLs can detect and respond to ‘wire on ground’ powerline faults that traditional non-REFCL network protection cannot ‘see’. For supply reliability and security purposes, traditional SEF protection is usually set to detect nine amps of fault current in rural networks. REFCLs detect two amps and tests demonstrated detection of less than one amp.

5. REFCLs can significantly reduce fire risk for a wide range of ‘wire on ground’ powerline faults. By detecting earth faults that traditional protection systems cannot ‘see’ and by dramatically reducing energy released into the local environment when earth faults occur, REFCLs reduce the chance of ignition across a wide range of earth faults.

6. There are some ‘wire on ground’ powerline earth faults where today’s REFCL products may not prevent ignition. High current faults may result in bounce ignition before a REFCL has time to reduce the fault current. The reduction in residual fault current may not be sufficient to completely remove the risk of slower ground ignition.

P a g e | 9


7. Though both REFCL variants reduce fire risk, GFNs offer superior fire risk reduction benefits compared to ASCs. Test results indicate GFNs reduce bounce ignition risk more than ASCs. The residual current with an ASC is higher than with a GFN. However, today’s GFN performs fault-confirmation and faulted feeder identification tests a few seconds after the fault and these tests require current flow that can sometimes be sufficient for ignition.

8. REFCL designs can be improved to further reduce fire risk. Four specific REFCL design improvement opportunities were identified that in total have the potential to eliminate fire risk from ‘wire on ground’ faults in worst case conditions. These opportunities apply more to GFNs than ASCs. They are:

a. Increased fault detection sensitivity with increased tolerance for network imbalance b. Faster residual current compensation c. More accurate residual current compensation d. Fast reliable fault-confirmation and identification of the faulted-feeder.

9. REFCLs offer benefits to public safety. REFCLs quickly reduce the voltage on a fallen conductor and can potentially transform high voltage electrocution risk of irreversible serious internal and external burns, to low voltage electrocution risk of reversible injury that is responsive to immediate first aid, especially CPR. A GFN has the potential to reduce voltage on a fallen conductor to levels where even low voltage electrocution risk is low.

10. REFCLs offer benefits to supply reliability. Improved supply reliability is a major motivator of utilities’ adoption of REFCLs around the world. Experience at Frankston South supports published studies that show substantial improvements in reliability indices such as SAIDI and MAIFI following REFCL installation.

1.4 Optimum settings for REFCLs to minimise fire risk Whilst it is not appropriate to prescribe detailed settings for REFCL systems as these must be set to suit local circumstances, the following considerations and recommendations are offered for those charged with responsibility for protection settings in Victoria’s rural networks when minimum fire risk is to be achieved:

1. Fully exploit modern technology in non-REFCL protection systems - ensure the capabilities of modern digital earth fault relays are used to provide the fastest most sensitive earth fault protection possible on days of extreme fire risk.

2. Monitor transient faults in REFCL networks to assess their relevance to fire risk - detect any instances of transient earth faults that cause ignition and review settings accordingly.

3. For sustained faults on REFCL networks trip the faulted feeder and do not reclose – reclose onto a permanent earth fault serves little or no purpose with a REFCL in service on days of extreme fire risk.

4. Temporarily increase REFCL fault detection sensitivity on high fire risk days – ensure as many earth faults as possible are detected and quickly addressed by REFCL systems on days of extreme fire risk.

5. Promote continued development of the GFN fault confirmation test – address the goal of reliable detection of a permanent fault and identification of the faulted feeder at current levels low enough to prevent fires.

6. Confirm network ‘hardening’ prior to each fire season – confirm by direct test that networks supplying high fire risk areas can withstand REFCL-induced voltage variations without equipment failures that constitute cross-country faults that start fires.

7. Calibrate RCCs regularly and before each fire season – ensure residual current compensation delivers minimum possible residual voltages and currents on faulted feeders on days of extreme fire risk.

8. Prove fire performance by real tests – verify all REFCL product developments and design improvements using rigorous objective ignition tests.

P a g e | 10


1.5 Challenges facing network owners in adoption of REFCLs For Victoria’s network owners to adopt REFCL technology, they must address a number of cultural and technical challenges. Wide-scale roll-out of REFCLs is likely to take at least a decade if the risks posed by these challenges are to be properly managed. Some of the challenges are:

1. Learn by doing – culture change for network owners and suppliers: The thinking patterns among engineering and operations staff required to get full value from REFCL technology are profoundly different from those that are prevalent today. Suppliers also face major challenges in understanding the priority of fire risk goals in Victoria and the implications for their products.

2. Harden networks to reduce risk of cross-country faults: Vulnerabilities to the over-voltages created by REFCL responses to earth faults must be identified and addressed so cross-country faults do not disrupt customer supply and start fires.

3. Upgrade networks to REFCL-compatible equipment: Many items of network equipment and network protection systems do not work with REFCLs and they must be upgraded or replaced to become REFCL-compatible.

4. Minimise network imbalance: Future REFCL products may be more tolerant of network imbalance (different capacitance to earth from each of the three phases). However, today’s products are not and network owners deploying REFCLs must act to minimise imbalance to achieve minimum fire risk.

5. Fault location: With a REFCL in service many faults draw so little current they leave no evidence of their presence, i.e. they are hard to find. This is a complex challenge and new products are emerging to address it.

Experience at Frankston South has demonstrated that these challenges can be successfully addressed over time. However, Frankston South is a single installation supplying a peri-urban area and Victoria has more than 100 zone substations that supply rural areas, some using long-established technical designs that are not present in the Frankston South network.

At the time of writing this report, two more Victorian network owners are preparing to install REFCLs in zone substations, so the journey towards wider adoption has already started. Both overseas and local experience indicates it will take at least several years before REFCL technology is a fully accepted and smoothly functioning part of Victoria’s electricity distribution infrastructure.

1.6 Key design decisions for the REFCL Trial For completeness, this report documents some key decisions made for the REFCL Trial. These drew upon the PBST’s 2011 arc-ignition tests, proof-of-concept arc-ignition tests carried out in February 2014 and separate experimental investigations in early 2014 into conductor falls and wind speeds at low heights. The key decisions included:

1. Selection of single phase ‘wire on ground’ earth fault as the fire cause to be investigated. 2. Selection of sideways (flat) conductor-soil impact as the fault geometry 3. Selection of conductor type, conductor impact speed and bounce height 4. Selection of soil, fuel and soil/fuel bed geometry 5. Selection of worst case fire weather conditions and fuel moisture content

The high level conceptual designs for the test site and test rig are also documented, together with a sample test procedure and test settings.

