Asset Management in Electricity Transmission

Enterprises: Factors that affect Asset Management

Policies and Practices of Electricity Transmission
Enterprises and their Impact on Performance

Jennifer J Crisp
BE Hons (UQ) M Eng (QUT) MIEAust MIEEE

Submitted for the degree

of
Doctor of Philosophy

School of Electrical and Electronic Engineering

Faculty of Built Environment and Engineering
Queensland University of Technology
Brisbane, Australia

February 2004
Key Words
asset management, infrastructure management, electricity transmission
systems, network performance, financial performance, fuzzy logic, rule-based
expert system, system dynamics modelling

Abstract
This thesis draws on techniques from Management Science and Artificial
Intelligence to explore organisational aspects of asset management in
electricity transmission enterprises. In this research, factors that influence
policies and practices of asset management within electricity transmission
enterprises have been identified, in order to examine their interaction and
how they impact the policies, practices and performance of transmission
businesses. It has been found that, while there is extensive literature on the
economics of transmission regulation and pricing, there is little published
research linking the engineering and financial aspects of transmission asset
management at a management policy level. To remedy this situation, this
investigation has drawn on a wide range of literature, together with expert
interviews and personal knowledge of the electricity industry, to construct a
conceptual model of asset management with broad applicability across
transmission enterprises in different parts of the world. A concise
representation of the model has been formulated using a Causal Loop
Diagram (CLD).

To investigate the interactions between factors of influence it is necessary to

implement the model and validate it against known outcomes. However,
because of the nature of the data (a mix of numeric and non-numeric data,
imprecise, incomplete and often approximate) and complexity and
imprecision in the definition of relationships between elements, this problem
is intractable to modelling by traditional engineering methodologies. The
solution has been to utilise techniques from other disciplines. Two
implementations have been explored: a multi-level fuzzy rule-based model
and a system dynamics model; they offer different but complementary
insights into transmission asset management. Each model shows potential
for use by transmission businesses for strategic-level decision support.

2
The research demonstrates the key impact of routine maintenance
effectiveness on the condition and performance of transmission system
assets. However, performance of the transmission network, is not only
related to equipment performance, but is a function of system design and
operational aspects, such as loading and load factor. Type and
supportiveness of regulation, together with the objectives and corporate
culture of the transmission organisation also play roles in promoting various
strategies for asset management. The cumulative effect of all these drivers
is to produce differences in asset management policies and practices,
discernable between individual companies and at a regional level, where
similar conditions have applied historically and today.

3
 Table of Contents
1 Introduction.................................................................................... 22
1.1 Questions prompting this research topic........................................ 22
1.2 Background to the research........................................................... 23
1.2.1 Transmission networks ................................................................23
1.2.2 Characteristics of transmission system plant ...............................24
1.2.3 Transmission enterprises .............................................................25
1.3 Defining asset management .......................................................... 25
1.4 Purpose of the research ................................................................ 28
1.5 Scope ............................................................................................ 28
1.6 Methodology .................................................................................. 28
1.7 Key Assumptions, Key sources .................................................... 30
1.8 Limitations ..................................................................................... 31
1.9 Structure of this thesis ................................................................... 31
2 Literature review ............................................................................ 34
2.1 The CIGRE Study .......................................................................... 35
2.1.1 Regional differences in maintenance practices............................36
2.1.2 Regional differences in maintenance spending............................38
2.1.3 Regional differences in replacement criteria ................................39
2.1.4 Regional differences in outage management...............................42
2.2 Strategic Level Asset Management ............................................... 43
2.2.1 Assessment of Regulatory impacts on transmission ....................44
2.2.1.1 Changes to structure of the electricity industry .......................... 44
2.2.1.2 Regulation within transmission................................................... 45
2.2.1.3 Research on regulation, structural reforms and governance...... 49
2.2.2 Assessment and management of market impacts .......................55
2.2.3 Interconnection of networks .........................................................57
2.2.4 Embedded generation..................................................................57
2.2.5 Risk management in transmission organisations .........................58
2.2.6 Level of performance ...................................................................61
2.2.7 Performance measurement..........................................................62
2.3 Process Level Asset Management ................................................ 62

4
2.3.1 RCM and its influence on asset management in transmission.....63
2.3.2 Condition based maintenance (CBM) ..........................................65
2.3.3 Justification and selection of condition monitoring .......................65
2.3.4 Maintenance period optimisation .................................................66
2.3.5 Strategies to increase circuit availability.......................................68
2.3.6 Criticality in transmission asset management ..............................69
2.3.7 Reliability modelling .....................................................................70
2.3.8 Life cycle costing..........................................................................71
2.3.9 Criteria for replacement of equipment ..........................................72
2.4 Equipment Level Asset Management ............................................ 74
2.5 Identifying the need for research ................................................... 74
3 Asset Management Relationships ................................................. 77
3.1 Factors of influence for asset management in transmission .......... 77
3.1.1 External / Business-related factors ..............................................77
3.1.2 Internal factors .............................................................................78
3.1.3 System-related factors .................................................................81
3.1.4 Environment-related factors .........................................................87
3.2 Important asset management relationships for transmission
enterprises ................................................................................................... 87
3.2.1 Factors affecting failure rate.........................................................88
3.2.1.1 The impact of climate and loading on failure rate....................... 88
3.2.1.2 The impact of equipment ageing on failure rate ......................... 92
3.2.1.3 The link between maintenance history and failure rates ............ 94
3.2.1.4 The effect of predictive maintenance on failure rates................. 95
3.2.1.5 Geographical factors affecting failure rate.................................. 95
3.2.1.6 The effect of voltage on failure rate............................................ 95
3.2.1.7 Other factors affecting failure rate .............................................. 96
3.2.1.8 Summary of factors affecting failure rate.................................... 97
3.2.2 Failure rate and corrective maintenance ......................................98
3.2.3 Use of time-base maintenance ....................................................98
3.2.4 Use of predictive maintenance.....................................................99
3.2.5 Use of online condition-based maintenance ................................99
3.2.6 Use of offline condition-based maintenance ..............................100
3.2.7 Factors affecting maintenance spending....................................100

5
3.2.8 Investment in new equipment ....................................................102
3.2.9 Drivers for equipment replacement ............................................102
3.2.10 Refurbishment of equipment ......................................................103
3.2.11 Drivers for system augmentation ...............................................104
3.2.12 Investment spending ..................................................................105
3.2.13 Ease of raising capital ................................................................107
3.2.14 Reliability of supply ....................................................................107
3.2.15 Availability ..................................................................................109
3.3 Summary of AM relationships...................................................... 111
3.3.1 Key elements in the model.........................................................111
3.3.2 Feedback loops in the model .....................................................112
3.3.3 Discussion on Chapter 3 and the CLD model ............................115
4 Modelling ..................................................................................... 117
4.1 Methodology ................................................................................ 117
4.2 Characteristics of the model ........................................................ 118
4.3 Desired outputs ........................................................................... 119
4.4 Potential modelling methods........................................................ 120
4.4.1 Fuzzy modelling .........................................................................120
4.4.2 System dynamics modelling.......................................................121
4.4.3 Markov analysis .........................................................................123
4.4.4 Time-based modelling of the decision-making process..............124
4.5 Selection of model methodology.................................................. 124
5 The multi-level fuzzy model ......................................................... 126
5.1 Development of the multi-level fuzzy model ................................ 126
5.1.1 Implementation of fuzzy modelling .............................................126
5.1.2 Model structure ..........................................................................130
5.1.3 Verification of the fuzzy model ...................................................134
5.1.4 Validation of the fuzzy model .....................................................136
5.1.5 Summary of data used in regional studies for the fuzzy model ..137
5.1.6 Refinement of the fuzzy model...................................................140
5.1.7 Summary of modelling methodology and implementation of the
fuzzy model..................................................................................................141
5.2 Results – Regional Comparisons using Fuzzy Modelling ............ 141
5.2.1 Failure rate.................................................................................142

6
5.2.2 Reliability of supply ....................................................................143
5.2.3 Availability of circuits ..................................................................147
5.2.4 Use of predictive maintenance...................................................148
5.2.5 Maintenance spending ...............................................................152
5.2.6 Refurbishment of equipment ......................................................154
5.2.7 Replacement of equipment ........................................................155
5.2.8 Discussion of results ..................................................................158
5.2.8.1 Efficacy of the methodology ..................................................... 158
5.2.8.2 Insights to be gained from these studies.................................. 160
5.2.8.3 Discussion of results of fuzzy modelling – regional cases........ 171
5.3 Case Studies for individual companies/countries ........................ 173
5.3.1 Australian case study – Powerlink Queensland .........................173
5.3.1.1 Description of the Powerlink Queensland case study .............. 173
5.3.1.2 Results of the Powerlink Queensland case study .................... 177
5.3.2 Malaysian case study.................................................................178
5.3.2.1 Description of the Tenaga Nasional Berhad case .................... 179
5.3.2.2 Results of the Tenaga Nasional Berhad Case Study ............... 185
5.3.3 Japanese case study .................................................................185
5.3.3.1 Description of the Japanese case study................................... 186
5.3.3.2 Results of the Japanese case study......................................... 189
5.3.4 US utility case study...................................................................191
5.3.4.1 Description of the Duke Energy case study ............................. 191
5.3.4.2 Results of the Duke Energy case study ................................... 194
5.3.5 UK utility case study..................................................................195
5.3.5.1 Description of the National Grid Company, UK case study ...... 195
5.3.5.2 Results from the NGC, UK case study ..................................... 200
5.4 Use of the model for strategic decision-making ........................... 201
5.5 Summary of results from fuzzy modelling .................................... 207
6 Investigations using system dynamics modelling ........................ 209
6.1 Simulation in system dynamics modelling ................................... 210
6.2 Model verification and validation in system dynamics modelling . 211
6.3 Repairable and non-repairable systems theory – application to this
model 216
6.4 Modelling population ageing and asset value .............................. 219

7
6.5 Condition deterioration, maintenance effectiveness and predictive
maintenance .............................................................................................. 221
6.6 Linking the age and condition models ......................................... 225
6.6.1 Impact of maintenance effectiveness on asset population and
value for a condition-based replacement/ refurbishment scenario ...............229
6.6.2 Impact of use of predictive maintenance on asset population and
value for a condition-based replacement/refurbishment scenario ................232
6.7 Age-based replacement and the impact on asset value and failure
rate 239
6.7.1 Impact of varying maintenance effectiveness fraction................240
6.7.2 Effect of (condition-based) refurbishment for age-based model242
6.7.3 Comparison of age-base and condition based replacement
models with refurbishment for different values of maintenance effectiveness244
6.8 Remedial maintenance of poor condition assets ......................... 246
6.9 Replacement on failure only ........................................................ 248
6.10 The impact of maintenance and replacement policy on debt and
profit 251
6.11 Comparison of profit outcomes for various replacement policies. 258
6.12 Sensitivity of profit to failure rate assumptions............................. 259
6.13 Comments on Maintenance efficiency and the impact of
replacement and maintenance strategies on maintenance spending ........ 261
6.14 Impact of failure rate on system performance.............................. 262
6.14.1 Effective redundancy..................................................................262
6.15 Impact of system augmentation rate on network composition and
performance............................................................................................... 265
6.16 Summary of findings from system dynamics modelling ............... 271
7 Discussion of regulation of transmission utilities supported by
modelling results ........................................................................................ 274
7.1 Desirable traits in electricity transmission enterprises ................. 274
7.2 Implications for regulation of transmission utilities ....................... 276
7.2.1 Overall revenue..........................................................................276
7.2.2 Revenue based on return on investment ...................................278
7.2.3 Drivers of Operations and Maintenance.....................................280

8
7.2.4 Performance-based incentives in regulation of transmission
enterprises 281
7.3 The effect of changing regulation ................................................ 282
8 Conclusions ................................................................................. 284
Appendix A Sources of Data.................................................................... 291
A.1 General Data Sources and treatment of data ...................................... 291
A.1 Development of Regional Case Data Sets .......................................... 294
Bibliography ............................................................................................... 313

9
List of Figures
Figure 1 Maintenance strategies employed by region.................................. 38
Figure 2 Comparison of direct maintenance spending by region. ................ 39
Figure 3 Estimation of operational life for replacement purposes ................ 40
Figure 4 Estimated rate of replacement of assets - compared by region ..... 41
Figure 5 Replacement of assets by category and region ............................. 42
Figure 6 Comparison of planned outages per circuit per annum ................. 42
Figure 7 Unplanned outages per circuit and per 100 km of circuit by region 43
Figure 8 Profit as intention compared by region. Data are from (CIGRE SC
23 and 39, 2000).................................................................................... 79
Figure 9 Maintenance relative to manufacturers' recommended levels
compared by region (CIGRE SC 23 and 39, 2000). Series from left to
right are: a) more than recommended; b) less than recommended; c)
neither specified. This information is relevant mainly for substation
equipment such as transformers, circuit breakers and protection
equipment. ............................................................................................. 81
Figure 10 Bus configuration schemes in use in various parts of the world... 83
Figure 11 Flexibility compared by region, calculated from data in (CIGRE SC
23 and 39, 2000). Results show significant variability in substation
design practices across regions............................................................. 84
Figure 12 Connectivity compared by region, calculated using data from the
CIGRE study .......................................................................................... 85
Figure 13 Geographic factor is compared by region. Values are calculated
from data in the CIGRE study. ............................................................... 86
Figure 14 Circuit breaker failure rates as a function of age.......................... 93
Figure 15 Influence diagram of transmission asset management relationships
............................................................................................................. 112
Figure 16 Membership functions for climatic factor.................................... 126
Figure 17 The membership functions for parameter load growth– an example
of membership functions estimated from real numeric data, in this case
percentage recent load growth............................................................. 127
Figure 18 The decision tree for the task Investment in new equipment ..... 128
Figure 19 Effect of translating model as a snapshot .................................. 132

10
Figure 20 Actual structure of maintenance spending, from fuzzy model.
(Note equipment replacement and loading subtasks have not been
elaborated, for reasons of diagram clarity)........................................... 133
Figure 21 Comparison of unplanned outages per 100km of circuit and
unplanned outages per circuit from CIGRE study with failure rate from
model. .................................................................................................. 142
Figure 22 Supply point reliability predicted by the model. The arrows indicate
the range of observed values from Table 15........................................ 147
Figure 23 Sensitivity diagram of offline CBM with attitude to new technology
and labour skill for the North American region. X marks the coordinates
used for the North American base case. .............................................. 151
Figure 24 Maintenance spending predicted by model output compared with
CIGRE study data ................................................................................ 152
Figure 25 Comparison of refurbishment as a percentage of direct
maintenance costs (from CIGRE study) with model predictions for
refurbishment levels............................................................................. 154
Figure 26 Replacement of equipment - predicted rate from model compared
with CIGRE Study data ........................................................................ 155
Figure 27 Availability as a function of flexibility for three cases from fuzzy
model ................................................................................................... 161
Figure 28 Availability as a function of maintenance actions increasing
availability for three regions. Availability for WE is sensitive to this
parameter about the current operating point (28)................................. 165
Figure 29 Sensitivity of reliability to connectivity for three regions ............. 166
Figure 30 Combined effect of connectivity and flexibility on reliability for WE
case. X denotes the existing situation................................................. 167
Figure 31 Failure rate as a function of maintenance history and proportion
aged equipment for South America. X marks the estimated levels of
maintenance history and proportion of aged equipment. ..................... 168
Figure 32 Sensitivity of routine maintenance and total maintenance spending
to maintenance relative to manufacturers' recommended levels ......... 201
Figure 33 Impact of use of predictive maintenance on reliability for TNB case
............................................................................................................. 202

11
Figure 34 Sensitivity of reliability to failure rate contribution of load and
climate for TNB case............................................................................ 203
Figure 35 Sensitivity of reliability to flexibility and connectivity for TNB case
............................................................................................................. 204
Figure 36 Availability as a function of live working and combining
maintenance activity by circuit. Current operating point is shown as X.
............................................................................................................. 205
Figure 37 Sensitivity of availability to use of online CBM for Powerlink
Queensland case study. The calculated value of use of online CBM for
current parameters is moderate (43).................................................... 206
Figure 38 Availability as a function of flexibility for the Powerlink case ...... 207
Figure 39 Simple main chain showing ageing process .............................. 220
Figure 40 Simple model of deteriorating condition, failure, refurbishment and
replacement ......................................................................................... 222
Figure 41 Condition deterioration model with deterioration rate a function of
maintenance effectiveness, and refurbishment and replacement of poor
condition equipment a function of predictive maintenance................... 223
Figure 42 Graphical representation of relationship between PCA detected
and use of predictive maintenance ...................................................... 224
Figure 43 Graphical representation of the relationship between maintenance
effectiveness and fraction deteriorating ............................................... 225
Figure 44 This diagram shows the links between failure of poor condition
assets and age-related failures. ........................................................... 226
Figure 45 This simplified diagram illustrates the links between replacement of
poor condition assets and the replacement of assets from different age
groups, for condition-based replacement. ............................................ 228
Figure 46 Variation in condition of assets with (preventive) maintenance
effectiveness fraction, and PCA failing fraction. ................................... 229
Figure 47 Impact of maintenance effectiveness on failure rate.................. 230
Figure 48 Number of assets in different age groupings as a function of
maintenance effectiveness fraction, for MTBF PCA of 2 years. ........... 231
Figure 49 Total depreciated asset value as a function of maintenance
effectiveness ........................................................................................ 231

12
Figure 50 Variation of the condition of equipment within the asset population
with use of predictive maintenance ...................................................... 233
Figure 51 Impact of predictive maintenance on failure rate, if condition-based
replacement/refurbishment criteria are used; MTBF PCA 2 years. ...... 234
Figure 52 Impact of varying use of predictive maintenance on asset age
distribution............................................................................................ 234
Figure 53 Depreciated value of assets decreases as use of predictive
maintenance increases ........................................................................ 235
Figure 54 Refurbishment increases with increasing use of predictive
maintenance ........................................................................................ 235
Figure 55 Depreciated asset value for the case of refurbishment of 50% of
detected PCA and no refurbishment of PCA. The replacement of
detected PCA is held constant at 50%................................................. 236
Figure 56 Comparison of failure rate for 50% replacement of detected PCA
and 100% replacement of detected PCA. Refurbishment of PCA is not
undertaken in either case. MTBF PCA is 2 years for both studies. ...... 238
Figure 57 Total depreciated value compared for cases with 50% and 100%
replacement of detected PCA. Refurbishment of PCA is not undertaken
in either case. MTBF PCA is 2 years for both cases........................... 238
Figure 58 Simplified age-based replacement model: fully depreciated assets
are replaced. Replaced items are assumed to come from PCA and any
remaining assets are replacements of GCA. Refurbishment of PCA is
allowed................................................................................................. 240
Figure 59 Failure rate as a function of maintenance effectiveness fraction for
an age-based replacement model........................................................ 241
Figure 60 Distribution of asset age as a function of maintenance
effectiveness fraction for an age-based replacement model (shown for
MTBF PCA of 2 years)......................................................................... 242
Figure 61 Total depreciated asset value as a function of maintenance
effectiveness for an age-based replacement scenario (illustrated for a
cased with MTBF PCA of 2 years). ...................................................... 242
Figure 62 Failure rate as a function of maintenance effectiveness for age-
based replacement case with no refurbishment and refurbishment at 20%

13
 of detected PCA, with predictive maintenance level 50%. MTBF of PCA
is 2 years in both studies. .................................................................... 243
Figure 63 Total depreciated value as a function of maintenance effectiveness
for age-based replacement with no refurbishment and 20% of detected
PCA refurbished. MTBF of PCA is 2 years for both studies. ............... 243
Figure 64 Comparison of failure rate for condition-based replacement 50% of
detected PCA, and 50% predictive maintenance and age-based
replacement of assets (assets > 40 yrs). ............................................. 244
Figure 65 Comparison for total depreciated value of condition-based
replacement (20% of detected PCA, 50% predictive maintenance) and
age-based replacement (>40yrs) ......................................................... 245
Figure 66 Total depreciated asset value with increasing use of remedial
maintenance of PCA ............................................................................ 247
Figure 67 Failure rate as a function of fraction failed PCA repaired........... 247
Figure 68 Effect of replacement on failure only policy, without remedial
maintenance and with 50% of failed PCA repaired to condition before
failure (MTBF PCA 2 years). ................................................................ 250
Figure 69 Comparison of failure rate for three replacement strategies. Other
parameters: MTBF PCA 5 years, use of predictive maintenance 50%,
refurbishment of detected PCA 20%.................................................... 250
Figure 70 Total depreciated value as a function of maintenance effectiveness
for a replacement-on-failure-only strategy, without remedial maintenance.
............................................................................................................. 251
Figure 71 Illustration of model dependencies when retained revenue and
debt calculations are added to the model. ........................................... 254
Figure 72 Impact on profit, debt and depreciated value of variation of
maintenance effectiveness................................................................... 256
Figure 73 Comparison, for the condition-based replacement policy case, of
retained earnings with and without refurbishment of PCA ................... 257
Figure 74 Effect on retained earnings of optimising asset value compared
with the case with no optimisation........................................................ 258
Figure 75 Comparison of retained earnings for three different replacement
scenarios. An optimisation factor of 0.9 is used, and there is no
refurbishment in each case. ................................................................. 259

14
Figure 76 Retained earnings is presented as a function of maintenance
effectiveness for three replacement strategies and MTBF PCA of 5 years.
(Other parameters: use of predictive maintenance 50%, replacement of
detected PCA 50% in condition-based replacement case; refurbishment
of detected PCA 20% in all cases; optimisation factor 1.0).................. 260
Figure 77 Retained earnings as a function of maintenance effectiveness.
(Parameters: MBTF PCA of 10 years, all other parameters same as
previous case.)..................................................................................... 260
Figure 78 For the North American base case this graph shows how
increasing redundancy reduces the impact of aging equipment on the
reliability of supply................................................................................ 265
Figure 79 A 3% rate of augmentation of assets per annum – clearly
impractical............................................................................................ 266
Figure 80 Simplified model incorporating system augmentation ................ 267
Figure 81 Modelled relationship between load growth and network maturity
............................................................................................................. 269
Figure 82 Modelled relationship between network maturity and value per
asset .................................................................................................... 269
Figure 83 The asset growth rate and total replacement value of assets as a
function of time, and with increasing network maturity......................... 270
Figure 84 Network maturity and load growth as a function of time............. 270
Figure 85 Age distribution of assets as a function of time.......................... 271

15
List of Tables
Table 1 Breakdown of maintenance spending by region ............................. 37
Table 2 Major failure frequency for SF6 circuit breakers 63kV and above,
placed in service after Jan 1978 ............................................................ 96
Table 3 Major failure frequency for GIS in service ....................................... 96
Table 4 Rules generated for the task investment in new equipment.......... 128
Table 5 Summary of model tasks and their inputs ..................................... 134
Table 6 Input Parameters used in regional studies for fuzzy model........... 138
Table 7 Reliability measures from various parts of the world..................... 143
Table 8 Effect of network configuration on supply reliability....................... 146
Table 9 Circuit availability .......................................................................... 147
Table 10 Availability – Model predictions compared with actual data ........ 148
Table 11 Comparison of model predictions with reported maintenance
strategies ............................................................................................. 150
Table 12 Substation equipment 20 and more years of age........................ 157
Table 13 Parameters used to assess availability and their values for AU, NA
and WE cases...................................................................................... 160
Table 14 Outputs from Fuzzy Model – Regional Studies ........................... 162
Table 15 Components of the task maintenance activities increasing
availability for AU, NA and WE............................................................. 165
Table 16 Parameters used to determine reliability and their values for AU, NA
and WE ................................................................................................ 166
Table 17 Parameters used to determine failure rate and their values for the
Asian and South American cases ........................................................ 168
Table 18 Parameters contributing to the task proportion poor condition
equipment with values for Asia and South America ............................. 168
Table 19 Factors contributing to maintenance spending, with values shown
for AU, NA and WE .............................................................................. 169
Table 20 Factors contributing to routine maintenance costs (for AU, NA and
WE)...................................................................................................... 169
Table 21 Inputs to the task refurbishment for three regions....................... 171
Table 22 Input parameters for the Australian Case study .......................... 177
Table 23 Outputs for the Australian case study ......................................... 177

16
Table 24 Inputs to fuzzy model for Malaysian case study.......................... 184
Table 25 Outputs from fuzzy model for the TNB case Study ..................... 185
Table 26 Input data for the Japanese Case Study ..................................... 188
Table 27 Results for the Japanese Case Study ......................................... 189
Table 28 Input data for the US Utility Case Study...................................... 193
Table 29 Results from the US Utility Case Study....................................... 194
Table 30 Input parameters for the UK utility case study............................. 199
Table 31 Results of the UK utility case study ............................................. 200
Table 32 Factors contributing to availability and their values for the Powerlink
Case .................................................................................................... 205
Table 33 Maintenance activities increasing availability for the Powerlink Qld
Case .................................................................................................... 205
Table 34 Data used in fuzzy model and their sources ............................... 293
Table 35 Enterprises in Asia – Business parameters at 1998.................... 296
Table 36 Production of electricity in billion kWh from (World Bank, 2001b) 298
Table 37 Labour cost in $US (World Bank, 2000b).................................... 299
Table 38 Debt and interest coverage for government owned enterprises in
Australia 1997-1998 and 1998-99 (Productivity Commission, 2002) ... 301
Table 39 Credit ratings from some former Soviet Block companies........... 303
Table 40 Credit rating of South American electricity companies from
(Standard & Poor's, 1999).................................................................... 309
Table 41 Enterprises in Western Europe – Business parameters at 1998
(Standard & Poor's, 1999).................................................................... 311

Supplementary Material
Software models on compact disc

17
List of Abbreviations
ACCC Australian Competition and Consumer Commission
AM asset management
CBM condition-based maintenance
CLD causal loop diagram
CT current transformer
DEA data envelopment analysis
DORC depreciated optimised replacement cost
DSS decision support system
FACTS flexible AC transmission system
FERC Federal Energy Regulatory Commission (USA)
FMEA failure modes effects analysis
FMECA failure modes effects and criticality analysis
IPP independent power producers
LCC life cycle costs
MTBF mean time between failures
MV membership value (fuzzy sets)
NERC National Electricity Reliability Council (USA)
ODV optimised deprival value
O&M operations and maintenance
ROCOF rate of occurrence of failure (repairable systems)
RCM reliability-centred maintenance
RBSE rule-based schedule evaluation
SFD stock flow diagram (in system dynamics modelling)
SVC static VAR compensator
TFP total factor productivity
TNSP transmission network service provider
VIMS visual interactive modelling system
VT voltage transformer
WACC weighted average cost of capital

18
Abbreviations used in System Dynamics models
AA aged asset
GCA good condition asset
MLA mid life asset
PCA poor condition asset
YA young asset
VAA very aged asset

Region identifiers
A Asia
AU Australasia and South Africa
EE Eastern Europe (former Soviet Block countries)
NA North America
SA South America
WE Western Europe

19
Statement of Original Authorship

I declare that the work presented in this thesis is, to the best of my
knowledge and belief, original, except as acknowledged in the text, and that
the material therein has not been submitted, either in whole or in part, for a
degree at this or any other university.

20
Acknowledgements
I wish to acknowledge the members of CIGRE Joint Working Group 23/39
Reliability and Maintenance for their efforts in undertaking “An international
survey of maintenance policy and trends” and thank them for permission to
use the survey data in my research.

Thanks also to Powerlink and Ergon Energy for technical and financial
support of this project. The financial support from these two companies
contributed to the cost of travel, enabling me to undertake interviews with
representatives of transmission companies in Europe and the USA.

I wish to acknowledge the assistance of the following people, who shared

with me their knowledge of the electricity industry:
Michael Allison and Peter Jay of National Grid Company UK
Andrew Bower of Yorkshire Electricity and Gas
Tom Sherard and Dan Melkior of Duke Power, North Carolina, USA
Patrick Lee of San Diego Gas and Electric
John Teixiera of NGC USA
Steven A Stewart of Scottish Power
Mohammed Junaizee Mohd Noor of TNB, Malaysia
Peter Brennan of Ergon Energy, Australia
Brian Sharp and David Gibbs of Powerlink, Queensland Australia.

My supervisors – Associate Professor David Birtwhistle (Principal

Supervisor), Adjunct Professor Nick Hastings and Dr Stewart Bell – have
provided editorial support and critical comment that has improved the
standard of the research. Thank you for this support.

Finally, I wish to thank my family Associate Professor Dorothy Watts, Dr

Keith Watts, my husband Nick and children Peter and Sarah for all their
support, encouragement and patience.

21
1 Introduction

1.1 Questions prompting this research topic

Electricity is an essential service that contributes to prosperity and quality of
life. To support the provision of high quality electricity supply at lowest
possible cost, a well-designed, efficiently operated and maintained, reliable
transmission network is required. For transmission enterprises, asset
management (AM) is a core function and is strongly linked with both revenue
and reliability of electricity supply. Factors that impact on performance are
therefore of great interest to those who regulate transmission and those who
provide electricity transmission services.

A recent international survey of AM practices in transmission enterprises

(CIGRE SC 23 and 39, 2000) revealed that the performance and policies of
electricity transmission enterprises vary significantly in key areas. These
differences are observable even when the statistics are compared by region.
Examples include:
• Significant differences in maintenance strategies and direct
maintenance spending;
• Differences in replacement criteria and rate of replacement of assets;
• Considerable variation in the management of planned and unplanned
outages.

This present research was prompted by the question: Why do the policies
and practices of transmission enterprises vary appreciably across different
regions of the world when all transmission networks serve essentially the
same purpose? A question that follows from this is: Do these variations
reflect differing circumstances or disparities in performance? It was
anticipated that exploration of these questions would reveal avenues for
improving the performance of transmission businesses and the networks they
manage.

22
1.2 Background to the research

1.2.1 Transmission networks

Transmission networks provide an essential service to the community. The
quality of that service directly impacts on work productivity and quality of life.
Thus the electricity network is seen by many countries as being of strategic
importance. Transmission networks are intended to be maintained, to
provide an adequate service quality, indefinitely.

Electricity transmission networks connect generators of electricity with users

of electricity. Domestic users of electricity are connected through distribution
networks. The point of connection between a distribution network and a
transmission network is often described as a bulk supply point. Sometimes
large industrial consumers are connected directly to the transmission
network.
The distinction between transmission or distribution network is not completely
clear - different jurisdictions use different voltages to define the point at which
the subtransmission/distribution network joins the transmission network.
However, transmission networks are often defined as operating at voltages
above 100kV or above 66kV. Such networks typically operate with two or
more different voltage levels, the higher voltage parts of the network intended
to transmit higher power flows. The design of transmission networks usually
incorporates a higher degree of redundancy than is generally found in
distribution networks. This is considered necessary to provide reliability of
supply to customers, since transmission lines usually supply more customers
per circuit1 than do distribution circuits.

The structure of the transmission network can be visualised as a set of nodes

joined by connectors. The connectors are overhead transmission lines and
underground cables; the nodes are the substations. Transmission line
circuits include the switchgear at either end of an overhead transmission line
or underground cable as part of the connector.

1
including those connected through bulk supply points

23
1.2.2 Characteristics of transmission system plant
Transmission equipment can generally be described as capital intensive,
robust, long-lived and not easily relocatable. Equipment lives of forty and fifty
years are not uncommon, and some networks have operating equipment up
to eighty years old.

Equipment in transmission networks can be classified as either line/cable or

substation equipment. Within the substation one would expect to find
switchgear comprising circuit breakers, used for switching and protection,
and disconnectors used for disconnecting and earthing circuits. Other
equipment types may also be present: transformers convert energy from one
voltage level to another; reactors and capacitors provide reactive control to
maintain the voltages in the network within acceptable limits and static VAR
compensators provide reactive support, and can be designed to contribute to
system stability. Measurement devices such as voltage transformers (VT)
and current transformers (CT) are also likely to be present in a typical
substation.

The substation equipment described in the previous paragraph is categorised

as primary plant. Secondary plant is also present. It includes the control
circuits, protection relays and communications equipment associated with the
primary plant, and backup power supply for the secondary systems. At
substations there are also infrastructure assets including buildings, fences,
earth mats and busbars.

Overhead transmission lines are distributed assets, the main parts of which
are conductors, insulators and towers. Conductors may or may not include
earth wires, depending on the design policy of the network, and the lightning
incidence rate. A recent trend has been to use overhead earth wire with an
embedded optical fibre for wide band-width communications. Series reactive
devices may be used on transmission circuits for the purpose of enhancing
system stability.

24
Underground cables can be single or three-phase, and at transmission
voltage levels, are usually constructed with a protective metal sheath around
the cable. The main elements of cables are the conductors and associated
terminations.

1.2.3 Transmission enterprises

There are many different structures in electricity industries world-wide. In
some jurisdictions transmission enterprises are stand-alone organisations,
which may be privately-owned (for example the National Grid Company of
the UK, SPI Powernet in Victoria, Australia), publicly-owned (Powerlink
Queensland and Transgrid New South Wales, Australia) or partially privately-
owned like Tenaga Nasional Berhad of Malaysia. Transmission entities may
be part of a vertically integrated structure, with generation and distribution
(many utilities in the US and all Japanese utilities), or with either generation
or distribution. These vertically integrated organisations may also be
privately (eg Japanese utilities, Endesa of Chile) or publicly owned (eg
Electricité de France). In many countries the structure of the electricity
industry and regulations pertaining to it have changed in the past twenty
years, and some countries are still undertaking structural reform within the
electricity industry. (This is discussed in detail in Section 2.2.1.1).

Most transmission organisations or business units within organisations have

the function of managing the transmission network, including all asset
management functions. In some cases another organisation may have
responsibility for augmentation decisions (eg Victoria, Australia). Under
some arrangements the transmission organisation is also the system
operator (eg NGC). In most arrangements the transmission organisation
does not trade in electricity, but there are exceptions to this too (eg Ukraine).

1.3 Defining asset management

Various definitions for asset management have been recently proposed.
According to Hastings (Hastings, 2000):
Asset management starts with a business or organisational
objective, and is the set of activities associated with
identifying what assets are needed, acquiring them,
supporting and maintaining them and disposing or renewing

25
 them so as to effectively and efficiently meet the desired
objective.

The International Infrastructure Management Manual (Association of Local

Gov. Engineering NZ and National Asset Management Steering Group Aus.,
2002) states:

Asset management [is] the combination of management,

financial, economic, engineering and other practices applied
to physical assets with the objective of providing the required
level of service in the most cost-effective manner.

The CIGRE Australasian Asset Management Working Group defines asset

management (for the electricity industry) as follows:
Asset Management is a set of business processes
concerned with developing, operating and maintaining the
assets of an organisation to meet the desired requirements
of the customers and the shareholders of the business.
These desired requirements usually encompass cost,
performance, safety and environmental outcomes as well as
a secure electricity supply.

Jones of Yorkshire Electricity uses a three tiered model of asset

management (Jones, 1997):
The nature of this model is that it reaches from the strategic
level business outputs to which the asset manager is seeking
to contribute; identifies the phases of the asset management
life-cycle that must be managed in order to deliver these
values and finally the model identifies the processes in which
the Asset Manager will be engaged as the life-cycle phases
are managed.

Beardow (Beardow, 2003) offers a broader view of asset management: An

asset is an item or property owned by an individual or business which has
monetary value. Three types of assets are identified: physical assets (plant
and equipment), financial (financial instruments, equity accounted
investments) and intangible assets (operating licence, knowledge and skills
of staff).
Asset management involves:
• Managing physical asset base;

26
 • Managing asset’s financial base i.e. its valuation; regulatory pricing
reviews; and
• Managing the brand, for new “brownfields” and potential “greenfields”
business.
The purpose of asset management is to turn assets into a revenue stream.

These definitions reflect different viewpoints – the academic, the public

infrastructure manager, electricity industry professionals and the economist,
but they have certain elements in common:
• Asset management must align with the nature and values of the
organisation.
• It must support the goals of the organisation and its shareholders.
• It includes the management of physical assets of the organisation over
the various stages of their life-cycles.
Beardow’s definition is consistent with this summary but also encompasses
non-physical assets.

Management of physical assets is the primary focus of this present study.

Asset management for physical assets may be categorised into three levels
of application:
• Equipment level asset management;
• Process level asset management; and
• Strategic level asset management.

Equipment level asset management includes such activities as analysis of

failure modes, development of condition-monitoring of equipment and
assessment of remaining life of equipment. The process level expands the
focus to include the whole life-cycle of plant. It encompasses processes
(discussed in Chapter 2) such as Reliability Centred Maintenance, Life Cycle
Costing, maintenance optimisation methods, maintenance management
systems and economic justification of online condition-monitoring and
optimisation of spares inventories. The strategic level of asset management
is closely aligned with the business function of the organisation. In strategic

27
asset management the emphasis is on developing policies and practices that
support the short and long-term goals of the organisation taking into account
a broad range of drivers.

1.4 Purpose of the research

The purpose of this research is to
• Identify factors that influence policies and practices of asset
management in electricity transmission enterprises;
• Investigate how these factors interact to constrain or enhance
the performance of transmission enterprises; and
• Develop methods to identify opportunities for performance
enhancement within the framework of existing external, system-
related and environmental drivers.

1.5 Scope
It is at the strategic level of asset management that this thesis seeks to make
a contribution to knowledge. Therefore the analyses and models presented
take a broad, high-level view of the subject. Although the study focuses on
the management of physical assets of the organisation, some references are
made to financial management when this impacts on management of
physical assets.

The study is intended to be applicable to a wide range existing transmission

enterprises and networks. An exploration of diversity in transmission policies
and practices in different parts of the world is an important facet of the
research. The temporal setting for data used in regional case studies is circa
1998, for reasons of convenience and data consistency. It is intended that
the model apply to transmission networks as they exist in the present day.
Applications set in the future or past might need to consider criteria additional
to or different from those used to develop the current models.

1.6 Methodology
The research may be described broadly as a data acquisition phase and a
model implementation and exploration phase. Within the implementation

28
stage two methodologies are presented, which enable different facets of the
research topic to be explored.

In the data acquisition stage, factors of influence for transmission asset

management are identified and their relationships are explored. The drivers
are classified according to external, internal and business-related, and
system- and environment-related categories. Literature review reveals little
documentation linking these factors of influence, or relating them to
performance of transmission enterprises. The structure of the model is
therefore painstakingly pieced together from many fragments of information;
the analysis draws on an extensive range of literature, together with expert
interviews of industry personnel in different parts of the world and personal
knowledge of the electricity industry. The conceptual model of transmission
asset management thus constructed is compactly represented as a causal
loop diagram (CLD).

In the second phase of the research, methodologies for implementing the

model and testing different scenarios are investigated. A process for
selecting appropriate methodologies for strategic-level research in asset
management, implementing, verifying and validating such models is
developed. The research problem is not amenable to solution by traditional
engineering modelling techniques, so modelling methods from other
disciplines are employed: a multi-level fuzzy rule-based model, and a system
dynamics representation. These quite dissimilar implementations allow
different aspects of the problem to be investigated.