P a g e | 11


2 The project The REFCL Trial was established by the Powerline Bushfire Safety Program (PBSP) as a limited duration research project.

2.1 Genesis and objectives In September 2011, the Powerline Bushfire Safety Taskforce (PBST) recommended that, subject to further trials on a real network to confirm their effectiveness in reducing fire risk, REFCLs be installed on the Victorian distribution network. The Taskforce’s arc-ignition research had, based on simulations of REFCL responses to powerline earth faults, provided indicative evidence that REFCLs might reduce energy release sufficiently to prevent fires under worst-case conditions.

The PBST estimated installation of a REFCL could reduce the likelihood of earth faults on multi-wire (three phase and single phase) powerlines starting bushfires by 70 per cent. Since multi-wire powerlines make up about two thirds of all rural powerlines (the remainder being SWER powerline which do not benefit from REFCLs), this could (if tests confirm anticipated benefits) cut total powerline fires by around 50 per cent.

Although installation of each REFCL (one per zone substation) including all requisite ancillary works is relatively expensive (estimates range from $4 million to $15 million), the delivered fire risk reduction benefit per dollar spent is comparatively attractive because each REFCL can provide protection against earth fault fires to all multi-phase (22kV) lines in an entire substation network – on average 400km of powerline route length per REFCL.

In December 2011, the Victorian Government established the PBSP for the express purpose of implementing:

• Recommendations 27 and 32 of the 2009 Victorian Bushfires Royal Commission (VBRC); and, • Six recommendations of the PBST that gave specific direction to the VBRC

recommendations.

The PBSP established the REFCL Trial project to rigorously test REFCL technology on a real network. The objectives of the REFCL trial are to:

1. Determine whether the REFCL is effective in reducing fires started by electric arcs in single phase to earth faults on multi-wire 22kV powerlines; and,

2. Determine the optimum operational settings for the REFCL for reducing fires started by electric arcs in powerline faults.

The trial aimed to quantify the benefits of REFCL technology in reducing the risk of powerlines starting bushfires in high fire risk areas of Victoria. This includes the comparative benefits of different REFCL variants.

2.2 Governance The PBSP team within the Department of State Development Business and Innovation (DSDBI) was given oversight of the REFCL Trial with the support of Energy Safe Victoria (ESV) and each of the electricity distribution businesses in the Victorian Electricity Supply Industry (VESI). A Technical Working Group (TWG) comprising representatives from each of these organisations supported the trial.

Arrangements for project governance are illustrated in Figure 1. The identities of the key roles in the Trial were:

• Lead Researcher: Dr Tony Marxsen of Marxsen Consulting Pty Ltd • Research team: HRL Technology Pty Ltd

P a g e | 12


• Host distribution network service provider business: United Energy Ltd • Independent Technical Advisor: Professor Alex Baitch of BES (Aust) Pty Ltd

Figure 1: REFCL Trial - governance arrangements

The Lead Researcher had operational control over the research program, including responsibility for managing consortium partners and direct engagement with the host distribution business for each stage of the testing, although all contracts were made directly with the PBSP team in DSDBI. The host distribution business had operational decision-making power on all aspects of the work that had potential to affect the broader distribution network and its connected customers.

Technical oversight was exercised in two forms:

• The TWG undertook a technical advisory role at key program milestones, providing advice to the PBSP Distribution Business Reference Group (DBRG), consisting of higher level representatives of each of the distribution businesses, ESV and the PBSP team.

• The Independent Technical Advisor reported to ESV on progress to provide additional assurance to Government that the research objectives were being fully met.

The PBSP Team, ESV and the DBRG exercised oversight of delivery against key milestones and authorised any strategic changes to the research program as required. The PBSP team in turn reported to a Program Control Board, which sponsored the research program and was the ultimate decision-maker.

2.3 Project timeline The REFCL Trial ran for 11 months from issue of the Request for Quotation for the position of Lead Researcher in July 2013 to completion of field test activity in June 2014. The timeline of project activities is shown in Figure 2.

P a g e | 13


Figure 2: Schedule of major activities

2.4 Project team The key project team members were:

Role Name Affiliation Role Name Affiliation

Program Director Ashley Hunt PBSP Director Network Protection David Wilkinson United Energy

Project Manager Peter Dobson PBSP project manager Test Rig Operator Adrian Graves HRL Technology

Lead Researcher Tony Marxsen Marxsen Consulting Data Manager Marc Listmangof HRL Technology

Test Manager Andrew Czerwinski HRL Technology Network Tester Andrew Wilson United Energy (Tenix)

Site Controller Trevor Dixon United Energy (Tenix) Camera Operator David Adermann MACS Images

May June July August September October November December January February March April May June JulyHost network agreement in principle Lead Researcher appointedTrial principles and concept approvedInternational survey of REFCL practice completedPrimary research partner appointedSpecialised equipment orderedFormal risk analysis completed by network hostTheoretical modelling work orderedProof of concept tests completedTest site design completedTest rig design completedTest site commissionedTranche 1 testsTranche 2 testsTranche 3 testsTrial report drafted

2013 2014

P a g e | 14


3 ‘Wire on ground’ fire risk: comparative NER/ASC/GFN analysis The test results and the specific findings derived from them, as detailed in the following sections of this report, confirm that REFCLs can reduce fire risk from powerline ‘wire on ground’ faults. The results have been used to support the following comparative analysis of overall fire risk from this class of fault between current network protection practices based on solid earthing or a Neutral Earthing Resistor (NER) and two REFCL variants: Arc Suppression Coil (ASC) and Ground Fault Neutraliser (GFN).

This comparative analysis illustrates that:

• Either REFCL variant can reduce fire risk compared to non-REFCL network protection • GFNs can reduce fire risk more than ASCs • A GFN with four specific design enhancements could have the potential to almost completely

eliminate fire risk from this type of fault.

This analysis relies on:

• Categorisation of faults and ignition types; and • A key assumption to enable particular test results to be specifically linked with worst case

powerline faults in real networks.

These dependencies must be borne in mind when considering the results of the comparative analysis. They have been clearly described in the following sections so individual network owners can if desired, substitute their own insights, experience and standards to derive results that are more meaningful for their particular circumstances.