The structure of the multi-level rule-based model is derived from the CLD,
and the rules are based on the relationships between factors of influence.
The model is validated against data sets representing composite regional
and individual utility cases. Reasons for variations between regional cases
and potential means of improving performance in individual cases are
explored.

29
The second implementation involves the use of system dynamics models to
explore core parts of the asset management model in more detail. The
models are implemented and tested incrementally. A model linking ageing
and condition degradation of assets is developed. From the age distribution
of the asset population, the value of the asset base and its impact on
revenue can be determined. From the condition distribution, the performance
of assets under various maintenance and replacement strategies can be
examined. By combining the age and condition models it is possible to
explore the effects of various combinations of strategies on the performance
of equipment and profitability of the company. The methodology employed
for this research offers significant potential for application to a range of asset
management problems.

1.7 Key Assumptions, Key sources

The “International Survey of Maintenance Policies and Practices” (CIGRE SC
23 and 39, 2000) provides a significant portion of the data used in the
research. Case studies of individual utilities have been developed combining
data from this CIGRE survey with interviews of industry representatives
and/or published information from technical papers, electronic sources and
business magazines.
Key assumptions made in the thesis are
• That data grouped into regions to form composite case data sets can
be used to test a fuzzy rule-based model of transmission asset
management. (This assumption is more valid for regions having less
diversity, particularly in the level of economic development. To avoid
over-reliance on this assumption a set of company-based case studies
is also used for validation);
• That the ability to predict outcomes across a range of data sets (both
regional and individual utility cases) is sufficient to validate the model,
to the extent that validation is possible with this kind of application;
• That available data are of sufficient quality to enable a fuzzy rule-
based model to predict AM outcomes; and

30
 • That conclusions can be drawn from system dynamics models using
assumed data, provided that the data are with reasonable ranges in
their context of use.

1.8 Limitations
Data limitations have been a major concern in this research:
• Regions tested have been limited to those represented in the CIGRE
study.
• Data quality and accessibility have been a major source of uncertainty.
Certain important parameters have had to be estimated. Other data
have been of questionable quality. Ambiguities also exist in the
CIGRE study data, and answers in many questions are incomplete.
• Mismatches arise between data from difference sources.
These limitations have led to the need to use methodologies that are robust
in the face of uncertainty and imprecision of data. A lack of data from certain
parts of the world has meant that the fuzzy model has not been tested for
some of the more extreme cases possible. Likewise, for the system
dynamics model, the investigation has been limited to well-documented
cases, for which sensible parameters can be obtained.

1.9 Structure of this thesis

Chapter two reviews the literature relating to asset management in the
electricity transmission sector, with particular emphasis on the policy or
strategic level of asset management, but also identifying important
developments in asset management processes. Included in this is a review
of differences in AM policies and practices of transmission enterprises
revealed in the CIGRE “International Survey of Maintenance Policies and
Trends” (CIGRE SC 23 and 39, 2000). There has been little work taking an
holistic view of the subject, so the chapter serves to highlight current issues
and concerns of asset managers. A case is made for the originality of and
need for research of this type.

The categories of influences on transmission asset management are defined

in Chapter three. This is followed by a detailed examination of the
relationships between different factors and how they influence practices and

31
performance of transmission enterprises. A conceptual model in the form of
an influence diagram (or causal loop diagram) is developed. This model
forms the basis for implementations described in Chapters 5 and 6.

Chapter four examines the nature of the model presented in Chapter 3, and
investigates ways of implementing this model with a view to
• Verifying and validating the model, and
• Gaining insights into the research topic.
The process of identifying suitable methodologies is explained and two
approaches are selected. One is to interpret the causal loop diagram of
Chapter 3 to a more detailed stock-flow diagram form, and use a system
dynamics icon-based program to examine the dynamics of the model. The
other approach employs the fuzzy-logic capabilities of a rule-based expert
system shell to examine the relationships in the model, and compare
predicted results with published performance information.

Chapter 5 describes the implementation of a multi-level fuzzy model in detail

and present results from application of this model. There are two sections on
results. The first examines the results from studies with data aggregated on
a regional basis, and the model’s ability to predict regional variations in
assets management, while the second looks at case studies of individual
enterprises. The potential value of such tools in strategic asset management
is then discussed, with reference to the individual case studies.

In Chapter 6 maintenance and replacement policy and practices are

examined in more detail using system dynamics techniques. Chapter 6 also
investigates possible effects of these policies on the depreciated value of
assets, revenue derived from return on assets, profitability (taking into
account capital expenditure and debt), and on the performance of the
network. Finally the impact of system augmentation is examined. The
findings from this research are then related back to observations about
regional differences in transmission asset management policies and
practices.

32
Chapter 7 contains a discussion of regulation, in the light of findings from this
research. Finally, in Chapter 8 the main findings from the research are
summarised, and the significant contributions to knowledge highlighted.

33
2 Literature review
Historically the focus of asset management was on existing equipment,
particularly maintenance of that equipment. After World War II, as
manufacturing became more mechanised, there was an increase in interest
in processes, particularly in the areas of maintenance and replacement.
Successful processes such as Total Productive Maintenance (TPM)
(Nakajima, 1986) and Reliability Centred Maintenance (RCM) (Moubray,
1992) were developed. Maintenance planning and scheduling was examined
to improve the efficiency of maintenance work.

The period since the 1960s has spawned a large number of research papers
on maintenance optimisation and replacement strategies (see Section 2.3.4).
The penetration of this research into industry generally has been slow
(Dekker, 1996), and the electricity transmission sector has not been faster to
take up these methods. However, Decision Support Systems, based on
operations research methods are starting to become commercially available
(Dekker, 1996). Use of these systems is likely to become more widespread in
the next ten years.

Over the past twenty years the narrow focus of asset management has been
expanded to include the whole life-cycle of plant – planning, procurement,
commissioning, operations, maintenance, replacement and / or refurbishment
decommissioning and disposal. Life cycle costing (Standards Australia,
1999b) has become an established process for evaluating different
purchasing options, operational and maintenance strategies.

The current understanding has evolved further to consider asset

management as a business function, as is evident from the AM definitions in
Section 1.3. In this thesis this will be referred to as Strategic level asset
management. For electricity transmission enterprises asset management is
a core part of the business. It is at this level that major business strategies
relating to assets are set. Strategic Level AM decisions are usually based on

34
a broader range of considerations than those taken into account at the
process or equipment levels of asset management. These decisions rely
heavily on information generated at the lower levels of asset management.
Strategic level asset management can be seen as driving much of the
research at the process and equipment levels.

The remainder of this chapter is in two parts: the first part introduces an
international survey that supplied the impetus for this research and much of
the data used in preparation of case studies. In the second part of the
chapter, other literature pertaining to strategic asset management of
transmission enterprises will be reviewed; trends in process and equipment
level asset management will also be described, but in less detail.

The survey, undertaken by international industry body CIGRE, relates to

asset management policies and practices in electricity transmission
enterprises. Although its report title refers to “maintenance” the survey
covers outage management, replacement policies, refurbishment, spares
policy, outsourcing and key performance measures as well as references to
maintenance strategy, but it does not extend to network augmentation or
disposal of assets.

2.1 The CIGRE Study

In February 2000 the Study Committees 23 and 39 Joint Working Group on
Maintenance and Reliability published “An international survey of
maintenance policies and trends” (CIGRE SC 23 and 39, 2000) which is also
referred to as “the CIGRE study” in subsequent pages of this thesis. There
were 64 respondents to the survey comprising 28 from Western Europe
(WE), 9 from Asia (A), 6 from Eastern Europe (EE) (former Soviet Block
countries), 7 from North America (NA), 5 from South America (SA), 8 from
Australasia and South Africa (AU) and one from the Middle East (ME). The
Middle Eastern response has not been used for analysis in this thesis.

35
Some attempt was made in the survey to obtain information on performance
outcomes (such as reliability and availability), but these results are very
incomplete. The survey also suffers from ambiguities in some questions
and/or responses which must be taken to account when comparing results.
Furthermore there is some information important to the present investigation
that is not recorded in the study. It is particularly sparse in the areas of
business/financial and environmental drivers. Nevertheless the CIGRE study
provides a window of insight into the drivers of asset management in different
utilities and regions that is not available from any other source.

The authors of the CIGRE study have kindly permitted use of the raw data in
this thesis. The company details have been stripped from this data for
reasons of confidentiality, but other details such as region and ownership-
type are available for comparison.

Despite the fact that transmission networks serve the same purpose world-
wide and all have similar types of equipment, the CIGRE study reveals
significant differences in the ways in which these networks are managed.
Some of these differences also manifest when comparison is made by
region, suggesting that common practices within a region may exist. In the
following sections some of the major differences highlighted by the survey
will be presented. Factors of influence identified from the survey will be
introduced in Chapter 3. Performance outcomes from the survey, in addition
to those mentioned here will be presented in Chapter 5, supplemented by
information from other sources.

2.1.1 Regional differences in maintenance practices

Comparison of the breakdown of maintenance costs between different
activities shows significant variation. For instance, in Table 1, North America
has much higher use of corrective maintenance than Western Europe, and
much lower proportion of spending on refurbishment. Eastern Europe, Asia
and South America spend much higher proportions of their maintenance
budget on refurbishment than Western Europe, North America and
Australasia/South Africa.

36
 Table 1 Breakdown of maintenance spending by region
Preventive Corrective Refurbish/Repair Other
A 54% 11% 29% 6%
AU 53% 19% 17% 11%
EE 46% 16% 30% 8%
NA 46% 35% 8% 12%
SA 57% 10% 30% 3%
WE 59% 15% 18% 8%

When maintenance strategy is compared there is also significant regional

variation. Respondents were asked to nominate major maintenance strategy
employed and were given the option of answering this question either as a
general answer and/or for specific equipment categories. A cross indicates
an affirmative; a blank can be negative or no answer. In some cases more
than one major strategy was nominated for the same equipment category.
Corrective maintenance has not been defined for this question, but in a
subsequent question it is defined as “the additional work it is found necessary
to do arising from equipment defects and failures, etc”. This includes
remedial (post-failure) and corrective maintenance (to avoid failure).

To analyse this data it has been assumed that a general affirmative is

equivalent to affirmative in all categories, unless individual answers have also
been given, in which case the total is the sum of the individual category
responses. If no response was given in any category then the answer was
not included in the average.

Figure 1 shows, that while time-based preventive maintenance is still the

most common strategy overall, it is less important in South America and
Australasia/South Africa. In Australasia/ South Africa more use is made of
predictive maintenance strategies (online and offline condition-based
maintenance), and Australasia has the highest use of online condition-based
maintenance as a strategy. In North America the use of predictive
maintenance as a strategy is relatively low – more reliance appears to be
placed on traditional routine preventive and corrective/remedial maintenance.

37
 6.0

strategy is used (max 7)

Avg categories where
5.0
4.0
3.0
2.0
1.0
0.0
A AU EE NA SA WE
Region

Corrective Time-based Offline CBM Online CBM

Figure 1 Maintenance strategies employed by region.

2.1.2 Regional differences in maintenance spending

The CIGRE study also asked respondents to give their direct maintenance
spending in US dollars. In the CIGRE study report a composite size
parameter was used to normalise direct maintenance costs for comparison
purposes. This size parameter is developed from two parts, Part 1 based on
energy transmitted, Part 2 based on assets. Total replacement value would
be a better comparator for maintenance spending, but the asset value data in
the CIGRE study are rather incomplete. Total number of assets could also
be used. As a compromise the asset component of the size parameter (Part
2) has been employed as a basis for comparison in this analysis (Figure 2).
A complete description of this formula may be found in Section 5.2.5.

38
 Normalised maintenance cost / 9
asset value component 8
7
6
5
4
3
2
1
0
A AU EE NA SA WE
Region

Figure 2 Comparison of direct maintenance spending by region.

The maintenance spending data from South America are very limited, there
being only two responses from this region. The South American data are
also a little surprising in the light of the fact that four of five utilities reported
doing more maintenance than recommended by manufacturers. Note also
that there is considerable variation within regions for this analysis, especially
in the Asian results. The analysis suggests that Western Europe has the
highest maintenance spending per asset value dollar, Australasia, Asia and
Eastern Europe a moderate level, North America slightly lower and South
America low. The CIGRE study published results contain an error in one of
the Australasian results (result 10 times higher than true figure) that has been
corrected for this analysis.

2.1.3 Regional differences in replacement criteria

The CIGRE study also asked respondents if they made a practice of
estimating life of assets for the purpose of replacement planning.
The respondents could answer in general or on an equipment category basis.
In general the most common equipment types on which life was estimated
were the transformers and switchgear. Absence of an answer was taken to
be a negative response in this case. A positive general answer was
assessed to be equivalent to all categories positive, unless individual

39
category responses were also given. The analysis in Figure 3 shows a wide
variation between regions. In North and South America the practice of
estimating plant life appears to be significantly less common.

6
assessment (max 7)
average equipment
categories for life
5
4
3
2
1
0
A AU EE NA SA WE

Figure 3 Estimation of operational life for replacement purposes

Figure 4 shows the estimated actual replacement of assets compared by

region. Respondents were given the option of a general response and/or
individual category responses. Answers were in percent over four years. In
this analysis, for cases where individual responses were given in some
categories, null responses in other categories were treated as zero percent
replacement. If only a general response was given this was taken to be the
average across all categories. For cases where only individual responses
were given, the average of all categories was taken to be the average
replacement for the system (approximate, since equal numbers of equipment
in each category are unlikely). Since values vary widely within regions the
median value of all responses for the region has been used. Since the
assessments in this graph give only estimates of replacement rates the
individual category data has been presented in Figure 5 to facilitate
comparison.

40
 Median replacement - all categories or
9.0

general (% over 4 years)

8.0
7.0
6.0
5.0
4.0
3.0
2.0
1.0
0.0
A AU EE NA SA WE

Figure 4 Estimated rate of replacement of assets - compared by region

The results show that Eastern Europe was undertaking a large amount of
replacement of equipment at the time of the survey. Inspection of individual
category data (Figure 5) suggests that this was mainly in protection
equipment category, but also included moderate levels of replacement of
switchgear and control equipment. The assessments in Figure 5 have been
prepared assuming that null values are equivalent to zeros if the respondent
has given numeric values in any other equipment category for this question.
The data shown in Figure 5 for South America represent only two responses.

41
 Median replacement % (over 4 years)
16
14
SWGR
12
TRANSF
10 LINES
8 TOWER
6 CABLES
PROTN
4
CONTRL
2
0
A AU EE NA SA WE

Figure 5 Replacement of assets by category and region

2.1.4 Regional differences in outage management

planned outages per cct

5
4
3
2
1
0
A AU EE NA SA WE

average median

Figure 6 Comparison of planned outages per circuit per annum

There are also variations in planned outages in different regions (Figure 6),
although the difference between average and median values suggests
variations within regions as well. Note that these data are rather sparse –

42
there are only two responses from South America and three from North
America.

5
4 Unplanned
outages
3 per100km

2
Unplanned
1 outages/cct
0
A AU EE NA SA WE
Regions

Figure 7 Unplanned outages per circuit and per 100 km of circuit by region

Likewise, as Figure 7 illustrates, there are significant variations in reported

forced outage rates between regions.

2.2 Strategic Level Asset Management

This second part of Chapter 2 reviews Strategic Asset Management topics
pertinent to electricity transmission systems. Typical themes include:
• Assessment and management of regulatory impacts;
• Assessment and management of market impacts;
• Risk management;
• Long term strategies related to replacement of equipment;
• Alignment of performance with customer needs;
• Use of external service providers to support asset management
activity;
• Skills retention and development;
• Alignment of maintenance activity with the operational strategy for
assets;

43
 • Performance measurement, performance indicators and
benchmarking;
• The impact of interconnection of systems;
• The impact of embedded generation;
• The potential benefits and opportunities associated with new
technology; and
• Likely developments in environmental management and regulation.
These will be examined in the following sections.

2.2.1 Assessment of Regulatory impacts on transmission

2.2.1.1 Changes to structure of the electricity industry

1990 to the present has been a period of upheaval for the electricity industry
world-wide, with many utilities subjected to changes in regulation, market
operation, and industry structure. In this period there has also been an
increase in the participation of private investment in the industry (Izaguirre,
1998), which has seen a reassessment of goals and priorities for some
organisations.

Changes in structure, and in some cases, ownership, within the electricity

industry have occurred in many countries including the United Kingdom
(Newbery, 1994), Chile, Argentina and other countries in South America
(Lalor and Garcia, 1996), Eastern Europe (von Hirschhausen and Opitz,
2001), New Zealand (Ministry of Economic Development, 2000) and Australia
(NEMMCO, 1999). Changes are occurring still in Canada and the United
States of America (EIA, 2000), parts of Europe (Sioshansi, 2001) and Asia eg
(EIA, 2001d, EIA, 2002c). Typically the electricity reform process has seen a
move from vertically integrated, publicly owned utilities towards structures
that separate the different functions, particularly generation separated from
the “wires businesses” of transmission and distribution. Exceptions to the
state of ownership include the USA, Germany and Japan, where privately
owned utilities are more common.

Motivations for these reforms have varied and the forms of the resultant
electricity industry structures have also varied from one jurisdiction to

44
another. Historically the electricity industry was viewed as an asset of
strategic importance to the country, and best managed as a single entity
(Cope, 2000). In the early 1980s this view was challenged by new economic
theory insisting that free and competitive markets were better at delivering
basic services than government agencies (Cope, 2000). For the electricity
industry it was perceived that competition, particularly among generators and
retailers, would lead to an improvement in efficiency of the electricity industry,
with resultant lower prices to consumers. In some countries (eg Argentina
(Lalor and Garcia, 1996 ), Victoria Australia (Theunissen et al., 1998)),
electricity restructuring and privatisation was seen as a way of providing
additional capital and/or a mechanism for retiring public debt by selling
assets to private investors.

Early reform initiatives typically involved the complete de-integration of

utilities into different functions (eg UK (Newbery, 1994), much of Latin
America (Fischer and Serra, 2000), NZ (Ministry of Economic Development,
2000) and some states of Australia (EIA, 1997)). An exception to this was
Chile, which initially did not require separation of functions (Fischer and
Serra, 2000). More recent approaches to reform have been more cautious,
recognising the individual social and economic circumstances in some
countries, and existing industry structure might not justify the implementation
of full restructuring (Cope, 2000). In Asia, notably, competition in the
electricity industry has been fostered by encouragement of Independent
Power Producers (private investors in generation plant) (EIA, 2002c, EIA,
2002b, EIA, 2001c, EIA, 2002a).

2.2.1.2 Regulation within transmission

The changes described in the previous chapter were in part brought about by
efforts to increase competition in the electricity sector. Competition policy
seeks to facilitate effective competition in the interests of economic efficiency
while accommodating situations where competition does not achieve
economic efficiency or conflicts with other social objectives (Hilmer, 1993).
Economic efficiency, as defined in the Hilmer report (Hilmer, 1993 p 4)
includes:

45
 • Productive efficiency (achieved when individual firms produce goods
or services at least cost);
• Allocative efficiency (achieved when resources used to produce goods
or services are allocated according to their highest value uses); and
• Dynamic efficiency (achieved when firms make ongoing changes to
technology and products in response to changes in consumer tastes
and productive opportunities).

The application of competition principles to the electricity industry has seen

the introduction of competitive markets in generation and retail sectors in
many countries, but the transmission sector is presently viewed as a natural
monopoly (Dismukes et al., 1998, Newbery, 1994). A natural monopoly
arises when the total cost of providing goods or services is less when
provided by a single firm than when provided by a set of firms acting in
competition. Under situations of natural monopoly it is recognised that
competition is not consistent with economic efficiency (Hilmer, 1993 p5). As
a result transmission is usually treated as a regulated monopoly, although the
degree and precise nature of regulation has varied considerably between
countries (Newbery, 1994).