3.1 Categories of ignition and faults As the test program progressed, insights emerged that prompted the categorisation of different ignition mechanisms and reinforced the need for categorisation of different severity levels of earth faults. To aid clarity and support analysis, the following categories were defined and adopted:

3.1.1 Ignition categories

Based on observations in tests, ignition causes in ‘wire on ground’ earth faults were categorised as follows:

1. Bounce ignition: ignition that results from electric arcs between conductor and soil associated with the first bounce of the conductor. Except in NER tests, these arcs generally lasted less than 100 milliseconds after initial electrical contact of the conductor with the soil bed.

2. Ground ignition: ignition that results from continuing electric arcs between the soil bed and a stationary conductor after all bouncing has stopped. Typically this period commenced about 500 milliseconds after the initial conductor impact and extended to the end of the defined fault duration set in the test procedure adopted for that test.

The factors that influenced ignition probability in these two ignition categories were different as detailed later in this report.

In the case of higher current tests with non-REFCL protection, the distinction between these two ignition categories was not always valid as arcs commenced from the first contact and continued without interruption through the bounce period and into the ground ignition period. However, this does not

P a g e | 15


materially diminish the value of the categorisation – the two types of ignition could be clearly distinguished in lower current tests with NER-based protection and in all tests with REFCLs in service.

A third category of ignition – fulgurite ignition – was observed in some ignition tests (see Appendix D). This ignition cause is a feature of faults where the soil current ‘punches through’ to buried metal (pipes, cables, foundations, etc.) and forms a high power arc in a hollow tube through the soil, ejecting incandescent molten material that, together with the arc, ignites surrounding vegetation. Fulgurite ignition was not the focus of the test program which was concerned with ‘wire on ground’ faults in rural settings away from buildings and buried metal. It was also observed that this type of ignition mainly occurred with NER-based network protection as REFCLs appear to inhibit fulgurite formation almost to the point of complete elimination.

3.1.2 Earth fault categories

In accordance with general industry practice, network earth faults have been categorised by fault current. To avoid ambiguity, specific category boundaries have been allocated to the usual less formal industry taxonomy. The boundaries used here are based on earth fault protection settings used by the host network owner, United Energy. Other network owners are invited to apply the method outlined in this section by substituting their own values:

1. Heavy earth faults: ‘wire on ground’ faults that draw fault currents greater than normally caused by conductor-soil contact in hot dry weather, i.e. those that involve earthed metal (either by fulgurite formation or by direct conductor contact), contact with low voltage wiring, or water on the ground. Anecdotal evidence indicates that typically such faults draw more than 200 amps. These faults were not the focus of this test program – they are not as common in hot dry conditions, so they are not usually the cause of major fires. Nevertheless, some of the higher current test results can be applied directly to them. Typically, heavy earth faults are detected and managed by earth fault protection systems that either act instantaneously or have a fast speed of operation inversely proportional to fault current.

2. Low impedance faults: faults that traditional (non-REFCL) inverse time delayed protection systems can detect, i.e. faults with current flow greater than about 50 amps but less than 200 amps. It is understood that some network owners set inverse time delayed earth fault protection at much higher minimum operating currents than others, e.g. up to 200 amps. In this analysis the 50 amps value used by the host network owner United Energy was adopted.

3. High impedance faults: faults that traditional (non-REFCL) sensitive earth fault (SEF) protection can detect, i.e. faults with sustained current flow greater than five amps but less than 50 amps. Again, it is understood there is a diversity of practice across Victoria’s network owners with minimum operating currents for sensitive earth fault protection set at five, seven, nine or eleven amps. Sometimes this setting reflects the setting of the master earth fault protection in the substation as this ‘gates’ the operation of the sensitive earth fault protection on individual feeders. SEF settings are also strongly influenced by the number and type of ‘downstream’ protection systems on a feeder, such as pole-mounted Automatic Circuit Reclosers (ACRs). As in the case of low impedance faults, the five amp value used by the host network owner, United Energy, was adopted for this analysis.

4. Very high impedance faults: faults that cannot be detected by traditional (non-REFCL) sensitive earth fault protection but which a high sensitivity REFCL can detect, i.e. faults with current flow greater than one amp but less than five amps. The usual fault detection sensitivity setting for REFCLs is two amps, but experience in the test program indicated this might be reduced to one amp temporarily in high fire risk periods with acceptable risk of ‘false positive’ detections. The

P a g e | 16


feasibility of increased REFCL fault detection sensitivity is an area for medium term investigation beyond this test program.

5. Undetectable faults: faults that cannot be detected even by a high-sensitivity REFCL, i.e. faults with current flow less than one amp. These faults were not the focus of this test program as they are not usually the cause of major fires provided they are ‘wire on ground’ faults, but some of the low current test results could apply to them. Whilst not necessarily creating a fire risk, such faults are a clear public safety risk as they involve a live conductor lying on the ground for an extended period.

The above categories of real ‘wire on ground’ faults used in the comparative analysis of NER/ASC/GFN fire performance are summarised in Table 1: Table 1: categories of real 'wire on ground' network faults

Real ‘wire on ground’ fault

Fault category Fault current

Heavy >200A

Low impedance 50-200A

High impedance 5-50A

Very high impedance 1-5A

Undetectable <1A

3.2 Relating test results to real network earth faults The total current seen by the REFCL or NER in real network faults would normally be greater than the soil currents measured in the tests reported here, primarily because there is more conductor in contact with the earth in a real ‘wire on ground’ fault.

The ignition tests measured soil current drawn by a 400 millimetre length of conductor striking and coming to rest on a relatively homogeneous soil bed supporting standing fuel. This contrasts with real faults, where the network protection system sees the total fault current drawn by perhaps hundreds of metres of conductor falling onto ground which may not be flat or even. Further, contact may not be homogeneous but may include a diverse mix of different soil types, rock and vegetation.

To apply the test results presented in this report to real network situations, a relationship between fault current and test soil current must be postulated so that for any real network ‘wire on ground’ fault, the most applicable test result can be identified.

3.2.1 Fault current and soil current

In relating ignition test results to real network faults, three different currents must be separately considered:

P a g e | 17


1. Test fault current1: the current seen at the source substation by the REFCL or NER during a test – this current determined the response of the Frankston South network protection to the test.

2. Test soil current: the current into the 400 millimetre (0.4 metres) long soil bed in an ignition test – this was the current actually measured on-site in the tests.

3. Fault current: the current that flows in a real network ‘wire on ground’ earth fault – this is the current used to categorise faults into heavy, low impedance, high impedance and very high impedance, etc.

The relationships between these three currents must be clearly understood if the test results are to be validly applied to real network earth faults to assess the fire risk benefits of various protection schemes.

3.2.2 The key enabling assumption: current proportional to conductor length

The key assumption adopted for this comparative analysis was:

Current into the soil is distributed evenly along the length of fallen conductor, i.e. fault current is proportional to the length of conductor on the ground.

This assumption allows the ratio of fault current to test soil current to be assumed equal to the ratio of the length of fallen conductor on the ground to the length of conductor in the test rig (400 millimetres).

The rationale for this assumption relies on the fact that compared to soil resistance, conductor resistance is small, i.e. in most ‘wire on ground’ network faults, all parts of the conductor will be at almost the same voltage with respect to the soil. Therefore the current into the soil from any part of the conductor should be the same as from any other part. Variations in local soil current levels should be due only to variation in soil resistivity, uneven terrain, rocks, stumps, bushes, etc.

The primary supporting evidence for the assumption is high speed video records showing multiple similar arcs at relatively uniform intervals along the length of the conductor, i.e. arc current appears to be spread uniformly along the length of the conductor. A typical example is shown in Figure 3:

Figure 3: Test 235 - distribution of conductor-soil arcs at instant of initial contact (16 amps, 19/3.25AAC conductor)

1 The fault current produced by the tests was the sum of the soil current and (in some tests) current in a second parallel path provided at the test site (comprising either a shunt resistor or a coil of conductor resting on an earthed ‘sandpit’) to simulate a longer length of conductor on the ground. This second path current was not measured directly at the test site except in calibration tests. With a 600 ohm shunt resistor, it could reach 21 amps, but in most ignition tests it was ‘sandpit’ current of 2.5 amps or less. An indicative initial value of the fault current drawn by tests could sometimes be deduced from records downloaded from the GFN. Of course, in tests where there was no second parallel path the fault current generated by the test was simply the soil current.

P a g e | 18


The rationale and available evidence both suggest this assumption is plausible, but it must be recognised that it has not yet been rigorously proven by objective measurements in suitable large scale ‘wire on ground’ tests. Network owners are invited to substitute other assumptions based on local knowledge to develop an understanding of their network fire risk landscape that better suits their local circumstances.

3.2.3 Real network ‘wire on ground’ faults on Code Red days

The ‘fault current proportional to conductor length on ground’ assumption can be used to relate test results to real network faults. To do this, it is necessary to define a worst case ‘wire on ground’ earth fault. Worst case in this consideration is that which would produce an interpretation of the test results which provides the most conservative assessment of fire risk. Generally, this means a fault that involves the minimum length of conductor contacting the ground.

Despite their certain occurrence, there is limited detailed information available on real network ‘wire on ground’ earth faults in high fire risk conditions. Information that is reasonably reliably known includes:

• Typical rural powerline span length is 80-100 metres. Span lengths in peri-urban areas tend to be shorter (down to 40 metres) and those in remote areas can be considerably longer (up to one kilometre).

• A conductor falling from a height of eight metres (the estimated minimum mid-span height during high demand periods on very hot days) will strike the ground at a speed of about 12 metres per second

This information implies that it is reasonable to interpret test results slightly differently for the two different types of ignition, bounce ignition and ground ignition. The rationale for this is as follows:

3.2.4 Bounce ignition – worst case real network fault

The most conservative estimate of fire risk in real network faults results from assuming that bounce ignitions result from arc processes that take place in the first 50 milliseconds of contact. A briefer period than this appears unrealistic from examination of high speed video records of tests showing the conductor partially embedded in the soil for about 30 milliseconds. A longer period will produce a more optimistic, i.e. less conservative, estimate of fire risk.

Figure 4 illustrates the possible impact of a falling conductor in a real network ‘wire on ground’ fault that is worst case for bounce ignition. At a speed of 12 metres per second, a broken conductor falls 600 millimetres in 50 milliseconds, so the length of conductor that impacts the ground in the first 50 milliseconds of a ‘wire on ground’ earth fault is the length that is within 600 millimetres of the ground when the first contact occurs.

High speed video records of line drop tests indicated that this length could vary widely. However, for conductors which fell relatively flat, i.e. without curling due to ‘drum memory’ and with multiple counter-twisted layers of strands that inhibit ‘unwinding’ when tension is released, perhaps four to ten metres of conductor could impact the earth in the first 50 milliseconds. The more conservative case would be to assume that four metres of conductor could be involved in the risk of a bounce ignition.

Assuming current into the soil is proportional to conductor length, this means that the most relevant test results for the assessment of bounce ignition risk will be those in which soil current is about ten per cent of the fault current in the corresponding real network fault, since the length of conductor in the test rig (400 millimetres or 0.4 metres) is ten per cent of the four metres contacting the ground in the real ‘worst case’ fault.

P a g e | 19


Figure 4: bounce ignition conditions in a worst case real network fault

Network owners with information or accumulated experience of ‘wire on ground’ earth faults are invited to substitute their own information to allow risk assessment to better reflect their own circumstances. Varying the assumptions in this ‘worst case’ scenario for bounce ignition, e.g. by assuming a longer time than 50 milliseconds or a longer length of conductor within 600 millimetres of the surface at the instant of first contact, will increase the ratio of fault current in the real fault to soil current in the corresponding test.

If the fault current were to be twenty times the test soil current instead of ten times, the only types of real faults that would involve bounce ignition risk would be heavy ones, i.e. possibly unrealistic in high fire risk conditions. It can be seen that a variation of this nature would reduce the estimate of fire risk from powerline faults and in effect, move the assessment away from potential ‘worst case’ conditions.

In this comparative analysis the worst case is taken to be four metres of conductor involved in bounce ignitions, i.e. for assessment of bounce ignition risk, the ratio of fault current to test soil current is ten to one.

3.2.5 Ground ignition – worst case real network fault

The considerations in worst case ground ignition risk are somewhat simpler, as shown in Figure 5. Given the average rural span is 80 metres and individual spans may be much longer and that most conductor breaks occur close to a pole, the length of live conductor that comes to rest on the ground when a powerline span falls has been assumed to be not less than 40 metres.