Regulation is generally seen as necessary to avoid the abuse of market

power by monopolies, in order to ensure fair prices to consumers. It is also
seen as a surrogate for competition (ACCC, 1999a) and a means of
promoting allocative efficiency (Bell, 2002). Allocative efficiency also implies
productive efficiency (Mc Taggart et al., 1996).

Regulation of electricity transmission organisations, usually includes an

obligation to supply and a guarantee of security, stability and safety
(Newbery, 1994). For regulation to be credible, it must also provide sufficient
and stable revenue to transmission service providers, in order promote
adequate investment and ensure the long-term viability of the transmission
network. Under electricity market systems an additional feature of regulation
is a requirement to provide non-discriminatory open access to the
transmission network to third parties.

46
The types of regulation existing today for transmission businesses may be
classified into the following categories, (listed most favourable to least
favourable for credit quality by Standard & Poor’s assessment) (Greer, 2001):
• Rate based regulation (Cost-plus regulation);
• Self regulation (Light-handed regulation);
• Revenue cap regulation;
• Price cap regulation (Rate cap regulation);
• Non-transparent regulation;

Under rate-based regulation the utility is allowed to achieve a reasonable rate

of return on investment and to recover costs of operations and maintenance.
This was traditionally the most common form of regulation, and is still
predominant for utilities in North America (Greer, 2001). Electricity rates are
set by the regulator. There is some risk exposure for the utility based on
changes in load and thus revenue (eg milder weather may reduce electricity
usage, severe weather is likely to increase usage). Newbery (Newbery,
1994) writes “rate of return regulation is an attractive way of underwriting
property rights and ensuring the ability to cheaply finance expansion, [but] it
does so at the cost of inflexibility, risk aversion towards new technologies that
may lower costs, and a reluctance to cross institutional boundaries to seek
out lower cost solutions, as well as the usual tendency to over-costly
investment and employment”.

Under self-regulation, utilities set their own price levels and pricing
mechanisms, generally with the requirement to justify these levels to
customers (eg Sweden (Greer, 2001), New Zealand (Read, 1997)).

For revenue cap regulation the total revenue which may be collected by the
transmission business is capped. Revenue is generally calculated based on
return of and on investment plus operations and maintenance costs. In CPI-
X schemes (such as used in Australia (ACCC, 1999a)) the operations and
maintenance component of revenue is adjusted for inflation and by a factor

47
“X“ to encourage efficiency improvements. Adjustments are made for over or
under-collection of revenue. Financial risk arises from a number of sources
including:
• The valuation of assets;
• The calculation of rate of return on assets; and
• Determination of the value of the “X” factor.

Price cap regulation is similar to revenue cap regulation except that no

adjustments are made for over- or under-collection of revenue within the
regulatory reset period. Under price capping, the utility is allowed to increase
the weighted average of the prices of its services by no more than the
increase in the CPI less a percentage amount, X. This implies that there is
an incentive to increase output (Albon, 2000).

Non-transparent regulation is where there are no procedures or guidelines on

how to set revenues.

Incentive-based regulation (also known as performance-based regulation) is

becoming more widespread (although application has been more common in
the distribution sector of the industry than in the transmission sector (Bell,
2002)), the most widely used schemes being based on price cap, revenue
cap models or targeted incentives (Jasamb and Pollitt, 2001). Operating
efficiency is often targeted by incentive regulation (Jasamb and Pollitt, 2001)
but quality of supply issues (such as reliability, forced outages, availability,
and/or voltage excursions), may also be the subject of incentives (Jasamb
and Pollitt, 2001, ACCC, 2002b, Nolasco and Ferreira da Silva, 2002, ENRE,
2001).

Under the Yardstick form of incentive-based regulation, utilities are

encouraged to strive for lower costs by inducing them to compete with one
another for cost reductions (Albon, 2000). Some form of benchmarking is
usually used but differences in circumstances of utilities complicate the
assessment of relative performance.

48
2.2.1.3 Research on regulation, structural reforms and governance
Research has been undertaken largely from a public policy/economics
perspective. Topics have included the structure, ownership and method of
regulation of the electricity industry, and whether reforms have achieved the
desired results, transmission pricing mechanisms, regulatory governance and
regulatory risk and the associated topics of congestion management and
stimulating investment.

There has been little research relating the extensive changes in regulation to
the asset management within utilities, even though there have been
significant changes to external drivers for asset management. However an
examination of the literature suggests several ways in which regulation
impacts or should impact on transmission asset management.

The first and most fundamental impact of regulation is on the adequacy of

revenue (Newbery, 1994). This has a major effect on the company’s ability
to operate, maintain and extend the network in order to provide reliable
supply.

Economists have studied this in relation to transmission pricing. According to

Sioshansi (Sioshansi and Morgan, 1999) there are two main considerations
affecting transmission pricing:
• Usage pricing – to compensate owners of the grid for the right to use
their network, and to encourage the building of more network; and
• Congestion pricing – charging a premium for using the facility during
times of peak demand. Ideally this premium should encourage some
users to shift their usage to other times.
Some economists and utility representatives in the USA have recently argued
that because transmission costs represent only a small fraction of total
electricity prices to consumers, and the impact of transmission constraints on
prices in a competitive market is large, transmission investment to alleviate
congestion should be encouraged through appropriate regulation (Hirst,
2000, Butler, 2000, Nasser, 1999). In Australia there is a proposal to provide
revenue incentives to transmission companies to reduce transmission

49
constraints that impact on the market (ACCC, 2002b). It is not yet clear how
this will be implemented.

There is a school of thought that says that transmission prices should provide
a locational signal to customers (generators and consumers) to encourage
them to use the network in a way that is efficient in terms of system losses
and use of transmission assets (Hogan, 1998, Camfield and Schuster, 2000).
Efficient use of transmission could potentially defer the building of more
transmission infrastructure. Elements of locational pricing are incorporated
into Australia’s zonal pricing scheme and the Pennsylvania- New Jersey –
Maryland nodal pricing scheme (Sioshansi and Morgan, 1999). However,
because transmission constitutes only a small fraction of total electricity
pricing it is unlikely that these signals are strong enough to have much
impact on the siting of major generators or loads. Nevertheless transmission
pricing signals may illuminate the need for transmission enhancements
(Kirby, 1999).

The third impact of regulation is on financial risk associated with the

transmission business. This has a direct impact on the ability of the
transmission enterprise to raise capital for investments. For instance,
Standard and Poor’s list
• Tariff-setting mechanism,
• Regulatory efficiency and investment requirements, and
• Transparency of regulatory arrangements
among the main factors distinguishing one transmission business from
another for the purpose of credit rating (Flintoff, 2001).

Transparency of regulatory arrangement is one aspect of regulatory

governance. Regulatory governance concerns (Stern and Holder, 1998)
• The objectives of infrastructure regulation and
• The specific institutional framework for regulation in any country.

50
Poor regulatory governance can expose transmission network owners to
considerable risk (Stern and Holder, 1998). In particular:
• Transmission businesses are highly capital intensive, and the assets once
installed are sunk assets. Hence investors are exposed to risk of adverse
policies set by governments eg by failing to allow maintenance of income,
or reneging on explicit or implicit contracts.
• Electricity services are vitally important to the welfare of all households,
and also to industrial output; this means that prices to customers are
highly political. Price changes (eg caused by abolition of cross-subsidies
or even recovery of full costs) can make a large impact on household and
industrial costs. Hence there is a temptation for governments to interfere
with the regulation of electric utilities.

Consideration of these factors can be avoided if the state owns the utility. In
such cases the functions of policy making, ownership and management and
regulation can become blurred, and end-user pricing tends to become highly
politicised. Under these circumstances there is often little incentive for
utilities to operate efficiently. Examples from Stern include Indian State
Electricity Boards, which are chronically loss-making, and Indonesian and
Malaysian energy companies, that are earn a low real rate of return on
assets and have pervasive cross-subsidies.

Stern argues that when it is desired to introduce private investment into an

infrastructure industry it is necessary for policy making, ownership and
management, and regulation to be clearly delineated. In particular,
regulation must be, and be seen to be, independent of government
interference in order to convince potential investors that the risk level is low.

The case of Ukraine (Lovei, 1998) illustrates the importance of regulatory

governance to the health of the electricity industry. After successful
restructuring of the electricity industry, the average electricity price rose by a
factor of three (from a level initially well below cost) from 1994-1996. This
caused a growing problem with non-payment of electricity accounts.
Electricity suppliers were pressured by central and local governments to

51
continue supplying electricity to strategically or politically important
customers. In 1996 the government also instructed the National Electricity
Regulatory Commission to leave retail prices unchanged until further notice.
Accordingly the regulatory commission instructed the National Dispatch
Centre to apply downward corrections to the daily average marginal price,
contrary to market rules. A proliferation of barter and other non-cash
payments further undermined the application of market rules, as the National
Dispatch Centre could only collect and allocate cash payments.

The adjustments to market rules, and tolerance of non-payment, government

interference in market pricing and implicit preference to non-cash payments
deterred lending investors. The European Bank for Reconstruction and
Development cancelled a US$60 million loan and the World Bank suspended
disbursement of a US$314 million loan (Lovei, 1998).

Under some tariff setting scenarios there is a significant element of risk

associated with asset valuation. A study of European Union transmission
pricing found that asset valuation was a major source of difference between
regulation in different countries with impact on tariffs and hence revenues
(Perez-Arriaga et al., 2002). Unfortunately the paper did not examine the
different asset valuation methods employed, but the implication is that there
is a wide variety of approaches to asset valuation, even within Western
Europe. In Australia a large component of transmission revenue is based on
Depreciated Optimised Replacement Costs (ACCC, 1999a). In New Zealand
the asset valuation is based on Optimised Replacement Costs. In both
cases the “Optimisation” term refers to the fact that valuation is based on an
ideal system, rather than the one that developed over time. Under-utilised
equipment may be subject to reduced valuation or being “optimised” out of
the total. When this transpires the transmission company cannot recover
capital nor gain a return on its investment in this plant. This risk is faced
when the equipment is first installed on the network – only investments
deemed by the regulator to be “prudent” are counted towards asset valuation,
and throughout the life of the equipment.

52
There is much less research published about the impact of regulation from
the perspective of the transmission company. Evidence of the impact of
deregulation and restructuring on AM issues can sometimes be found in
descriptions of strategies employed by different companies. Certain common
threads can be observed in these documents:
• In the past ten years there has been strong downward pressure on
operation and maintenance costs (Davies et al., 1998, Jefferies,
1997). This has occurred not only in enterprises that have been
subjected to extensive restructuring and re-regulation, but also where
the regulation and structure of the industry is largely the same as it
was in 1990 (See for example (Percebois and Wright, 2001),
discussing Electricité de France (EdF)). In the USA in the period 1990-
96 maintenance spending in transmission declined by 22% (Kirby,
1999), even though restructuring was only in its infancy in the US
during this period.
• For transmission entities, particularly those formerly part of vertically
integrated utilities, there has been an increase in uncertainty in
planning, and lead times for installation of new plant have been
reduced (Powerlink Queensland, 2000) (Urwin, 1999).
• The pressure to improve efficiency in operations and maintenance,
together with increased uncertainties in planning network
augmentation have focussed attention on risk management (Jay and
Williams, 2001, Australasian CIGRE Asset Management WG, 2001)
within transmission organisations to a greater extent than ever before.
Jones et al (Jones et al., 2000) suggest that for Australia and New Zealand
the regulatory drivers towards higher availability and reliability but lower
operations and maintenance cost, encourage network owners to consider
new technologies. It is also argued that the high capital component (80%) of
revenue could be considered to encourage a strategy of capital investment,
especially if this also reduces operations and maintenance costs. This
results (Jones et al., 2000) in inclusion of the following criteria for the
selection of new technologies:
• Minimal inherent safety risk;

53
 • High self-monitoring and remote interrogation capability;
• Low maintenance requirements; and
• Integration with other substation systems.

Some useful research has come about as a result of these drivers, especially
from industry-based organisations. For example in (CIGRE WG B2.13,
2002) the authors draw on decision science methods to examine options to
replace or upgrade a transmission line, in the light of uncertain generation
developments. The paper advises the analysis of threats and opportunities
from external sources, in addition to assessment of the company’s own
strengths and weaknesses in order to assess different options. It is an
excellent example of the application of strategic thinking to asset
management.

Another strategic asset management topic, arising from efforts of utilities to

reduce operating costs, is that of maintenance outsourcing. There is a
potential trade-off between skills retention within the transmission business
and minimising maintenance and operating costs (Australasian CIGRE Asset
Management WG, 2001). This problem is also being addressed by CIGRE
Working Group 23/39.14 (CIGRE JWG 23/39.14, 2002): this group is
developing guidelines to assist utilities to identify the optimal level of
outsourcing, dependent on their individual circumstances. The authors of
this paper propose the use of a competency mapping model, and a risk
versus effectiveness model to provide an understanding of what services
should potentially be outsourced. Five competency classes are identified:
• Distinctive – the most important capability of an organisation (For
an asset manager type utility this might be ability to manage
contracts.);
• Essential – Necessary for the organisation to operate;
• Spill-over – Allows a utility to obtain profits in a related activity;
• Protective – Related to activities that cause considerable risk to the
organisation (eg protection maintenance); and

54
 • Parasitic – Activities that waste organisation resources (eg tower
painting).
Three overriding issues are identified for the analysis of outsourcing
opportunities: risk, cost and quality/effectiveness.

2.2.2 Assessment and management of market impacts

Under the new competitive arrangements adopted in many countries the role
of the transmission provider is (usually) to own, operate and maintain the
transmission network. In some cases (such as the National Grid Company)
the role also includes system operator. The generators and retailers trade in
a market for electricity, and although the rules for the market arrangements
vary significantly from one jurisdiction to another, a key feature of electricity
markets is open access to the electricity network. Market operation is based
on bidding and contracts rather than the cost of operation of generation plant.
Hence compared with economic dispatch scenarios, in market operation
there is typically a much higher degree of variability in generation outputs and
in the flow of power across the network.

In (Jones et al., 2000) the impact on transmission asset management of

deregulation and the market in Australia and New Zealand is described with
particular reference to switchgear. The new risks and opportunities for
transmission businesses identified include:
• New liabilities arising from potential impacts on the operations of other
market participants;
• Risk of stranded costs, the inability to obtain a return on or of capital
on under-utilised assets is increased. (Changes in utilisation may
arise from market operations, such as changes in generation
patterns.)
• Greater difficulties in obtaining outages for maintenance are
experienced: this drives changes in maintenance activities, such as
reduction in outage duration, use of live maintenance work, and
transferring of work to periods of low load.

55
 • Planning uncertainties: the lack of central planning of generation
development means that lead times for transmission planning are
reduced.
• Operation of the network in ways not intended by original designs:
Changing transfer patterns, for example, can cause large swings in
reactive support requirements and consequently higher numbers of
switching operations on reactive plant, with the result of greater wear
on circuit breakers.
• Increased opportunity for embedded generators: This could reduce
use of parts of the transmission network, increasing the potential for
stranded costs.
• Competition with other asset owners for development of contestable
assets could impact on profitability and
• Competition between regulated and unregulated asset developments
could change utilisation of plant. For example, in Australia
unregulated interconnectors may be built and can operate in direct
competition with regulated interconnectors owned by transmission
network service providers.
In the UK, changes to generation patterns, and short lead times for projects
have led to installation of significant reactive plant and the development of
relocatable static VAR compensators (Urwin, 1999). The use of relocatable
equipment reduces the risk that plant will become redundant in the future.

Laycock et al (Laycock et al., 1998) argue that de-regulation and market

operation have put higher stresses on the performance of protection systems
in transmission networks. They identify reduction of operational constraints
and improvement of voltage quality as drivers for transmission utilities. They
suggest that the use of numeric protection relays can assist in resolving
these problems, and can provide valuable data for the analysis of power
quality problems. They also propose the use of intelligent analysis tools for
the purpose of analysing data from the numeric protection relays. However,
no quantifications of benefits are identified; the article is largely an
advertisement for research projects being undertaken.

56
2.2.3 Interconnection of networks
Another major trend in electricity industries world-wide has been the
interconnection of electricity networks (Hammons et al., 1998, CIER et al.,
2001). There are significant benefits for interconnection of networks:
• Interconnection promotes more efficient use of generation capacity by
using structural differences in load profiles (for instance time of peak
load differences in different time zones).
• Reserve capacity required to cope with loss of generation (or
transmission outages) is shared between interconnected systems,
thus reducing the level of reserve required on individual systems.
• In an electricity market the interconnection of networks potentially
increases competition between generators, and can lead to reductions
in average price of electricity and price volatility.
For transmission network service providers interconnections pose a number
of challenges:
• The differences between actual and design network power flows can
be exacerbated when previously separate well-developed networks
are interconnected.
• When networks are interconnected operating and reliability standards
have to be harmonised (CIER et al., 2001). This may require
upgrading of equipment.
• Interconnection of networks can result in new stability limits (dynamic
or transient stability) which may impact on the operation of the
network, or may mean a requirement to install new control equipment
such as FACTS devices or Static VAR Compensators (SVCs).
• Despite potential benefits it may be difficult to build interconnections,
for reasons including the need to deal with multiple jurisdictions,
difficulty justifying the costs in terms of benefit (this has been the
Australian experience) and public opposition to the building of
transmission lines.

2.2.4 Embedded generation

In the past ten years there has also been a trend toward the location, within
distribution networks, of generation using renewable energy sources or co-

57
generation projects. This has arisen from acceptance of “Global Warming”
as a phenomenon of human origin, recognition of the need to reduce the
emission of “Greenhouse Gases”, and also an effort to decrease the use of
non-renewable resources for energy generation.

The use of embedded generation appears to have been adopted most

enthusiastically in parts of Western Europe (Leonhard and Muller, 2002),
where sympathetic regulation and subsidies have encouraged the
development of wind generation and co-generation projects, but government
initiatives have also been made in this direction in other parts (eg
Queensland government requires electricity retailers to source 15% of their
electricity from gas-fired or renewable generation from 1 January 2005
(Queensland Government, 2000)). In Germany about 10 000MW of wind
generating capacity is now installed with many more giga-watts of power
planned from offshore wind-farms in the North and Baltic Seas(Leonhard and
Muller, 2002). In Denmark the installed capacity of wind and combined heat
and power projects together is larger than the maximum demand (Bak-
Jensen, 1999). The variability and lack of output control associated with
these embedded generation plants add another degree of uncertainty for
transmission reliability assessment and planning. There is concern in the
Scandinavian system that fluctuations from heavy wind-power in-feeds and
the local heat-and-power plants in Denmark might necessitate strengthening
of the high voltage grid (Leonhard and Muller, 2002).

In other parts of the world there has been concern expressed that embedded
generation may reduce transmission system utilisation with possible
repercussions for revenue. In (CIGRE WG B2.13, 2002) the Working Group
identifies “spreading distributed generation” and “use of renewable energy
resources” as two of the threats/opportunities that need to be considered
when planning transmission augmentations.