Network owners with information or accumulated experience of ‘wire on ground’ earth faults are invited to substitute their own information to allow risk assessment to better reflect their own circumstances. Varying the assumptions in this ‘worst case’ scenario for ground ignition, e.g. by assuming a longer length of conductor lying on the surface, will increase the ratio of fault current in the real fault to soil current in the corresponding test.

P a g e | 20


Figure 5: worst case 'wire on ground' fault for ground ignition

Comparing the assumed 40 metres to the 400 millimetres (0.4 metre) of conductor in the test rig, the most relevant test results for the assessment of ground ignition risk will be those in which soil current is about one per cent of fault current.

If the fallen conductor length were assumed to be 200 metres, fault current would be five hundred times the test soil current instead of one hundred times. Then fallen conductors that draw less than 30 amps would not start fires, even in worst case fire risk conditions. This would reduce the estimated fire risk from powerline faults and in effect, move the assessment away from potential ‘worst case’ conditions.

In this comparative analysis, the worst case is taken to be forty metres of conductor involved in ground ignitions, i.e. for assessment of ground ignition risk, the ratio of fault current to test soil current is one hundred to one.

3.2.6 Relating test results to real network faults

Based on the relationships shown in Figure 4 and Figure 5 between fault current in the assumed worst case real ‘wire on ground’ network faults (four metres and forty metres) and the test soil current, the test results most relevant to each category of fault are listed in Table 2:

Line resting on ground: 40 metres

Test bed:0.4 metres

Ground ignition: fault current is one hundred times test soil current

Ground ignition: fire (flame) that starts after the conductor has come to rest on the ground

P a g e | 21


Table 2: Guide for interpretation of test results

Real ‘wire on ground’ fault Test soil current for real worst case fault

Fault category Fault current Bounce ignition Ground ignition

Heavy >200A >20.0A >2.0A

Low impedance 50-200A 5.0-20.0A 0.5-2.0A

High impedance 5-50A n/a 0.05-0.5A

Very high impedance 1-5A n/a 0.01-0.05A

Undetectable <1A n/a n/a

The ‘n/a’ entries in Table 2 are for conditions which the test program identified as unlikely to produce ignition, e.g. bounce ignition was not observed for first bounce peak soil currents below ten amps. Ground ignitions were not observed at soil current levels below 0.06 amps.

3.3 Network fire risk landscape – inherent fire risk When ignition test results are overlayed onto Table 2, a ‘risk landscape’ can be produced showing the types of faults that are likely to produce a vegetation fire from a ‘wire on ground’ earth fault under worst case fire risk conditions. This landscape ignores the effects of network protection systems and should be viewed as raw or inherent risk, i.e. before application of any risk controls. The inherent fire risk landscape is shown in Figure 6:

Figure 6: inherent fire risk from 'wire on ground' earth faults

The green areas on the risk landscape are those where test results indicate a fire is unlikely, i.e. near-zero chance of a fire start. For bounce ignitions this was with less than ten amps peak soil current spread over 400 millimetres of conductor/ground impact; for ground ignitions it was at sustained soil current less than 0.06 amps over 400 millimetres of conductor/soil contact.

The red areas are those where test results indicate fires are likely, i.e. probability of a fire is significant, not near-zero. If the assumptions for worst case fallen conductor lengths are made more conservative, e.g. less than four metres contact in the first 50 milliseconds of conductor bounce and less than 40 metres lying on the ground at rest, the red areas would extend downwards to cover more of the

P a g e | 22


network fire risk landscape. Hence, the decision as to what is a realistic ‘worst case’ fault establishes the boundaries of the fire risk territory on the network fire risk landscape.

3.4 Fire risk equation The inherent risk landscape shown in Figure 6 can be used to illustrate the benefits of different network protection schemes which act to control the risk. The effectiveness of controls that might potentially prevent fires generally relies on the relativity of three specific time intervals set out in the simple powerline fire-risk equation shown in Figure 7.

Figure 7: The powerline fire risk equation

For a particular ‘wire on ground’ fault, two of the terms in the fire risk equation can be determined from data held by network owners:

3.4.1 Time to detect

The ‘time to detect’ is the time it takes a network protection system to detect that a network fault has occurred and to confirm it is a real fault and that it is the best protection system to respond. These three elements are set by network owners to suit local circumstances:

1. Detection of the fault: The time taken for a modern relay to detect that a current has exceeded a set threshold is very short (typically of the order of 20 milliseconds or less) and is usually a minor part of the total ‘time to detect’.

2. Confirmation the fault is real: All networks experience a constant low level of transient activity due to switching of loads by customers, switching of powerline segments by network operators, etc. Many protection systems, especially the more sensitive ones, include inbuilt time delays to ensure these transient events do not cause false detections, i.e. the fault has to remain present for some time before the protection system recognises it as a genuine fault upon which it must act.

3. Confirmation this protection system is the best one to act: Different time delays are also used to ensure different protection systems are ‘graded’ along the length of the feeder, i.e. to ensure the one closest to the fault acts quickest, so no more than the minimum necessary number of customers loses supply in the incident.

Earth fault detection is normally done by integrated feeder protection packages with multiple protection elements. Some of the more common functions used to detect earth faults on Victoria’s rural networks are:

P a g e | 23


• Definite time sensitive earth fault protection (SEF): This function detects any earth current above a set threshold and flags a fault if the current exists for longer than a set duration. In sensitive earth fault systems, ‘time to detect’ is usually set in the range 500 milliseconds to 3.0 seconds for normal days and may be reduced to 500 to 800 milliseconds for Total Fire Ban and Code Red days.

• Inverse time delayed earth fault protection (ITDEF): In this function, the minimum detection current is higher and the ‘time to detect’ is generally much shorter and often in the range 100 to 500 milliseconds. The delay time also varies inversely with the fault current magnitude. However (refer Finding 2) ITDEF protection may not be very relevant to fire risk reduction from ‘wire on ground’ faults.

• Instantaneous earth fault protection (IEF): This function detects high current earth faults, typically with fault currents above 400 amps. It is not relevant to fire risk reduction as ‘wire on ground’ earth faults at this current level will always start a fire if vegetation is present.

Modern feeder protection packages offer improvements that potentially allow time settings to be shortened. For example, they filter transients out of the current they monitor so they are less likely to require ‘fault confirmation’ time. They incorporate highly accurate and reliable digital timers, so ‘grading’ margins can be reduced. It is not known to what extent these improvements have already been exploited by network owners to speed up operation of earth fault protection systems on high fire risk days.

3.4.2 Time to act

The time it takes for a protection system to disconnect the fallen conductor once it has ‘decided’ to do so is essentially the operating time of the high voltage switch used to interrupt supply.

For older feeder circuit breakers, this is typically about 80 to 120 milliseconds. For newer switches such as vacuum switches in modern Automatic Circuit Reclosers (ACRs) it can be much faster – less than 50 milliseconds.

From the above, it can be seen that the (time to detect) + (time to act) part of the fire risk equation can range from less than 100 milliseconds to more than 3.0 seconds. If ‘time to ignite’ exceeds this range then at least some fires can be prevented by existing non-REFCL technology.

3.4.3 Time to ignite - new data from test results

Prior to this test program, there was little if any data available on ‘time to ignite’ for a fallen live conductor in worst case fire risk conditions.

A wide variety of ground ignition times was recorded in the test program. Though the timing was subject to wide random variations, the instant of ignition in a test was usually well defined – video records show increasing production of smoke until the concentration of flammable gas being produced reaches the ignition threshold, whereupon they ignite in a quick flare before a steadier flame starts to grow. ‘Time to ignite’ was taken to be the duration of arcing from the time the conductor settled on the soil bed up to the instant of the initial flare. A typical ignition event is shown in Figure 8.

P a g e | 24


Figure 8: Test 242 - initial flare in ground ignition with 3/12 steel conductor

Immediately prior to flare Initial flare of accumulated gas Flame 0.5 seconds after initial flare

Consideration of all tests that resulted in ground ignitions yielded the ‘time to ignite’ results shown in Figure 9.

Figure 9: 'time to ignite' curve from ignition tests

Figure 9 shows that ground ignition is unlikely at currents below 0.06 amps (spread over 400 millimetres of conductor/soil contact) and though ignition was observed to occur as quickly as 100 milliseconds at quite low currents (0.3 amps over 400 millimetres of conductor/soil contact), it often took much longer – often up to five or ten seconds. The longest recorded time to ignite was more than 50 seconds.

The ‘time to ignite’ test results indicate that traditional non-REFCL protection systems could have the capability to act in time to prevent at least some fires. The one caveat that must be borne in mind is that

P a g e | 25


along the length of any fallen conductor there will be a diverse variety of micro-situations, some of which will be conducive to very quick ignition while others may need longer times. This means the first ignition is more likely to occur at the shortest time shown in the test results for that value of soil current, not the average time. For example, at soil currents close to 0.3 amps, tests showed ‘time to ignite’ ranging from 100 milliseconds to 54 seconds. The reality would most probably be ignition somewhere along the length of fallen conductor in about 100 milliseconds.

The ‘time to ignite’ test results can be combined with the powerline fire risk equation and the relationship between assumed worst case faults and specific tests to derive a network fire risk landscape for residual risk, i.e. network fire risk after application of controls.

3.5 Network fire risk– res idual risk with existing (non-RE FCL) protection With the exception of Frankston South zone substation which is fitted with a GFN, all networks serving rural areas of Victoria are either solidly earthed (about 30 to 50 per cent of zone substations) or earthed via an NER. This comparative analysis uses NER-based protection as the baseline because it was available for tests at Frankston South with the REFCL out of service and also because it produces a more conservative estimate of the fire risk benefits of REFCLs. Fire risk will generally be higher in solidly earthed networks than with NER-based protection because earth faults in such networks can draw more current.

Apart from the single REFCL at Frankston South, all earth fault protection in Victoria is based on detection of earth fault current using three schemes: instantaneous operation for high earth fault currents, time delayed operation for medium earth fault currents and definite-time sensitive earth fault protection (SEF) for low fault currents. SEF is sometimes ‘gated’ by master earth fault protection to confirm that there is not only earth fault current on a particular feeder but there is also earth fault current into the transformer neutral connections. This reduces the risk of tripping on false positive detections.

Consideration of the powerline fire risk equation shows that SEF offers some reduction in fire risk from ‘wire on ground’ earth faults as illustrated in Figure 10. Analysis outlined in Finding 2 indicates the instantaneous and inverse time-delayed earth fault protection schemes are likely to offer little, if any, fire risk reduction in worst case conditions.

Figure 10: network residual fire risk landscape - non-REFCL network protection

It can be seen that the main fire risk benefits from non-REFCL protection is through reduction of ground ignitions. Given the spread of ‘time to ignite’ results in tests, it is a probabilistic reduction, i.e. it reduces the probability of fire starts from high impedance faults but it does not entirely eliminate them.

P a g e | 26


3.6 Network fire risk– res idual risk with ASC protection (REFCL variant) An Arc Suppression Coil (ASC) is usually combined with modern digital earth fault detection relays to trip the faulted feeder when it is identified. The analysis and testing of such relays was beyond the scope of this test program and for the purposes of this comparative analysis their performance is assumed to be of the same order as the digital fault detection, fault confirmation and faulted feeder identification algorithms in the GFN product at Frankston South.

The primary fire risk reduction benefits of the ASC variant of REFCL technology derive from two features of its operation:

1. Fast collapse of the voltage on the faulted phase - even during the first bounce if the fault current is high enough

2. Reduction of residual current in the ground ignition phase leading to reduced probability of ignition in the period before an earth fault detection relay can trip supply to the faulted feeder.

These benefits appear on the fire risk landscape for the ASC REFCL variant shown in Figure 11.

Figure 11: network residual fire risk landscape - ASC variant of REFCL technology

It must be noted that this assessment is subject to two significant sources of uncertainty:

3.6.1 ASC response depends on fault current

Unlike the NER-based and GFN-based protection technologies, an ASC relies on the fault current itself to collapse the voltage on the faulted phase. In effect, it inserts a passive high impedance element into the fault current path to reduce the fault current. For a test to be a valid indication of ASC fire performance, the test fault current must equal the fault current in the real network fault. This was not practically feasible in this test program, so ASC ignition test results must be interpreted with care.