2.2.5 Risk management in transmission organisations

Risk management is defined as the culture, processes and structures that
are directed towards the effective management of potential opportunities and

58
adverse effects (Standards Australia, 1999a). Considerations of risk
management are pervasive in many of the Strategic Asset Management
topics discussed so far. Risk management entails analysing and controlling
risks to which an organisation is exposed. Within operations and
maintenance areas the focus on risk management has translated into
changes in maintenance strategies
• Greater use of Reliability-Centred Maintenance (RCM)(Sherard, 2001,
Basille et al., 1995, Balzer et al., 2000, Allison and Jay, 2001, Lee,
1997) (See also section 2.3.1);
• Importance (criticality) of equipment on the network being taken into
account for operations and maintenance strategies (Davies et al.,
1998, Stewart, 2001);
• A trend towards the use of predictive maintenance strategies, and
away from more expensive, but low risk, time-based replace-overhaul
strategies (Davies et al., 1998).
The efficiency and risk management focuses have been drivers for some
useful projects and developments including some drawing on operations
research techniques. Interesting developments in this area will be discussed
in more detail within the context of Process Level Asset Management
(Section 2.3).

Risk is a function of likelihood and the consequence of an incident. For

transmission companies several types of risk are present:
• Regulatory risk may arise from poor regulatory governance and
regulatory practices (see Section 2.2.1.3).
• Market-related risk for transmission mainly relates to changed
operation of assets compared with their original intended use, whether
this is over-utilisation or under-utilisation, including the risk of asset
stranding (Sections 2.2.2, 2.2.3 and 2.2.4).
• Financial risk includes currency-related risk (when borrowing is in a
currency other than the local one), and risk of loss of revenue as a
result of external factors (such as weather-related factors or economic
growth). Factors such as these are frequently considered for credit

59
 rating determination (eg (Standard & Poor's, 1999)), which in turn
impacts on the cost of capital.
• System risk relates to actions taken within the power system that
increase the probability of not being able to fulfil a specific
transmission service (Damstra et al., 2000). This includes risks
associated with transmission constraints or outages.
The CIGRE Joint Working Group 23/39-07 (Damstra et al., 2000) categorises
system risk as three types:
• Type 1 System risk associated with outages for maintenance
activities;
• Type 2 System risk through delaying maintenance resulting in
unknown component deterioration; and
• Type 3 System risks arising from not taking an outage and operating
the system with known defective or depleted plant or equipment.
The paper suggests that Type 1 risks are influenced by system topology, age
and state of equipment, generation priorities, difficulties in short-term outage
planning, need to reduce outage times and use of weekend working.
Type 2 risks are influenced by the effect on equipment lifetime of use of
online condition monitoring and new technology to reduce equipment
maintenance and the number of opportunity outages (through plant failure).

Conceptually Type 1 risk decreases with maintenance interval and Type 2

risk increases, so that it is theoretically possible to minimise risk by
optimising the maintenance interval.

The paper recommends the use of RCM techniques to identify and quantify
risk. The use of commercially-available Decision Support Tools for
maintenance optimisation is also suggested. Bundling maintenance (doing
as much maintenance as possible in the same circuit outage) is identified as
an option that reduces Type 1 risk without increasing Type 2 risk.

Type 3 risks and certain Type 1 risks (including risk of tripping circuits) are
typically managed by operational procedures.

60
2.2.6 Level of performance
Level of performance issues are starting to assume a high level of
importance to transmission businesses. There are a number of aspects to
this:
In general there is an impression among transmission companies of
increasing customer expectations for quality of supply. For de-integrated
entities formerly part of a vertically integrated industry structure, increased
expectations may be accompanied by a greater exposure to litigation for
supply loss than previously, especially from generators. For instance Jones
(Jones et al., 2000) says “the major impact on service levels has been the
increased expectations of some market participant and the increased liability
associated with operating in a market environment”.

Different customers have different requirements for quality of supply (Sullivan

et al., 1996). Cost to customers also differs for unplanned and planned
outages (where notice is given). In some market arrangements the customer
can determine the level of service he requires and may choose a lower
service level for a reduced cost (eg New Zealand (Jones et al., 2000)). In
(CIGRE SC 23 and 39, 2000) some 27% of respondents offered a tariff
option permitting limited supply operations, and 6% offered a tariff penalising
supply interruptions. In some utilities, mainly those with distribution as well as
transmission functions, profiles of customers are developed to identify
reliability needs at various supply points. An example of this is a probabilistic
planning tool used by Duke Power called Value-Based-Reliability-Planning
(Dalton et al., 1996). Underlying this program is a realisation that there is a
cost associated with improving reliability and quality of supply, (capital and
O&M), and a cost associated with customer outages. The tool aims to
minimise the combined costs of these in order to provide acceptable
reliability to customers at minimum overall cost.

There is a growing trend for regulators to set targets for reliability and quality
of supply levels, and for incentives/penalties to apply. It is therefore
necessary to balance potential incentives/penalties against the costs
associated with achieving the regulatory targets.

61
2.2.7 Performance measurement
Performance measurement is of great interest to regulators, particularly
comparison of performance between different utilities. A survey of
benchmarking techniques and studies was made by Jasamb and Pollit
(Jasamb and Pollitt, 2001). The authors report few benchmarking exercises
in transmission. Four countries report use of international benchmarking of
electricity transmission companies: Columbia uses Data Envelopment
Analysis (DEA) to compare construction costs; the Norwegian regulator
compares its regional transmission networks with those of Sweden using
DEA to set firm-specific efficiency requirements and revenue caps; the
Netherlands plans to use DEA for setting transmission tariffs using yardstick
regulation; for the National Grid Company comparison has reportedly been
made with 40 utilities using DEA and Total Factor Productivity (TFP).

Comparison of transmission performance is complicated by a lack of

consistent performance measures, and other data (especially operating and
capital costs (Jasamb and Pollitt, 2001)) between different companies. (See
Section 5.2.2 for a detailed discussion of performance measures). This
problem is partially addressed by the use of third parties for benchmarking
studies. Companies contribute data to these studies and in return are given
information on their ranking relative to other participants on various
performance (or partial performance) indicators. Comparison is further
hampered by differences in conditions pertaining in different countries –
benchmarking works best when similar utilities are compared.

2.3 Process Level Asset Management

At the process level of asset management the emphasis is on processes and
statistical understanding of the equipment. Process-level asset management
issues tend to have narrower scope than those at the strategic level: they
tend to focus on one aspect of asset management such as maintenance
efficiency, planning for system reliability or equipment replacement
strategies. The current directions of process level asset management
research are being driven by the needs of the business and the issues
identified at the strategic level of asset management. In particular:

62
 • Maintenance cost drivers and risk management are fostering research
towards maintenance optimisation.
• Cost and reliability drivers are driving increased use of RCM and a
trend toward the use of condition-based maintenance.
• Risk management, risks associated with the regulatory system and
low return on capital are encouraging investigations into replacement
policies, and encourage strategies that take into account the criticality
and condition of equipment when considering maintenance or
replacement.
• Life cycle costing is being used to justify the expenditure of extra
capital in order to reduce operation and maintenance costs. The focus
on reducing maintenance is in line with regulatory drivers.
• Reliability drivers have fostered research into probabilistic reliability
modelling, and investigations into the effects of different configurations
of equipment on reliability.
• Market drivers and in some cases regulatory drivers, (through
performance standards for example) are placing pressure on utilities
to develop strategies to increase availability, leading to changes in
maintenance strategies.

2.3.1 RCM and its influence on asset management in transmission

Reliability centred maintenance (RCM) is a process that originated in the
aviation industry and was widely publicised by John Moubray (Moubray,
1992). Moubray defines RCM as a “process used to determine the
maintenance requirements of any physical asset in its operating context.”
The RCM process involves the following steps:
• Define the system, its functions and required performance standards;
• Determine the functional failures (failures to achieve required
performance);
• Identify the failure modes associated with functional failures (root
causes of failures);
• Describe the effect of each failure mode;
• Identify the consequences of the failure mode;

63
 • Identify appropriate actions to prevent each failure mode (Tasks must
be both technically feasible and worth doing – a function of
consequence and cost);
• Decide what to do if a preventive task cannot be found to address a
failure mode. (This may involve rebuilding or discarding the item
periodically, redesigning the item or doing nothing, depending on the
consequences of failure).

The first four steps are also known as Failure Modes Effects Analysis
(FMEA), and the first five steps, including consequence, are referred to as
FMECA.

The concepts involved in FMEA and RCM have become pervasive in asset
management, even though wide-scale adoption of the RCM method has not
occurred in distribution and transmission sectors of the electricity industry
(ECAR Electric Equipment Panel, 1999, Endrenyi et al., 2001). The main
reason for this appears to be the resources required (financial and personnel)
for full implementation of RCM. As a result there are now variants and
computer tools to streamline the RCM process available on the market.

One variant of RCM has recently been employed by the National Grid
Company of the UK. The strategy known as Rule Based Schedule
Evaluation (RBSE) (Allison and Jay, 2001) involves matching current
maintenance schedules with known failure models and deterioration rates in
order to eliminate unnecessary, ineffective or excessive maintenance. RBSE
can also flag maintenance work that can be undertaken without an outage. It
is expected that this process will reduce data collection and condition data
analysis, and the amount of work done during outages. It is also anticipated
to reduce protection system maintenance by up to 95%. A similar technique
was recently described in detail in (Anderson, 2002). These approaches are
likely to be effective only if current maintenance practices address all
important failure modes and if excess maintenance is being undertaken.

64
2.3.2 Condition based maintenance (CBM)
While time-based maintenance (or more precisely scheduled
repair/replace/overhaul task-based maintenance (Hastings, 2000)) remains
the most common form of maintenance process in the transmission sector
(CIGRE SC 23 and 39, 2000), condition based maintenance (CBM) or as
Moubray (Moubray, 1992) describes it, the “Scheduled On-condition Task” is
becoming increasingly important in the industry. Condition-based
maintenance comprises condition monitoring, assessment and maintenance
actions (including refurbishment and replacement), based on condition.
Condition assessment can potentially provide the following benefits (Urwin,
1999) when integrated in an overall asset management strategy:
• Improved risk management of assets;
• Improved management of environmental risks;
• Ratings and performance improvements leading to avoidance of
constraints (and associated costs); and
• Better targeting of assets for refurbishment and replacement.

CBM has two main purposes, either

• To prevent a functional failure; or
• To mitigate the consequences of that failure.
From the transmission perspective, CBM, properly applied, has the potential
to increase availability of equipment, reduce the incidence of major failures,
and reduce routine (ECAR Electric Equipment Panel, 1999) and remedial
maintenance costs. In order for CBM to work, it is necessary to be able to
predict the onset of failure early enough to rectify the problem or avoid the
consequences of failure. Interest in condition-based maintenance has
spawned much of the current equipment-level research on failure modes and
condition monitoring.

2.3.3 Justification and selection of condition monitoring

At the process level, considerable effort has been put into economic
justification for CBM. CBM, and particularly online CBM, often requires a
substantial capital outlay. The cost penalty is often highest for retrofitting of
on-line CBM to older equipment, even though benefit is likely to be realised

65
sooner when CBM is applied to older equipment than when applied to new
equipment (CIGRE WG 13-09, 2000). In (IEEE, 2000, Bergman, 2001,
Bergman, 1999) Bergman and the IEEE Guide (IEEE, 2000) advocate use of
RCM to identify economic condition monitoring opportunities. Likewise, the
CIGRE guide (CIGRE WG 13-09, 2000) identifies failure modes, effects,
condition monitoring techniques to address these failure modes, and
suggests methods for assessing the potential benefits of CBM applied to
switchgear.

2.3.4 Maintenance period optimisation

The concept of maintenance period optimisation arises from the realisation
that routine maintenance costs are reduced as maintenance period
increases, but remedial costs increase with increased maintenance period.
The combination of these cost curves suggests a roughly bath-tub shaped
curve with some minimum cost at the point of intersection of the curves. The
actual point of minimum depends on the failure period of the dominant failure
mode. Calculation of this point is complicated by multiple failure modes, and
also by failures that depend on the time since last operation (eg “sticky”
mechanical relays). Maintenance optimisation has been extensively
researched in operational research areas since the 1960s, with many
hundreds of applications reported eg (Valdez Flores and Feldman, 1989,
Scarf, 1997, Wang, 2002) but the modelling techniques do not appear to
have been widely applied in the electricity transmission sector. Many
reasons for the lack of penetration of optimisation techniques in maintenance
management in general are suggested by Dekker (Dekker, 1996) including:
• Many models have been developed out of interest in the mathematics
more than its application.
• Many of the case studies are tailored to particular applications and not
easily translated.
• Obtaining data to support the modelling is often difficult and time
consuming.
• A high level of expertise is often required to run these applications
successfully.

66
 • In some cases the potential benefits do not justify the expense of
modelling.

However, software to facilitate maintenance optimisation has recently

become commercially available, and is starting to find application in the
electricity industry (for instance Yorkshire Electricity and Gas have
successfully applied an inspection optimisation module from the MACRO
suite (Bower, 2001), and the National Grid Company has made use of the
maintenance optimisation module from the same package (Allison and Jay,
2001)). The development of general Decision Support Systems (DSS) is
facilitating this process. There are two main types of DSS: Operational and
Strategic (Dekker and Scarf, 1998). The Operational DSS is developed to
solve a repetitive problem; it is generally associated with a database of
relevant information. Examples of applications include road maintenance,
and a decision support system for opportunity maintenance of gas turbines
on offshore oil rigs (Dekker and Scarf, 1998). The Strategic DSS is designed
to analyse one-off problems at a higher level (systems or subsystems).
Examples cited in (Dekker, 1996) include KMOSS from KEMA, RELCODE,
OPTIMON, and MAINOPT (precursor to MACRO). A recent applications
targeted specifically at the electricity industry is RiBAM (Risk based Asset
Management from Kinetrics (formerly Ontario Hydro Technologies)) (Anders
et al., 2001).

A simpler approach to optimising maintenance intervals is used by NV Sep of

the Netherlands (Davies et al., 1998). Maintenance is initially performed
according with manufacturers’ recommendations. Once maintenance is
completed on a piece of equipment, maintenance staff assign to it one of
three categories, A, B or C:
• ‘A’ means the maintenance was unnecessary; it could have been done
later;
• ‘B’ indicates that the maintenance was performed at the right time;
only normal maintenance was required and
• ‘C’ means the maintenance should have been done earlier; major
faults were found.

67
Annually these simple results are analysed for each type of equipment, also
taking into account the manufacturer and age of the equipment. For
equipment that scored mainly ‘A’s the maintenance interval is increased. For
equipment that scored mainly ‘B’s no change is made. For each ‘C’ score an
immediate investigation is made to determine if other equipment of the same
type has the same problem. A decision is made on whether to decrease the
maintenance interval, avoid buying equipment of the same type or let the
manufacturer prove the problem will not recur.

This strategy (Davies et al., 1998) has resulted in increased maintenance

intervals in most bays, phasing out of air blast circuit breakers and early
discovery of major problems in surge arrestors and insulators of particular
types. It provides a simple but effective feedback mechanism from
maintenance to policy, providing maintenance staff with a direct role in policy
development.

2.3.5 Strategies to increase circuit availability

Optimising maintenance intervals is one strategy to increase availability of
equipment. Another strategy specifically targeted at circuit availability is to
reduce the duration and number of times a circuit is taken out of service for
maintenance. To this end NGC has developed a genetic algorithm based
spreadsheet model called COMPAC, which groups together those
maintenance activities on a circuit which can be performed concurrently,
potentially reducing unavailability on key circuits by up to 30% (Urwin, 1999)

Other strategies being employed include use of live-line and live-substation

maintenance techniques. In particular, helicopters are now being employed
for live line inspection and maintenance by a number of utilities (Ostendorp,
2000, Marshall and Visser, 1998, de Parny, 1993, Soler, 1993, Pelacchi,
1998)

In some cases the driver is not so much to increase circuit availability as to

reduce the risk of supply outage in the event that a circuit fault occurs while

68
maintenance is undertaken on another circuit supporting the same part of the
transmission network, or alternatively to reduce constraints on network
operation that occur during maintenance of circuits. In these cases
maintenance may be scheduled for periods of lower load, such as weekends
or over-night or during periods of milder weather (CIGRE SC 23 and 39,
2000).

2.3.6 Criticality in transmission asset management

Within a network there will be plant that is critical to the performance of the
whole, the absence of which can cause constraints on network operation.
There may also be customers who require higher quality of supply than
others – manufacturers of integrated circuits, for example. There is a trend
for these critical parts of the network to be assigned a greater priority in order
to provide a satisfactory electricity supply to all customers at a reasonable
price. Another trend is towards regulators making transmission companies
pay for the cost of constraints they cause to network power flows. The
management of critical plant is being addressed in different ways by different
utilities. Electricité de France address customer expectations by grading
maintenance on substation bays according to the quality of supply required
by the client (Davies et al., 1998).

At Scottish Power the concept of criticality has been applied to all aspects of
asset management (Stewart, 2001). In this strategy the commitment of
resources depends on the criticality rating assigned to the equipment. In the
area of inspection for example, equipment rated low criticality would receive
inspections to meet statutory requirements, equipment rated medium
criticality would have targeted information collected and only highly critical
equipment would be considered for on-line condition monitoring.

A similar concept, but with different method of application has been

developed in Germany by Damstadt University of Technology in conjunction
with ABB Calor Emag Schaltanlagen AG and in collaboration with German
utility Energie-Versorgun Swaben AG (Balzer et al., 1998). In this
application, equipment is assigned an importance rating on a scale of 1 to

69
100 and a condition rating, also in the range 1 to 100, which may be
constructed from a number of different parameters. Importance and
condition pairs of coordinates are plotted on a graph. A high value of
importance means that the equipment has a major influence on the network.
When an axis 45 degrees to the importance and condition axes is drawn, the
priority for maintenance or replacement is determined from the distance of
the (Importance, Condition) coordinates to this new axis. The greater the
distance, the higher is the priority given to this equipment.

2.3.7 Reliability modelling

The North American Electricity Reliability Council (NERC) describes reliability
in these terms(Reliability Criteria Subcommittee of the Engineering
Committee, 1985):
Reliability in bulk power electric systems is the degree to
which the performance of the elements of that system results
in power being delivered to consumers within accepted
standards and in the amount desired. Bulk power electric
system reliability can be addressed by considering …
adequacy and security. Adequacy is the ability of the bulk
electric system to supply the aggregate electric power and
energy requirements of the consumers at all times taking into
account scheduled and unscheduled outages of system
components. Security is the ability of the bulk power electric
system to withstand sudden disturbances such as electric
short circuits or unanticipated loss of system components.

From this definition it can be seen that for transmission systems, the design
of the network plays a major role in providing for both security and adequacy.
In transmission systems reliability is built into the network by the use of
redundancy. Traditionally an N-1 planning criterion was used, and this is still
the most common method of ensuring reliability today (CIGRE SC 23 and 39,
2000). An N-1 criterion means that the system must be able to operate
without loss of supply when any single contingency occurs. Examples of
single contingencies can include a line fault or an abnormal weather
condition.

70
Alternatively it is possible to model the reliability of the network using
probabilistic methods. The main methodologies employed for the
assessment of network reliability are
• Monte Carlo simulations, and
• Markov models.
Modelling has been applied to serve a number of different purposes:
• A number of applications have been developed to examine network
planning options to improve reliability (Bak-Jensen, 1999, Dialynas et
al., 1996). EPRI has developed a Probabilistic Reliability/Risk
Assessment tool for transmission systems (Lee and Hoffman, 2001).
This tool calculates risk indices for overloading, under-voltage,
transient and dynamic stability conditions. It may have application for
planning or as a real-time tool for system operation security
assessment. Other applications have focussed on the distribution
system (Brown et al., 1996, Goel, 2000).
• Applications have been developed to explore reliability within various
subsystems of the transmission network, such as protection, and
substation configuration (Atanackovic et al., 1999, Karlsson et al.,
1997, Meeuwsen and Kling, 1997b, Meeuwsen and Kling, 1997a,
Vermeulen et al., 1998).
• Reliability assessment including condition deterioration in equipment
has been the subject of a number of research endeavours (Hoskins et
al., 1997, Hoskins et al., 1999, Endrenyi and Wellssow, 2001,
Endrenyi et al., 2001).