Neither of the arrangements which increased the test fault current above the test soil current provided a way around this limitation:

• Ignition tests with 12.5 metres of conductor lying on a ‘sandpit’ added a 5,000 ohm shunt resistance across the test rig. This was enough to reduce the voltage on the faulted phase by about 45 per cent in ASC tests. However whilst soil current varied widely due to variations in soil bed moisture content, ‘sandpit’ resistance was constant, i.e. it did not validly simulate the effect of additional conductor length lying on the same soil.

• Ignition tests with 600 ohm and 1600 ohm shunt resistors added 21 amps or eight amps respectively to test fault current. However, the linear resistors used did not replicate the non-

P a g e | 27


linear conductor-soil resistance characteristics of the soil bed and residual soil current was reduced disproportionally, sometimes to zero.

This issue was not material in NER or GFN tests: in NER tests, additional test fault current did not materially reduce conductor voltage; in GFN tests, the 2.5 amp ‘sandpit’ current was enough to ensure the GFN detected the fault and engaged the residual current compensator, whereupon both ‘sandpit’ current and the soil current virtually disappeared, which is what would happen in a real fault. However, the issue affected GFN tests during any period when the RCC was turned off since in that condition the GFN acts like an ASC.

This dependence of conductor voltage on fault current in ASC tests essentially invalidated the assumptions applicable to NER and GFN ignition tests to relate particular test results to real worst case network faults.

3.6.2 Typical ASC protection systems were not tested

The ‘raw’ performance of an ASC was tested by suppressing the operation of the residual current compensation on the Frankston South GFN, i.e. the additional coil tuning and protection technology normally included in an ASC installation was not available to be tested. This technology is similar in function to digital algorithms embedded in the GFN product, but employs different products from a range of manufacturers. Any differential fire risk benefit of ASC ancillary systems compared to the same functions in the GFN was not able to be assessed.

In summary, the ASC ignition test results are potentially useful but require a more complex analysis before the results of any particular test can be reliably related to a real fault on an ASC-protected network.

Despite the above limitations, two findings related to ASC fire risk were clear to test observers:

1. An ASC causes conductor voltage to collapse during the first bounce. Hence the ASC will reduce the probability of bounce ignition.

2. During the ground ignition phase, an ASC will reduce residual voltage and current compared to NER-based protection but not to levels as low as a GFN would achieve. The extent of the fire risk reduction from the lower residual current remains uncertain to some extent due to the above-described limitations of the test program, but some reduction in ground ignition risk will result when compared to the fire performance of NER-based protection.

On that basis, the residual fire risk landscape in Figure 11 is considered qualitatively valid for ASCs in the context of comparative analysis against NER and GFN fire performance levels.

3.7 Network fire risk– residual risk with GFN protection (REFCL variant) The GFN adds residual current compensation (provided by the RCC) to an ASC together with a range of digital fault detection algorithms similar to those found in other modern digital earth fault detection relays. Its fire risk performance depends on all three elements: ASC, RCC, and algorithms.

Today’s GFN product offers some additional fire risk benefits over an ASC alternative, as illustrated in Figure 12. These include a reduced probability of bounce ignitions at high fault current levels and a reduced incidence of ground ignitions due to lower residual current. However, test results indicated it does not yet fulfil its full potential - some remaining fire risks are directly traceable to design features that could be improved. Comments by the manufacturer indicate it is already contemplating ways and means to do this should the market demand it.

P a g e | 28


Figure 12: residual fire risk landscape - today's GFN variant of REFCL technology

The GFN ignition tests generated valuable insights into the challenges of improving GFN fire risk performance.

3.7.1 Ground ignitions with a GFN

Successful management of an earth fault incident on a GFN-protected network in high fire risk conditions requires the GFN to perform the following sequence of actions:

1. Detect the fault 2. Ascertain which phase of the faulted network is carrying the fault current to earth 3. Compensate residual current by injecting a voltage into the network neutral connection to move

the voltage of the faulted phase close to zero 4. After a few seconds, test to see if the fault is still present (the fault-confirmation test) 5. If the fault is still present, identify the feeder on which it is located 6. Trip the faulted feeder to remove the fault from the network 7. Switch off the residual current compensation so network voltages return to normal levels.

The test program confirmed the first three of these actions:

• Fault current of just one amp was detected by the GFN, albeit with sensitivity heightened above normal levels by reducing the neutral voltage detection threshold from 30 per cent to 20 per cent. With normal sensitivity settings, the GFN detects earth faults of two amps magnitude.

• Residual current compensation reduced the soil current to zero (the voltage on the conductor dropped to 200-400 volts - insufficient to drive any current across the conductor/soil interface).

The next two tasks in the sequence (to confirm whether the fault is still present and if so, reliably identify the feeder on which it is located) pose fire risk challenges as they require some soil current to flow for the GFN to make the required measurements. The final two tasks (trip the faulted feeder, switch off the residual compensation to return the network to normal operating condition ready to deal with another fault) are relatively straightforward.

The tests confirmed the extent of the challenge of managing fire risk during Steps 4 and 5 of the earth fault response set out above. An earlier version of the GFN fault-confirmation test switched off the residual current compensation for one second to generate enough fault current flow to allow accurate measurement of network conditions. Tests confirmed this is likely to produce ground ignitions in many circumstances. A newer version of the GFN fault-confirmation test which involves controlled adjustment of the residual current compensation was also tested. In the very limited number of tests performed

P a g e | 29


with this feature enabled, it proved challenging to find a combination of settings that would reliably detect the fault and identify the faulted feeder without starting a fire.

Despite mixed test results, the degree of fine control of residual conductor voltage demonstrated in tests of today’s GFN product encouraged expectations among test observers that suitable settings will eventually be found to achieve the goal of ignition-free fault confirmation and faulted feeder identification.

In summary, the GFN offers virtual elimination of residual fault current into the soil but then faces a challenge during confirmation of the continued presence of the fault and identification of the faulted feeder, both of which require some current to flow. The challenge is to perform these functions at a level of current flow too low to produce a ground ignition event or to do it faster than ‘time to ignite’ at the particular current level used.

3.7.2 Bounce ignitions with a GFN

The GFN also reduces the probability of bounce fires. Bounce ignitions recorded in ignition tests using the five-year old GFN at Frankston South all involved delayed RCC response which has since been remedied in today’s GFN product. Allowance has been made for this in Figure 12.

The program goal of a direct ASC-GFN comparison proved elusive and a number of uncertainties remain, including those outlined in the section above on the ASC residual fire risk landscape. However, the conclusion reflected in Figure 11 and Figure 12 (see Finding 7) is considered to be valid. Even at its most aggressive, the GFN fault-confirmation test applies for just one second the same residual fault current that an ASC applies for the whole fault period. It is natural to conclude the GFN’s shorter duration of a similar level of residual current will result in lower fire risk.

3.8 Network fire risk– res idual risk with improved GFN (REFCL variant) If the specific GFN design improvements described below were in place, the network residual fire risk landscape would be as shown in Figure 13.

Figure 13: residual fire risk landscape - improved GFN

This would indicate that a suitably improved GFN may have the potential to virtually eliminate fires from ‘wire on ground’ faults in worst case fire risk conditions.

The GFN product has been developed in Europe to address public safety and supply reliability in overhead networks and the issue of restriking faults in underground cable networks in major cities. Victoria is the first potential REFCL market in the world that has placed fire risk reduction at the top of the design priorities list. The specific GFN design improvement opportunities revealed by the test program should not be seen as deficiencies in the current design – it is simply that the market has never before demanded them.

P a g e | 30


These opportunities have been briefly discussed with the manufacturer and while significant development effort might be required, these very preliminary discussions revealed no major barriers to delivery of the desired fire risk performance. As usual it can be expected that product development will only occur in response to market demand for it. The design improvements that might offer highest fire risk reduction benefits in Victoria’s context are:

3.8.1 Increased sensitivity with improved tolerance for network imbalance

If the worst case ‘wire on ground’ earth fault is not 40 metres of conductor lying on the ground, but four metres, then undetected faults could start fires. Fault currents of less than one amp are undetectable even with today’s GFN product. Increased GFN fault detection sensitivity would reduce fire risk in faults that involve shorter lengths of conductor on the ground. The same benefit would be delivered for ‘wire into vegetation’ faults that also typically draw low fault currents.

Discussions with the GFN manufacturer indicate that changing to an adaptive fault detection algorithm that detects faults by sudden changes in the neutral displacement voltage may offer potential for increased sensitivity. This change might also offer increased tolerance for standing network imbalance, which is an issue for remote rural networks with very long two-wire spur lines.

3.8.2 Faster onset of residual compensation

The test results indicate ASC/GFN bounce ignitions are marginal, i.e. a small increase in speed of RCC response might entirely eliminate bounce fire risk in GFN-protected networks. The current GFN product activates and connects the RCC 60 milliseconds after the fault. Once the RCC is activated, it starts to have effect immediately and reaches full effect in a further 25 milliseconds to reduce the voltage on the fallen wire to a level close to zero. This period can be shortened by using a higher rated inverter to enable the RCC to inject a higher current into the ASC.

It was noted that the GFN employs an electro-mechanical contactor (an inherently slow device) to connect the RCC when response to a fault is required. Modern power electronics options would seem to offer an opportunity for faster RCC response. Test results indicate that shaving just 40 milliseconds off the RCC activation time might entirely prevent bounce ignition.

3.8.3 Fast reliable fault-confirmation and faulted-feeder identification

Tests confirmed the GFN’s new ‘gentle’ fault-confirmation test could avoid ground ignition. However as currently configured, it occasionally experienced difficulty detecting the continuing presence of the fault. The level of discrimination in its identification of the faulted feeder was also less than required for high-confidence feeder tripping to allow a return to normal network operation. If either of these tasks failed in a test, the GFN switched off the RCC and the consequent increase in residual current sometimes led to a ground ignition event.

Part of the challenge is the highly non-linear characteristics of the conductor-soil contact which obstructs the current flow required for GFN measurement processes. Test observers tended to adopt a pragmatic standard along the lines of ‘if I can tell from inspection of the voltage and current records that the fault is still there and which feeder it is on, then it is a realistic expectation that the GFN algorithm can do the same’. This criterion would indicate that reliable fault-confirmation and feeder-identification should be available with further GFN algorithm development.

P a g e | 31


3.8.4 Automated RCC tuning

For minimum residual fault current, the RCC must be calibrated to match current network voltage levels. At the start of the test program this requirement was not recognised and in early tests, compensation was incomplete with residual conductor voltages in some GFN tests exceeding 1,000 volts. When RCC calibration was performed, residual conductor voltage levels dropped below 250 volts. In today’s GFN product, the calibration procedure is partly automated. However, it still requires manual initiation and operator acceptance of the results before the RCC settings are changed.

The test program confirmed that following network changes, e.g. when a feeder is tripped, the RCC must be recalibrated if residual voltage levels are to remain as low as possible on succeeding faults. The GFN includes a fully automated ASC tuning process that automatically occurs whenever a network change is detected. A fully automated procedure to also automatically recalibrate the RCC after network changes would ensure residual fault currents in high fire risk conditions remain as low as possible in a dynamically changing network environment. This would minimise the risk of ground ignitions.

3.9 Conclus ion: comparative fire risk analysis The comparative analysis set out above leads to the following conclusions:

1. Even some of today’s non-REFCL network protection systems can deliver potentially valuable reductions in fire risk from powerline ‘wire on ground’ earth faults.

2. The different network protection schemes tested can deliver different levels of fire risk reduction. In order of highest to lowest fire risk, they are listed in Table 3 below.

Table 3: comparison of fire risk of different network protection schemes

Network protection scheme Fire risk from ‘wire on ground’ earth faults

Non-REFCL (traditional existing) Highest

REFCL: ASC

REFCL: GFN (today’s product)

REFCL: GFN (with improvements) Lowest

3. Whilst tests showed today’s GFN product delivers more fire risk benefit than a broadly equivalent ASC-based scheme, the test program did not succeed in quantifying the difference between these two variants of REFCL technology.

4. An improved GFN design appears capable of virtually eliminating fire risk from ‘wire on ground’ faults in worst case conditions.

A key assumption of uniform conductor-soil current distribution along the length of a fallen conductor was used in this comparative analysis. Further, selection of specific ‘worst case’ wire on ground’ faults was used to apply test results to real network faults and perform a comparative assessment of network fire risk. Network owners can use local knowledge to vary this assumption and this selection to suit their circumstances. However, while such variations might lead to adjustment of the network fire risk landscapes shown above, they would be unlikely to change the ranking shown in Table 3.

refcl trial: ignition tests - energy · 2017-03-10 · this report contains test results,...

Documents