2.3.8 Life cycle costing

Life cycle costing (LCC) is the process of assessing the cost of a product
over its life cycle. Life cycle costs include all costs associated with
acquisition and ownership of a product over its life. Australian standard
AS/NZS 4536 (Standards Australia, 1999b) defines the following objectives
for life cycle costing:
• Calculate a dollar value representing the LCC of a product as an input
to a decision making or evaluation process.

71
 • Support management considerations affecting decisions during any
life-cycle phase.
• Identify attributes of the product which significantly influence the LCC
of the product so that they can be properly managed.
Considerations include the value of cost elements, timing of cash flows and
time value of money. The process should be an on-going one over the life of
the product – actual costs should be fed back into the evaluation process to
improve cost estimates and identify improvements.

Literature suggests that life cycle costing is gradually becoming an integral

part of electricity transmission asset management (Allison et al., 1995, Amato
et al., 1999, Dwivedi et al., 1988, Flatt, 2000, Hastings et al., 1999, Janssen
et al., 1993, Karlsson et al., 1997, Nordin et al., 1995, Pettersson et al., 1986,
Pryor et al., 1998). Results from a CIGRE survey of maintenance policies
and trends (CIGRE SC 23 and 39, 2000) suggests a trend towards the use of
life cycle costs as a key performance indicator (KPI).

In (Anders et al., 2001) a program is described that uses Markov chains to

investigate the effect of different maintenance and replacement options and
timings on the life and value of assets (assuming that the value is related to
the condition of assets – a debatable point, since book value seldom follows
condition). The paper does not refer to the concept of life-cycle costs but
describes converting total costs to Net Present Value (including capital and
maintenance), is easily translated to life-cycle costing.

2.3.9 Criteria for replacement of equipment

Questions of when and on the basis of what criteria to replace equipment are
of considerable interest to transmission asset managers. Interest in these
topics have been fostered by economic pressures to extend the life of
equipment, coupled with the recognition that in developed networks (Europe,
North America, and Australia), the average age of equipment is steadily
increasing (CIGRE WG 37.27, 2000). Concerns have been raised about the
potential cost burden associated with replacing large portions of the network
within a short timeframe (CIGRE AP13, 2001). Most transmission entities

72
have developed a long term plan for equipment replacement (for example,
some 92% of responses from the CIGRE study). This study identified that
the most important criteria influencing the system-wide replacement plans of
transmission businesses were:
• Deterioration of transmission system performance
• Load growth
• Availability of capital (particularly among respondents from North
America, Eastern Europe and Australasia).
Other criteria cited include changing transmission design criteria, changing
customer expectations and regulatory decisions. This list is curious in that
there is no mention of economic end of life, condition, criticality or age as
factors contributing to the transmission replacement plan. Deterioration of
system performance (as defined in this survey question) includes reliability,
power quality, safety and customer interruptions. These parameters are
lagging measures of risk, and it is strange that they would be used as criteria
influencing the long-term replacement plan. Perhaps the survey question
design is at fault for not offering other choices of criteria.

The survey additionally asks respondents to select from a list of options

important criteria for replacement of a specific piece of equipment. Results
show that
• Cables and switchgear are most often replaced because of frequency
or number of failures.
• Transformers and towers are mainly replaced because of asset
condition.
• For overhead lines public safety is the predominant reason for
replacement.
• In protection and control equipment replacement is influenced by the
availability of spare parts.
As in the long-term replacement criteria, economic reasons for replacing
equipment do not rate highly in this list. Examination suggests that in all
categories, except for protection and control equipment for which

73
 technical obsolescence is important, risk is the dominant reason for
replacement.
This explains why the myriad of capital replacement and economic life
models found in operational research have not been extensively adopted
by transmission enterprises.
A few applications including elements of risk and having an electricity
industry and focus have recently been published. In (Hastings et al., 1999)
risk cost is included in a techno-economic analysis, using equivalent annual
cost and repair limit theory.

2.4 Equipment Level Asset Management

Research at the equipment level can also be seen to support the needs of
the process and strategic levels of asset management.
Common themes throughout all types of equipment include:
• Determining the modes of failure;
• Deciding which failure modes are most important;
• Calculating failure rates;
• Development of various types of condition monitoring to detect onset
of failure;
• Development of diagnostics to interpret data from condition
monitoring;
• Use of condition data to predict remaining life.
The themes identified here, in general, support the information requirements
for process and strategy level AM analysis. For example, information on
failure modes and rates is useful for RCM and remaining life estimates
support replacement decisions.

2.5 Identifying the need for research

In this chapter a review of current research in asset management has been
made, illustrating the range of issues and topics discussed in research
literature at the strategic level of asset management and showing how these
relate to research at the process- and equipment-levels of asset
management. The review suggests many common themes among asset
management drivers for transmission businesses. Despite this, as the
CIGRE study of maintenance policies and trends shows, there are significant

74
differences in asset management practices world-wide. Perhaps this is a
function of the level to which different drivers operate on different
organisations. For example in a discussion paper (Australasian CIGRE
Asset Management WG, 2001) on asset management in Australia and New
Zealand the following statement:
Business processes established within each electricity utility
to deal with asset management are based on a number of
factors specific to that utility. These factors, and their impact,
vary from utility to utility and consequently result in diverse
approaches to asset refurbishment and replacement
decisions.
Since the transmission utilities, in Australia at least, are operating under the
same market and regulatory conditions, this statement could imply that other
drivers, such as internal, business-related or system related drivers may also
play a significant role in defining asset management policies. Furthermore it
is proposed here that the interaction of all drivers may impact on AM policies.
Its is also hypothesised that existing conditions (eg debt level, system
configuration, load growth to name a few) and the effect of interacting drivers
may define or limit the effectiveness of asset management policies for
particular utilities.

Little, if any, research has been done to explore asset management policies
in an holistic manner, even though the literature suggests that managers are
actively considering strategy level AM issues, taking into account a range of
factors. While external drivers such as regulation and market impacts have
been, to some extent, described in literature, internal drivers such as
corporate culture issues and historical practices, and the influence of system-
related and environmental factors have been largely ignored. Yet these
factors may not be insignificant: substation configuration for example can
have a significant influence on reliability and availability of circuits, all other
conditions being the same: transmission businesses with high levels of
flexibility built into substation configurations will find it easier to achieve high
circuit availability.

Comparison of performance of utilities through benchmarking is important,

but of greater value to utilities is the knowledge of how to improve

75
performance, including financial performance. Benchmarking studies may
reveal areas that could be improved, but not whether the strategies employed
by other utilities will be effective when applied under different conditions.
Since the capital value of transmission assets is high, and revenue is directly
related to asset management, decisions on asset management policy can
have a major impact on the financial well-being of transmission enterprises.
However there appear to be few tools available to assist the transmission
asset manager to make such assessments. The proposed research is aimed
at providing the necessary links between input factors, policies and practices
and performances needed to explore strategic level decisions. The
processes and models developed here should, therefore, potentially be
adaptable for individual companies for use in strategic decision support.

In summary, the research outlined in Chapter 1

• Addresses the gaps in knowledge of strategic level transmission asset
management identified in the literature review and
• Has the potential to provide useful tools or methodologies to support
strategic decision making within transmission utilities.

76
3 Asset Management Relationships
This chapter is structured in three parts. In the first part the factors impacting
on asset management will be examined; the second part explores the
relationships between these factors; and the third part draws together these
relationships to construct a model of transmission asset management in the
form of an influence diagram.

3.1 Factors of influence for asset management in transmission

Factors impacting on asset management and system performance in
transmission enterprises may be divided into three groups:
• Internal practices and polices;
• External factors including the business environment; and
• System and environmental factors.

The relative importance of these factors varies from one transmission

business to the next; the interaction of these parameters results in significant
variations in overall asset management and system performance. Variations
can sometimes also be seen when comparison is made on a regional basis.

3.1.1 External / Business-related factors

There are some business-related factors that have a very strong impact on
asset management within transmission networks. In particular, revenue
sufficiency and historical revenue sufficiency are of major importance.
Revenue sufficiency relates to the revenue that the organisation obtains from
transmission of electricity and other sources, the reliability of that revenue,
and whether it meets the needs of the organisation. It will affect the day-to-
day operation and maintenance of the network, and also the organisation’s
ability to raise capital for new equipment. The level of revenue required will
depend on a range of factors, such as the size and geographical spread of
the network, network design factors, the current level of network performance
and any perceived need to change the level of performance: in the CIGRE
study, “deteriorating system performance” was ranked highest in a list of
factors influencing maintenance budgets. Just as revenue sufficiency can be

77
seen as affecting adequacy of maintenance, historical revenue sufficiency
will have affected historical maintenance effectiveness. Historical
maintenance effectiveness, in turn, is a factor contributing to current levels of
reliability.

Issues relating to the cost of capital can limit the level of system
augmentation and equipment replacement undertaken on the network. This
is discussed in greater detail in Section 3.2.8.

Labour costs can also play a part in asset management decisions. High
labour costs may make decisions to use online condition-based maintenance
more economically desirable. In areas with low labour costs enterprises may
undertake higher levels of labour-intensive maintenance activities.

There are also country specific external drivers. Regulations, for example,
may mandate replacement of assets at a certain age (for example in the
former USSR, standards mandated replacement of circuit breakers at 25
years of age (Batyayev et al., 1994)), or regulations may discourage
replacement. In the New Zealand regulatory framework, revenue is based on
the Optimized Replacement Value of assets (Energy Policy Group, 1995),
and there is less incentive to replace assets than under Australian
regulations where the Optimized Depreciated Replacement Value is used
(Crisp, 2001). Another example of external factors is the taxing of Sulphur
hexafluoride (SF6) or the proposed banning of (new) equipment using SF6
for insulation in some Western European countries, because of its
contribution to global warming as a “greenhouse gas” (Anonymous, 2002)

3.1.2 Internal factors

Internal practices and polices include management goals and priorities,
historical practices and attitudes. They relate to the structure and culture of
organisations, their level of expertise and understanding of the transmission
network.

78
A major driver of asset management is the goal of the organisation – whether
this is profit or meeting of some other defined targets. There is a strong
regional variation in profit as intention (Figure 8), with North America the only
region with more than 50% of respondents to the CIGRE study claiming this
goal, and no positive responses to this question from Eastern European
survey participants. Management priorities can also be deduced from the
choice of performance indicators. For instance, having profit as a goal and
maintenance cost as a performance indicator suggests a downward pressure
on maintenance costs.
Enterprises with profit

60
50
motive (%)

40
30
20
10
0
A AU EE NA SA WE
Regions

Figure 8 Profit as intention compared by region. Data are from (CIGRE SC 23 and 39,
2000).
Practices related to the replacement of assets affect financial- and system
performance. For example, selection of age or frequency of failure as the
dominant criterion for replacement of assets could result in quite different
network age profiles over time, with implications for debt, depreciation and
possibly, revenue. (Chapter 6 contains a more detailed examination of these
relationships.) Furthermore, results from the CIGRE study show variation in
accounting practices for replacement of equipment: some transmission
entities treat replacement of equipment as an investment under some or all
conditions. This may affect revenue or depreciation, depending on

79
regulations pertaining to the organisation, and in turn may make transmission
entities more or less willing to replace equipment.

Another subtle but important driver is the attitude to technology adopted by

the organisation. This may have an effect on decisions such as whether to
replace equipment, or whether to invest in online condition-based monitoring
equipment.

The level of maintenance undertaken, relative to manufacturers’

recommended levels, strongly influences routine maintenance spending. A
low level of maintenance activity might result from a highly targeted
maintenance program, using reliability centred maintenance (RCM) for
example. It might also result from inadequate maintenance expenditure, in
which case performance of the system is likely to deteriorate over time.
Similarly, a high level of maintenance relative to recommended levels implies
high spending, but might be the result of historical practices rather than need.
Figure 9 shows the regional spread of responses from (CIGRE SC 23 and
39, 2000) on this subject. It suggests that only in North America is there a
tendency to undertake less maintenance than recommended, whereas in
Eastern Europe, Asia and South America the tendency is to do more than
recommended levels.

80
 Percent of respondents
120
100
80
60
40
20
0
A AU EE NA SA WE
Regions

Figure 9 Maintenance relative to manufacturers' recommended levels compared by

region (CIGRE SC 23 and 39, 2000). Series from left to right are: a) more than
recommended; b) less than recommended; c) neither specified. This information is
relevant mainly for substation equipment such as transformers, circuit breakers and
protection equipment.

3.1.3 System-related factors

Load related factors impacting on asset management include the loading

level on equipment, the rate of load growth, and the load factor. Loading
level determines the heat dissipation within equipment. Equipment that is
frequently overloaded or overloaded by a large amount will have reduced life
expectancy and higher failure rates. The rate of load growth combined with
loading levels will impact on the need for system augmentation. In cases of
extremely high load growth (such as regional transmission in India) the rate
of system augmentation may not keep pace with the growth of demand (EIA,
2002a), in which case system overloading will occur and reliability of supply
will be reduced.

Load factor is the ratio of average electrical load on the transmission network
to peak load. If a network has a low load factor the equipment maximum
capacity will have to be high in order to meet maximum demand, but for
much of the time it will be loaded at much lower levels. This results in an

81
inefficient use of capacity, and a network with high replacement cost for its
energy throughput. A high load factor implies a more efficient use of
equipment, but if loading levels are also high it also makes taking equipment
out of service for maintenance more difficult to do without reducing system
security. On systems with high load factors, more maintenance might need
to be done at weekends or during the night when loading is lower.
Alternatively, use might be made of live working techniques and/or online
condition-based maintenance. The impact of load factor on maintenance is
also influenced by system configuration issues.

Transmission systems are generally designed to be highly reliable. A high

proportion of transmission networks (CIGRE SC 23 and 39, 2000) is
designed to an “N-1 planning criterion (CIGRE SC 23 and 39, 2000). This
stipulates that the network should be able to withstand the loss of any one
contingency (eg a failure of a piece of equipment) without loss of supply to
customers. However, within the general description of “N-1” there is a
multitude of variants as to the precise conditions applied, usually prescribed
in planning codes, for example (North American Reliability Council, 1994). In
addition, the system redundancy may be eroded during periods of high load
growth, if the rate of system augmentation is inadequate. Thus, while the “N-
1” level of redundancy is the norm, some networks will have higher or lower
levels of redundancy in practice. This can have a major effect on system
reliability.

Two other system configuration issues that affect system performance are
the flexibility of the system and the connectivity of the system. Transmission
circuits are generally connected together in substations. Circuits are
connected to busses via circuit breakers. Different arrangements of busses
and circuit breakers are used, giving different levels of flexibility for the
reconfiguration of circuits (Giles, 1970). Some of the common arrangements
are double bus, the breaker-and-a-half scheme, single breaker and ring-bus
systems. These are illustrated in Figure 10. Some configurations (eg
breaker-and-a-half and double bus) provide the flexibility to maintain a circuit

82
breaker without disconnecting the associated circuit from the network. This
can reduce outage time and fault restoration time.

single bus double bus

breaker and ring bus

a half scheme

Figure 10 Bus configuration schemes in use in various parts of the world

In flexible systems the circuit can be left in service while maintenance is
performed on the circuit breaker or other equipment within the substation.
This allows maintenance to be scheduled more easily and reduces system
restoration times in event of multiple coincident faults. The CIGRE study
asked respondents to enter the percentage of circuit ends capable of being
taken out of service for maintenance without taking the circuit out of service.
This information, described as “flexibility”, was recorded at three different
voltage levels. Combined flexibility may be calculated as an (weighted)
average from flexibility at each voltage, weighted by the number of bays at
each voltage level. In Figure 11, the combined flexibility is shown, and
compared by region. Flexibility is highest in Asia and South America, and
lowest in Western Europe.

83
 100

Percent flexibility
80

60

40

20

0
A AU EE. NA SA WE
Regions

Figure 11 Flexibility compared by region, calculated from data in (CIGRE SC 23 and

39, 2000). Results show significant variability in substation design practices across
regions.

The interconnectedness or connectivity of the network also has an effect on

its performance. A highly connected network has a larger number of
connection points between nodes, will be more robust in the face of faults,
and thus will have higher reliability, if all other factors are the same (See also
Section 3.2.14). A highly interconnected system may also permit loading of
individual components to higher levels, while still meeting the “N-1” reliability
criterion: upon loss of a single circuit, the load is spread across remaining
circuits, and on average, the more highly interconnected the system, the
greater the number of circuits feeding each node. The ability to load
equipment more highly presents greater opportunity for efficient use of
network capacity.

A measure of inter-connectedness of the system is shown in Equation 1. In

this thesis, the term bay refers to a group of associated equipment within a
substation. For instance a circuit breaker bay, terminating a transmission
line, might comprise one (or more) circuit breakers, one or more isolators (or
disconnectors), an earth switch, a voltage transformer and a current
transformer.

84
 total number of bays
connectivity factor =
demand in GW (1)
The regional distribution of connectivity factor is shown in Figure 12. Data
are from the CIGRE study.
Connectivity factor (GW )
-1

200
160

120
80

40
0
A AU EE NA SA WE
Regions

Figure 12 Connectivity compared by region, calculated using data from the CIGRE
study

Another factor that affects asset management in transmission enterprises is

the geographical spread of the network. Large distances between
substations and maintenance depots can increase the cost of maintenance,
particularly if associated with high labour costs. Using data from the CIGRE
study a geographic factor (Equation 2) may be defined, to compare relative
importance of geographical spread. This is illustrated in Figure 132.
total overhead line
geographic factor = (2)
total number of bays

2
Average circuit length might also be used, but these data were not available from the
CIGRE study

85
 50

Geographic factor (km)

40

30

20

10

0
A AU EE NA SA WE
Regions

Figure 13 Geographic factor is compared by region. Values are calculated from data
in the CIGRE study.

Equipment age and technology can also impact on the cost of maintenance
and on system performance. Some transmission equipment, such as the
circuit breaker, has seen great technological change over time. This has
(generally) resulted in improved reliability and lower maintenance costs for
newer equipment compared with older equipment. One manufacturer, for
instance, claims an increase in mean time between major failures (MTBF) of
their high voltage circuit breakers from approximately 20 years in 1972, to
more than 700 years in 1996, owing to technology and design changes
(Knobloch et al., 2000). Even for equipment such as transformers, where
there has not been any major change in technology for several decades,
there can still be age-related reliability considerations. Such equipment can
exhibit age-related degradation, which will reduce its reliability gradually over
a period of time (Tanaka et al., 1993, Allan, 1991). Transmission systems
having high proportions of newer or older equipment therefore should exhibit
respectively higher or lower system reliability, all else being equal.
Alternatively, if preventive maintenance is undertaken and effective, they
might exhibit similar reliability but different levels of maintenance cost.

86
3.1.4 Environment-related factors
Environmental conditions to which transmission networks are subjected vary
significantly from one region to another and even within regions.
Environmental conditions can have a significant effect especially on failure
rates of equipment and the costs associated with maintaining the system.
They can also impact on the cost of infrastructure. Environmental factors
include the following:
• Climate type and severity can affect failure rates and associated costs.
(This will be discussed in more detail in Section 3.2.1.1.)
• The presence of corrosive atmospheres (salt and/or sulphur) affects
maintenance costs and can be a significant consideration in
replacement decisions (Jones et al., 2000);
• Unstable geologic conditions affect cost of new assets (IEA, 1999);
• Rugged terrain makes installation and maintenance more costly;
• In rocky terrain with high resistivity it is more costly and difficult to
protect overhead transmission against lightning damage;
• Highly urban conditions make building new assets more difficult and
expensive (discussed further in section 3.2.12). According to an IEA
study, one of the factors contributing to the high price of electricity in
Japan is the high cost of land (IEA, 1999);
• Large distances between load and generation make networks more
costly to build and maintain, and increases likelihood of system
security problems such as poor dynamic stability.
• Remoteness of transmission from population centres can impact on
the cost of maintenance and response time to equipment problems
(Jones et al., 2000)

3.2 Important asset management relationships for transmission

enterprises
This second part of Chapter 3 deals with the relationships between factors of
influence in transmission asset management.

87
3.2.1 Factors affecting failure rate
In the context of this study, failure rate means the level of unplanned
equipment outages occurring on the transmission system per unit time, rather
than a specific hazard rate for a particular type of equipment. In general
failure rates can be represented as three possible phases: a wear-in or infant
mortality phase, a random failure phase and a wear-out phase, in the well
known bath-tub curve (See Section 6.3 for more detail). In this section
attention is largely given to the wear-out phase and the impact of
environment and loading and maintenance management techniques on
overall failure rates.

3.2.1.1 The impact of climate and loading on failure rate

An extreme tropical climate exposes equipment to high temperatures and
humidity, tropical storms, possibly hurricanes or cyclones, and potentially
high lightning flash levels. In such conditions failure rates of equipment are
likely to be higher than under mild or moderate climatic conditions. A
moderate or mild tropical climate will, however, still expose equipment to high
temperatures. The climate can therefore have an effect on both random
(non-age related) failure rates and condition-related failure rates that occur
mainly on aged equipment.

There is an interaction between climate and loading. When high ambient

temperatures are combined with overloading of equipment the effect is more
severe than if the same loading conditions occur under cool and/or windy
conditions, because the maximum temperature that develops within the
equipment will be higher. In the immediate period of the overload, the
loading condition can increase the likelihood of failure. For example
transformer oil subjected to temperatures above 140 degrees Celsius
produces gas bubbles which reduce dielectric strength, and increase the risk
of failure if transient overvoltages occur(IEEE Transformers Committee,
1991). In transmission lines, overloading can cause sagging of lines and in
the longer term, creep. High loading also exposes higher impedance joints to
overheating, with potential for failure to occur. Overloading of equipment can
be thought of as reducing the effective remaining life of equipment or

88
increasing its effective age, and under high ambient temperatures the effects
will be more severe. For instance the IEEE guide for loading mineral-oil
immersed transformers recognises aging of conductor insulation and
cumulative permanent degradation of mechanical properties of structural
insulation materials as functions of the time-temperature history of the
equipment (IEEE Transformers Committee, 1991). As mechanical strength
of the insulation degrades it becomes more vulnerable to mechanical
stresses, and will finally reach the point where even a small mechanical force
will cause dielectric failure (CIGRE WG 12.09, 1993). If predictive
maintenance is used and takes account of these effects, overall failure rates
are not likely to be increased significantly. However, equipment may need to
be replaced at an earlier calendar age than if overloading had not occurred,
and this could lead to an increase in the rate of replacement of equipment.

In tropical and sub-tropical climates, extreme climatic factors can be

associated with high lightning incidence. In equipment with paper insulation
(transformers, reactors), the likelihood of failure after a short-circuit fault is
higher if the paper is aged and significant de-polymerisation has occurred
(IEEE Transformers Committee, 1991). The impact of system-wide failure
rate is then related to the number of aged pieces of equipment (with oil-
impregnated paper insulation) and the lightning incidence (rate).
Transmission enterprises in climates with a high lightning incidence may
therefore find it necessary to replace some types of equipment at a lower age
than companies operating in milder climates.

Transmission businesses operating in milder climates might also observe

less of a relationship between failure rate and age of equipment. For
instance, CIGRE Working Group 12.09 (Thermal aspects of transformers)
(CIGRE WG 12.09, 1993) notes:
Among the reported transformer failures, comparatively little
information is available about transformers which fail
primarily as a result of thermal ageing of the insulation
materials. Instead we find a consensus of opinion that
abnormal events such as overvoltages and system faults are
much more detrimental than long-term ageing.

89
 Nevertheless, long-term ageing may make the transformer
more prone to failure through such events.

This view was also expressed in discussions with US utility representatives

for example (Sherard, 2001).

In other equipment, such as circuit breakers and surge arresters, the

incidence and severity of faults from lightning or other sources affects the
condition of the equipment. Wear on arcing contacts in circuit breakers is
related to the energy of the fault cleared; it is cumulative and therefore also
dependent on the number of faults cleared. Mechanical wear on moving
parts of circuit breakers is related to the number of operations. This depends
on the duty of the circuit breaker, as well as the number of protection
operations. For example, circuit breakers switching reactive devices may
operate far more frequently than circuit breakers protecting a transmission
line.

Failure rates on transmission overhead lines owing to lightning are also

dependent on line design, and especially the use of earth-wires for lightning
protection. In general most transmission networks in high lightning incidence
locations would be designed with overhead earth-wires. However, in areas
of lower lightning incidence they may not be used, or may be used only on
higher voltage circuits. The decision to use or not use earth wires becomes a
trade-off between cost and equipment reliability. For instance, the South
Australian regulator reports comments from Electranet SA about its service
outages (SAIIR, 2001):
Electranet SA has noted that the 132kV system in SA was
typically developed with limited lightning protection.
Reducing the risk of supply interruptions due to lightning
would require significant additional capital expenditure …
A recent study from Sweden (Karlsson, 2001) suggests that even for systems
with low lightning incidence the use of earth wires on overhead lines provides
significant benefits in reducing the number of flashovers involving more than
one phase (which reduces the severity of voltage dips, thus improving
voltage quality), and in reducing the failure rate of wood poles. Use of

90
overhead earth wires also reduces the voltage stresses experienced by
substation equipment, which may reduce failure rates.

Apart from lightning, another possible source of fault is from trees falling
across lines. Vegetation is more likely to be a problem in places with high
rainfall, and forested landscape, and especially in warm to hot wet climates
where vegetation growth is rapid. This source of fault is not outside the
control of the transmission company, and need not result in higher failure
rates provided an effective vegetation management program is in place.

High levels of humidity may be a problem, especially for cyclically loaded

equipment. As equipment cools moist air is sucked into equipment, and can
condense inside. Moisture degrades the insulating properties of liquid
insulation (eg oil). For example, CIGRE WG12.09 (CIGRE WG 12.09, 1993)
attributes moisture ingress to some of the random (non-time related) failures
of transformers. Moisture can also accelerate the degradation of cellulose in
paper insulation (CIGRE WG 12.09, 1993). Moisture ingress in transmission
substation equipment may be controlled by use of silica gel or heaters.
These drying mechanisms (especially the silica gel) need to be maintained
regularly to ensure adequate performance. If moisture condenses on solid
insulation subject to particulate pollution, this can cause electrical breakdown
across the surface of the insulator. In XLPE and PE cables subjected to high
moisture contact, water treeing can occur, although this is more common in
distribution cables than transmission cables because distribution cables are
typically constructed without a protective jacket (Bernstein and Brancato,
1993).

Equipment operating in dry dusty conditions can also be susceptible to failure

from pollution. Insulator strings on overhead transmission lines can become
coated with dust. When light rain or heavy condensation occurs this can
significantly reduce insulation level, and may cause flashovers. There are
ways to manage this problem:
• Lines may be built with longer insulator strings – this increases capital
costs.

91
 • A program of washing insulators may be set up – this increases
maintenance costs.
• In times of dry weather a monitoring program can be set up to
ascertain the risk of flashover, and initiate washing as required – a
(potentially smaller) increase in maintenance costs.

For moderate or sub-arctic climates extreme climatic factors might relate to

ice storms – these tend to increase the mechanical loading of overhead lines.
In cooler climates loading level is only likely to affect failure rates when
overloading is frequent or severe.

3.2.1.2 The impact of equipment ageing on failure rate

There are some failure modes that are strongly related to the ageing of
equipment. For example Kogan et al (Kogan et al., 1988) show an age-
related hazard rate for 345kV transformers above 25 years of age. There are
also other failure mechanisms of a cumulative nature that will result in an
increased failure rate over time, depending on the duty of the equipment, and
whether preventive maintenance is undertaken and effective. Wear on
contacts of circuit breakers and within mechanical components are examples
of this.

Transmission conductors are also susceptible to ageing. Corrosion of

aluminium strands is a major time-related cause. Another factor is material
fatigue from wind-induced aeolian vibrations (Aggarwal et al., 2000).
Insulators on transmission lines age faster than conductors and towers, and
have a number of failure modes. In the state of Queensland, Australia,
where towers and conductors generally experience lives beyond sixty years,
the life of insulators can vary from up to 50 years at 110kV in clean
environments to around 10 years at 275kV in heavily polluted environments
(Lee, 1997). The impact of insulator failures depends on the particular failure
mode: a cracked or shattered insulator will reduce electric strength of the
insulator string, but by itself is unlikely to cause serious consequences. Pin
corrosion, by contrast, can result in a broken insulator string causing the line
to drop and possibly flashover.

92
It is common practice, once a major failure is identified, to check other
devices of the same type and eliminate the possibility of failure by this
mechanism by either maintenance or design changes. A documented
example illustrating this practice is the NV Sep description in (Davies et al.,
1998). This results in a lower failure rate than otherwise would have been
recorded, a phenomenon known as data censoring.

It is also recognised that some equipment technologies have shorter life

expectancies than others and some designs have higher mortality rates than
others. There may also be a period after the introduction of a new
technology, in which design problems occur that are eliminated in later
models. These factors tend to cloud the issue of age-related failures.
However, overall there is still a trend to higher failure rates with age. Figure
14 illustrates the failure rate for circuit breakers of oil and SF6 types,
extracted from the database of one transmission company (Hastings and
Crisp, 2001).

50
(per 100 CB years)

40
Failure rate

Oil
30
20 SF6
10
0
1950s 1960s 1970s 1980s 1990s
Decade of installation

Figure 14 Circuit breaker failure rates as a function of age

93
3.2.1.3 The link between maintenance history and failure rates
It is logical to assume that there is a link between maintenance and failures in
equipment, and thus reliability. Why else are predictive and routine
maintenance done, other than to decrease the potential for failures, and
improve or maintain reliability? The empirical link has rarely been shown.
One important reason for this is that it is often maintenance history, rather
than current maintenance or spending that determines failure rate.

Cracked insulators on transmission lines can be used to illustrate this point:

a single cracked insulator is unlikely to cause an outage, but as more
insulators become cracked over time, the electrical strength of the insulator
string will be degraded, and failure will eventually occur if no corrective
maintenance is undertaken. Another example is maintenance of silica gel
breathers on transformers with free-breathing oil expansion systems. Failure
to maintain the silica gel will result in increasing moisture content in the
transformer. This in turn accelerates degradation of the paper insulation
(degradation rate is proportional to moisture content (CIGRE WG 12.09,
1993)), and degradation of the insulation makes the transformer more
susceptible to failures from mechanical stresses such as might be caused by
voltages surges or faults.

Maintenance history includes three components: the extent and frequency of

maintenance, and the quality of maintenance. The time lag between a
change in maintenance activity (level or quality) and a change in failure
trends will depend on the failure modes involved, and other factors such as
climate and loading. Discussions with industry personnel suggest that, for
transmission system equipment, the time lag is probably of the order of 5-10
years, depending on environmental factors. In Kazakhstan (World Bank,
1999a) for instance, suspension of maintenance brought about a marked
increase in failures on the system in approximately 7-8 years. Another way
of expressing this is that the degradation rate depends on the maintenance
effectiveness and conditions under which the equipment is used. (See also
Chapter 6).

94
3.2.1.4 The effect of predictive maintenance on failure rates
Predictive maintenance can reduce the number of failures by
1. Predicting that a failure will occur, based on condition assessment,
possibly coupled with history of equipment duty, and
2. Taking actions to restore the equipment to good condition (or
replacing the equipment) to avoid the predicted failure.

Predictive maintenance can be as simple as inspection and assessment of

corrosion on transmission towers, or as complex as Dissolved Gas Analysis
(DGA) for transformer oil degradation. Therefore the general level of use of
predictive maintenance will have an effect on failure rates on the system as a
whole. A low level of use of predictive maintenance will have little effect on
failure rates for the system, whereas a high level of predictive maintenance
will reduce failure rates over what would have occurred otherwise.

3.2.1.5 Geographical factors affecting failure rate

Geographical factors can also increase failure rate. Specifically, salt-laden
air associated with coastal installations can accelerate corrosion of metallic
components including overhead transmission lines and towers, and the
structures supporting substation equipment. Sulphur compounds generated
by geo-thermal activity are also known to accelerate corrosion. New Zealand
has particular corrosion problems associated with the combination of these
factors (Jones et al., 2000). Atmospheric contaminants especially sea-salt
and certain sulphates are known to attack Portland cement, used in the
construction of cap and pin insulators that support the conductors of
overhead transmission lines (Aggarwal et al., 2000). Furthermore, surface
contamination of insulators from sea salt or industrial contaminants can
cause flashovers of overhead lines in a manner similar to that described for
dust (Section 3.2.1.1).

3.2.1.6 The effect of voltage on failure rate

As the voltage of transmission is increased the electrical stresses on
equipment can also be increased. This is particularly relevant to insulation
systems. As Berstein states (Bernstein and Brancato, 1993) the functional

95
life of electrical insulation systems is directly influenced by the intensity of
stress or stresses -voltage, temperature, mechanical and environmental -
imposed on the insulation during the operation of the equipment. For
instance Aggarwal (Aggarwal et al., 2000) reports that the failure rate of
insulators on 400kV lines is far greater than that for 275kV lines. Most
defective units are at the high voltage end of the string, adding weight to the
argument that the failure rate is voltage-related.

Some dependence between voltage and failure rate has also been reported
for circuit breakers and transformers. Results from the Second International
Survey of Reliability for Circuit Breakers above 63kV (Janssen et al., 1995)
clearly illustrates a voltage dependency for failure rate, as shown in Table 2.
Results from the Second International Survey of GIS in Service (Chan et al.,
1998) (Table 3) are less clear.

Table 2 Major failure frequency for SF6 circuit breakers 63kV and above, placed in
service after Jan 1978
Failure rate (failures per 100
Voltage range (kV) Sample Size (CB years)
breaker yrs)
63-<100 0.275 24335
100-<200 0.680 23520
200-<300 0.814 10933
300-<500 1.210 9917
500+ 1.967 1983

Table 3 Major failure frequency for GIS in service

Failure rate (failures per 100 CB-
Voltage range (kV) Sample size (CB-bay yrs)
bay yrs)
60-<100 0.05 56884
100-<200 1.45 32048
200-<300 0.86 16040
300-<500 2.81 6371
500-<700 1.08 4525
700 + 6.0 200

3.2.1.7 Other factors affecting failure rate

Equipment quality and specifications can also play a part in failure rate. For
example an international survey on performance of GIS equipment (Chan et
al., 1998) found that the failure rate of equipment in Japan was significantly
lower than that of the rest of the world. The survey report attributes this to
very conservative specifications for equipment in that country. A corollary to
this is that poor quality equipment can impact on overall failure rates (Mohd

96
Noor, 2002). This may be caused by design failures or quality control
problems during manufacturing. Purchasing of poor quality equipment can
result from failure to undertake life cycle cost analysis, or, particularly in
some developing nations, companies may be compelled (by governments) to
purchase from particular suppliers.

In some cases quality problems may arise from the failure of usually reliable
suppliers to master a new technology. For instance higher failure rates of
GIS in the USA compared with other countries has been attributed (among
other causes) to supply of prototype designs to customers, design and
production problems in early stages of manufacture and lack of appreciation
of the need for site assembly under clean conditions, especially at high
voltages (Eriksson et al., 1994).

In some cases also there may be a lack of skill in engineering or

maintenance personnel that results in inappropriate application of condition
monitoring, or ineffective maintenance practices. This can increase the cost
of maintenance without gaining the benefits of the additional expenditure,
and may actually increase the failure rate.

Skill and purchasing issues are often difficult to detect without first-hand
knowledge of the transmission organisation concerned, because they are not
likely to be admitted in writing by the company. However they may be quite
significant in some cases.

3.2.1.8 Summary of factors affecting failure rate

Factors affecting failure rate can be categorised as those leading to
immediate failure and those resulting from cumulative effects. Failure rates
can be reduced by use of predictive maintenance that facilitates detection of
equipment in poor condition, and action being taken to avoid failures.

Cumulative degradation effects are influenced by the effectiveness of

preventive maintenance, loading and climate (especially high ambient
temperatures and humidity), as well as general time-related ageing.

97
Cumulative degradation effects can be thought of as contributing to the
proportion of poor condition equipment on the network.

Immediate failure triggers include climatic effects, particularly lightning

incidence and high winds, but also ice storms. Excessive overloading can
also contribute to failures by overheating components such as joints.

Voltage as a factor contributing to failure rate has not been modelled in this
study, because most utilities have a range of transmission voltages, and EHV
equipment which has the highest failure rates is comparatively rare.
Therefore it is considered that overall transmission voltages will not play a
major role in system-wide failure rates.

3.2.2 Failure rate and corrective maintenance

The level of corrective or remedial maintenance (and replacement of failed
equipment) is directly related to the failure rate. Even with perfect
maintenance it will not be possible to eliminate all remedial maintenance,
because of the existence of random failures.

3.2.3 Use of time-base maintenance

Scheduled repair/replace/overhaul and scheduled test and inspection are
traditionally commonly applied maintenance methods in the transmission
sector. They are commonly referred to as time-based maintenance.
Maintenance is scheduled based on age, operating life or number of
operations. Overhaul involves stripping an assembly down, and rebuilding
using only serviceable (new or used) components. Inspection and test
methods are necessary to detect hidden failures (Moubray, 1992) (for
example, failure of a relay or circuit breaker to operate on command).

Typically manufacturers provide guidelines recommending maintenance

intervals for their equipment. These guidelines cannot make distinctions for
the different environments and operating conditions under which the
equipment will be used, and therefore may be conservative. For instance,
after assessing the feedback on maintenance activities from maintenance
staff, NV Sep found it possible to increase maintenance intervals on most

98
bays from two to three years, and for some maintenance activities the
interval was increased from two to six years (Davies et al., 1998).
Companies which are willing to reduce maintenance below recommended
levels, and which operate their equipment under conditions that make this
possible without increasing failure rates, may be able to reduce maintenance
costs in this way.

3.2.4 Use of predictive maintenance

Predictive maintenance has two components, online condition-based
maintenance and offline condition-based maintenance. The key difference
between inspection and condition monitoring is the ability to track the
changes in condition between one measurement and the next. This enables
the rate of change of condition to be determined leading to a more pro-active
response. This may either avoid a failure or mitigate the consequences of
the failure.

3.2.5 Use of online condition-based maintenance

The use of online condition-based maintenance (CBM) is limited by the
opportunities for the technology to be employed and the inclination of the
company to use it. Online CBM is often easier and cheaper to implement on
new equipment, even though there may be more benefit in implementing it on
aged equipment (CIGRE WG 13-09, 2000). It often has a high capital
component – the monitoring hardware can constitute a significant proportion
of the equipment purchase price. Recently methods have been developed to
assess the value of online CBM ((CIGRE WG 13-09, 2000, IEEE, 2000,
Bergman, 2001)) and these are becoming easier to apply as a body of data is
built up on the performance of online CBM systems in practice. However
earlier decisions by transmission entities were based less on hard data and
well-developed analysis techniques, and more on attitude to technology.
Since use of online CBM technology is usually associated with installation of
new equipment, opportunity to employ online CBM can be identified by the
adequacy of the company’s financial resources over a number of years and
the proportion of young equipment on the system, and inclination to employ it
by the company’s attitude to new technology.

99
Another, perhaps less significant, driver is geographic factor. A high
geographic factor, implying long distances between population centres, may
be an incentive to use more online CBM, especially when high travel times
combine with high labour costs (Jones et al., 2000).

3.2.6 Use of offline condition-based maintenance

Use of offline CBM has become quite common – it is especially suited to
equipment that does not have a high routine maintenance component such
as transformers. Techniques often require a certain level of skill in testing
and assessment, for example dissolved gas analysis (DGA) for oil testing. In
general they do not have the same high capital requirements as online CBM,
but have a higher labour component. The CIGRE study found some
correspondence between use of maintenance cost as a KPI and use of
offline CBM. This was attributed to Offline CBM being a less conservative
(and hence less expensive) maintenance strategy than time-based
maintenance.

3.2.7 Factors affecting maintenance spending

The drivers for maintenance spending are a complex blend of many different
factors. In one sense maintenance spending is dictated by the amount of
money available. If revenue is inadequate then maintenance spending is cut,
especially if maintenance involves use of external resources. When revenue
is sufficient or better, other factors come into consideration. The perception
that efficiency in operations and maintenance could improve has led
regulators to apply downward pressure on these costs. This has been
explicitly addressed by the X factor in the rate-capped and revenue capped-
regulation in the UK and Australia (see also Sections 2.2.1.2 and 0).

The maintenance budget is driven down by a number of factors including the

desire to make a profit, especially for privately owned companies that have
shareholders (and share prices) to consider. Internal drivers on costs – for
instance the use of maintenance costs as a performance indicator – can also
push costs down.

100
In investor-owned utilities where maintenance is seen as a negative by
investors and investment as a positive, there is also an incentive to account
as much as possible to investment and as little as possible to operations and
maintenance. For example FERC (the US federal regulator) allows the
capitalisation of the use of a product called Cable Cure by Utlix, used to
restore the dielectric strength of underground cables (Teixiera, 2001). This
practice may contribute to the very low values of direct maintenance cost
reported for North America in the CIGRE study. A wide range of accounting
practices with regard to maintenance and capital has been observed in
Australian utilities (Beardow, 2003). Under the Australian regulatory
framework there may be an advantage in treating expenditure as
maintenance, up to the amount permitted in the regulatory determination.
The rationale for this is that expenses are recovered immediately, but
capitalised items run the risk of having their value reduced in subsequent
regulatory resets.

Maintenance management issues and system factors also impact on

maintenance spending. The failure rate dictates the amount of spending on
corrective maintenance. The other costs are for time-based and condition
based (predictive) maintenance activities. A high level of predictive
maintenance should reduce maintenance costs, if properly applied, because
it should reduce unnecessary invasive maintenance. However, offline-
condition based maintenance may add to maintenance expenditure,
especially if labour costs are high, while the effect of automated on-line CBM
in reducing maintenance expenditure is less ambiguous. Energie Noord
West of the Netherlands achieved a reduction in overall maintenance effort of
approximately 20% by changing from a time-based to a condition-based
maintenance policy (Davies et al., 1998). Targeted maintenance, through the
use of RCM or other means to optimise maintenance activities, should help
to keep costs low. On the other hand, doing more maintenance than
recommended by manufacturers is likely to result in high maintenance costs.

System factors impacting on maintenance costs include redundancy,

flexibility and connectivity. If these are at high levels there tends to be a

101
reduction in maintenance costs, because it is easier to undertake
maintenance during normal working hours. The effect of connectivity on
maintenance cost is tempered by the loading level (and load factor). At times
of very high loading, equipment may not be accessible for maintenance, even
on systems with high connectivity, because of the reduction in system
security that occurs when equipment is taken out of service. Hence high
connectivity will only reduce maintenance cost significantly on lightly to
moderately loaded systems.

Maintenance spending tends to increase with high geographic factor,

because of potentially higher travel costs. The effect is magnified if labour
costs are high. A high level of usage of online condition-based maintenance
will tend to reduce the cost penalty associated with high geographic factor.

3.2.8 Investment in new equipment

Investment in new equipment comprises two components – replacement of
existing equipment, and augmentation of the network.

3.2.9 Drivers for equipment replacement

Investment in new equipment has two elements: equipment replacement and
augmentation of the system. In practice, there are a number of reasons for
replacement of equipment:
• The equipment has failed, or is likely to fail soon. Assessment of
condition, and/or statistical analysis of equipment population may be used
to determine the risk of failure.
• The equipment is not fit for duty (insufficient capacity or fault rating).
• The equipment is obsolete (including factors such as inability to obtain
spares, or lack of technical skill required to maintain it).
• The equipment has reached its end of life, determined by some criterion
such as age, economic end of life, risk cost, or the requirement of
regulations. Economic end of life may be influenced by factors such as
the depreciation method employed, or the way that revenue is determined
(eg Australian system where revenue is based on deprival value of assets
(Crisp, 2001)).

102
The rate of replacement owing to failure is determined by the failure rate.
The conditions that cause equipment to be in such poor condition that it
needs to be replaced are the same as those that cause failure in the wear-
out phase. However, because of the use of predictive maintenance, failure is
averted and the equipment is replaced instead. Fitness for duty may be
related to load growth, but this could be considered as system augmentation,
rather than purely replacement. Equipment replacement for reasons of
obsolescence is probably a very small percentage of cases, and is unlikely to
impact much on overall replacement rate of equipment (except, perhaps for
secondary systems such as communications equipment and protection
relays). End of life considerations will be influenced by the criteria used for
replacement. If age is a replacement criterion, then the proportion of aged
equipment on the system will also influence the overall rate of equipment
replacement. Economic end of life could also be influenced by the treatment
of replacement as either investment or maintenance. In investor-owned
companies, investment is usually seen as positive by the share market,
whereas increased maintenance cost can be seen as negative (Sherard,
2001). If revenue is not adequate, the level of replacement for reasons other
than failure is likely to be very low.

3.2.10 Refurbishment of equipment

The purpose of refurbishment is to restore the equipment to good condition,
so as to extend its life. Refurbishment may be considered as an alternative
to replacement of equipment.

Criteria for refurbishment may be condition, age or possibly failure rate. The
ability of a company to ascertain the need for refurbishment or replacement
of a piece of equipment, on the basis of condition, will depend on its level of
usage of condition-based maintenance. However, if preventive maintenance
is effective some detection of poor condition equipment will be possible even
if the use of predictive maintenance is not at a high level.

Because refurbishment is generally major maintenance it may be treated as

capital expenditure. In this case the book value of the asset is likely to be

103
revised to reflect the expected remaining life of the refurbished asset. This
allows capital to be recovered and a return on investment to be achieved.

The choice between refurbishing and replacing equipment is a complex one.

It may depend on the company’s assessment of the design and/or
manufacturing quality and performance history of the particular equipment,
whether it is likely to suffer technical obsolescence (eg spare parts no longer
supported by manufacturer), or whether its capacity is likely to be adequate
to meet future needs (current loading, predicted load growth). It may also
depend on the current revenue sufficiency of the company and the current
overall level of capital expenditure on replacement equipment and system
augmentation. Because only a portion of capital expenditure is funded from
revenue (the rest from debt) there may be a need to smooth capital
expenditure. In this case it may be advantageous to undertake refurbishment
(low capital cost, or perhaps treated as expense) than replacement (higher
capital cost). If revenue is inadequate both replacement and refurbishment
may be curtailed.

Other possible criteria influencing the decision between refurbishment and

replacement include:
• The desire to increase the level online condition monitoring on the
system (this tends to be much easier to achieve and less expensive
on new equipment than on old equipment); or
• Pressure to reduce the size of equipment by using compact
technology equipment.

3.2.11 Drivers for system augmentation

The main reason for system augmentation is to maintain or improve reliability
of supply, including connecting new customers (loads or generators) to the
grid. A second reason for augmentation which has arisen with the advent of
electricity markets is to facilitate market operation. In Australia these two
justifications are specifically addressed in (ACCC, 1999b). The need to
augment the system may arise from insufficient capacity, as a result of
• Load growth;

104
• Increasing system fault levels; or
• Changing patterns of network loading, as a result of energy trading, new
generation or new interconnections (including network constraint
management).
Augmentation may also be a solution to security of supply issues, such as
dynamic and transient stability.

In many regulatory frameworks the revenue obtained by a transmission

company is linked to the value of its assets, so there may be an incentive to
augment the system.

Other factors, related to determination of future revenue, may also play a

part. For instance, in the Australian system, augmentations to the network
are included in the asset base for calculation of revenue, once they are
operational, only if the regulator deems the investment prudent. To convince
the regulator of this, the transmission network service provider must
demonstrate the problem and consider alternatives such as demand-side
load management solutions or building of additional generation, in order to
show that system augmentation is the most appropriate solution (ACCC,
1999b).

Non-financial issues may also reduce the level of augmentation of the

network. In the United States of America, much attention has been given to
the slow pace of transmission network augmentation in recent years,
compared with load growth for example (Hirst, 2000). Reasons cited include
• Public opposition to the building of transmission lines;
• Difficulties arising from dealing with multiple bureaucracies for trans-
state interconnectors; and
• Uncertainty over the direction of regulatory reform.

3.2.12 Investment spending

The ability of the transmission business to raise capital is an important factor
impacting on the level of investment in new equipment. If capital is difficult to

105
obtain or is too expensive, this will limit the potential for system augmentation
and equipment replacement.

The cost of implementing augmentations to the network or replacing existing

assets also depends on the cost per item. This is dependent on a number of
factors:
• Land costs in urban terrain can be much higher than in rural areas.
For example Heidari (Heidari and Heidari, 2002) reports that
transmission line costs in urban areas in Iran are ten times higher than
in rural areas, all other things being the same. In networks operating
in highly urbanised regions (eg Japan, Western Europe), the average
overhead transmission line costs per kilometre are therefore likely to
be significantly higher than in less urbanised regions.
• In urban areas underground cables are much more likely to be used,
because of the scarcity of land, the cost of land and the visual impact
of towers. Underground cables are significantly more costly to install
than overhead lines.
• Similarly, land scarcity and cost encourages the use of compact
substation design, especially the use of gas insulated substations
(GIS) which have higher capital cost than equivalent outdoor air
insulated equipment.
• In earthquake zones structures will be more costly to build than in non-
earthquake zones.
• In areas prone to ice storms, transmission tower design must take into
account much higher wind-loading, resulting in higher capital cost for
transmission lines.
• Long transmission lines are more expensive than short ones, and thus
networks with a large geographical spread will incur higher capital
costs to build new lines than more compact networks.

Cost per item can be reduced by purchasing efficiencies that result from
quantity of purchases. In some cases transmission entities can develop
special relationships with vendors, for instance by being pioneers for

106
 introduction of a new technology into the market. They may therefore be
able to negotiate significantly lower prices for equipment purchases.

3.2.13 Ease of raising capital

If the risk associated with lending to an organisation is high, the cost of
capital is also likely to be high, and capital may be difficult to obtain. Sources
of risk include
• Debt level of the company (Risk associated with debt level is
generally lower if the debt is government guaranteed.);
• Revenue sufficiency (Does the company have sufficient funds
available to meet its needs? Is the effective rate of return
commensurate with the level of risk associated with the business?);
• Regulatory governance (factors such as consistency, transparency of
regulatory decisions and absence of political interference) (Stern and
Holder, 1998);
• Regulatory risk (appropriateness of future regulatory decisions -
contributing to future revenue sufficiency)(Larsen and Bunn, 1999);
• Currency risk (financial exposure of the company to volatility in
exchange rates, and related to the economic performance of the
country involved);
• The quality of financial management and asset management;
• Access of the company to credit at low interest rates (eg Japanese
companies) may also be taken into account;
• Business management skills level.

The combined risk associated with lending to a company, and hence the cost
of capital, is reflected in the company’s credit rating, if it has one.

3.2.14 Reliability of supply

From the NERC definition of reliability given in Section 2.3.7 it can be seen
that the design of the system is a major factor in providing for both security
and adequacy of the system. This has been represented in the model with a
parameter, redundancy. As described in Section 3.1.3, redundancy level is
specified in planning codes for individual jurisdictions. Redundancy can be
altered by a mismatch between demand growth (and hence required growth

107
of transmission capacity) and augmentation of the network. A reduction in
redundancy might arise from very high load growth and inadequate capital to
finance augmentation. Occasionally demand can fall: many countries of
Eastern Europe experienced reduced demand for several years after the
collapse of the Soviet Union (Lovei, 1998, Dobozi and Pohl, 1995). In this
case redundancy and reliability for the system might actually increase, in the
short term, beyond design levels3.

The NERC definition also shows the dependence of network reliability on

failure rate (unscheduled outages of system components) and maintenance
activities (scheduled outages), but only if these impact on the customer.
Flexibility of the network (defined in Section 3.1.3) affects both scheduled
outages and restoration times for unscheduled outages.

Connectivity of the system (defined in Section 3.1.3) reduces the likely

impact of outages on the supply to customers. This is easily seen by
considering reliability theory:
For parallel systems the probability of failure of the system (in the
transmission network case, at the supply point) is given by the product of
probability of failures of each parallel element (Equation 3) (Hastings, 2000)
Fsys = F1 x F2 x F3 … (3)
Since all probabilities of failure are less than (or equal to) one, a system with
higher number of parallel circuits supplying a node will have lower probability
of loss of supply.

For series systems where all probabilities of failure are small (Equation 4)
(Hastings, 2000)

Fsys ≈ F1 + F2 + F3 … (4)

This has implications for circuits. A circuit can be considered to have a

failure rate that depends on its length, because elements such as
conductors, towers and insulators are distributed along its length. Therefore

3
In fact problems of sustained overvoltage occurred (Batyayev et al., 1994) as a result of the
lightly loaded lines, which may have increased failure rates, offsetting the effects of
increased redundancy.

108
the longer the conductor, the more series elements contained in the circuit
and the higher its potential unreliability. Thus for systems having high value
of average line length the proportion of unreliability owing to lines will be
higher compared with the proportion owing to substations.

In practice there are many different ways of measuring reliability used

throughout the world. This makes it difficult to compare reliability statistics.
(See Section 5.2.2 for a more detailed discussion of this.) A related problem
is the extent of the network considered in the calculation. Many networks
have different levels of redundancy at different voltage levels – generally
higher at higher voltages where loss of supply can affect greater number of
people. At sub-transmission and distribution level voltages there may be little
duplication of supply circuits. Therefore reliability measurement will be
significantly affected by the definition of the transmission/sub-transmission
boundary.

Reliability is a measure of great interest to regulators. Some measure of

reliability is often used as a measure of performance of the network. In some
cases there are also incentives or penalties associated with reliability, in
which case there is a financial driver on the transmission enterprise to
maintain or improve reliability. However, reliability is not something that can
be instantly improved: it can be improved gradually by taking measures to
reduce failure rates or making changes to system design.

3.2.15 Availability
Availability is a measure of the proportion of time that a component or system
is available to do its job. Circuit availability has historically been considered a
useful measure of performance in some countries, notably Australia (ACCC,
2002b). It includes planned and unplanned outages of circuits and does not
take into account impact on the customer. Unplanned outages relate to
failure rate and associated remedial maintenance actions. Maintenance
practices that reduce the number and duration of scheduled and
unscheduled outages will also impact on availability. These include:

109
 • Use of predictive maintenance, and particularly online condition-based
maintenance – this reduces routine maintenance outages and
• Combining maintenance activities by circuit – this reduces the number
of circuit outages; and
• Use of live maintenance techniques for lines and/or substations.
Flexibility within substation configuration also reduces restoration time and
the number of scheduled circuit outages by enabling substation equipment to
be maintained without removing circuits from service.

Major maintenance work, especially at the level of substation refurbishment

or replacement and the construction of network augmentations, can have a
significant impact on the availability of circuits. The extent to which
availability is affected depends on the duration of the outages associated with
the work, as well as the number of projects. Some utilities are able to rebuild
substations adjacent to existing infrastructure either within the substation
boundary or in an adjacent site, and then reconnect existing equipment to the
new equipment. This practice minimises the outage time. However if land is
not available to do this then a significant reduction in circuit availability is
likely. Similarly major overhauls of circuits will reduce availability.

Circuit availability is also sometimes used as a performance measure by

regulators, and in some cases (eg UK, Australia) incentives/penalties can
apply. In situations where transmission lines are privately owned the
availability of individual circuits may be used as a performance measure for
revenue determination (Nolasco and Ferreira da Silva, 2002). Availability
can be quickly improved by avoiding taking circuits out of service for
maintenance. If this translates to reducing maintenance below levels
necessary to support reliability then it will be counter-productive in the long
term. If reduction in planned outages is achieved by introduction of live work
practices, there is an associated cost and safety risk penalty. Alternatively
availability could be improved gradually by taking actions to reduce failure
rate. Therefore regulators should use caution when setting targets for
availability – optimum availability is not necessarily the same as maximum
availability, and short-term solutions may have long-term detrimental

110
consequences for network reliability and/or maintenance costs. This and
other issues relating to the regulation of transmission enterprises are
discussed in detail in Chapter 7.

3.3 Summary of AM relationships

In Figure 15 the relationships discussed in the foregoing sections of this
chapter are combined as an influence diagram of the causal-loop-diagram
(CLD) form (Pidd, 1996). This construction is not intended to be a precise
mathematical representation of the problem, but an aid to understanding the
structure of the system (Lane, 2000). In this diagram the arrows indicate the
direction of influence. A positive sign indicates that a change in the element
at the tail of the arrow will produce a change in the same direction in the
element at the head of the arrow. A negative sign indicates that the change
is in the opposite direction.

3.3.1 Key elements in the model

Several key elements are evident from the diagram. The central role of
failure rate is highlighted. It features in replacement decisions, maintenance
spending, performance of the system in terms of reliability and availability.
The aspects of failure rate that are related to degradation of condition are
incorporated in the model element proportion of poor condition equipment.

Another key element is available financial resource, which has an influence

on maintenance activities and investment activities and through these also
indirectly influences performance of the network. Available financial resource
is largely determined by the level of permitted revenue. This in turn is
influenced by the level of regulatory support. The extent to which
performance incentives affect available financial resource depends on the
targets set, whether they are met and the level of penalties or rewards. The
debt exposure and interest payments also influence available financial
resource.

Replacement and augmentation of the system are influential parameters.

Replacement, in particular has a significant role in feedback loops within the
system. This aspect of the model is described in the next section.

111
 +
+ routine + maintenance rel
skill labour maint cost to
geographic
cost recommended
spread - +
attitude to + +
refurb.
technology
- + cost
+
+ use of offline +
+ CBM - maintenance
proportion of spending +
equipment + financial
installed since use of
+ drivers
1990 predictive - + -
maintenance -+
+ refurbishment -
replacemt
+ ++ of equip. + + maint.cost

+ + +
maint.
+ + - corrective
effectiveness
taxes &
- + maint. cost dividends
use of -
online CBM -
+ + -
proportion
+ live - revenue
+ or failure rate + poor condit.
+ equip. + available
+ a/hours + financial
work + resource
construction
- - perform.
outages - + + + incentives
- +
- + + + replacement
- availability as
++ + investment
+ proportion of +
aged
load freq. of failure
combined equipment
climate & factor replacement
maintenance + severity criterion
activities - -
+ +
- loading + +
+ replacement of
flexibility reliability equipment
+
+
+ + age as + +
replacement
+ + criterion
+ + interest &
connectivity + + repayment
redundancy
+ system ease of
cost of augmentation + raising + +
land capital debt
investment + cost
+ +
spending - rate of
load
++ - debt
external growth level
factors size of
increasing cost + company
barriers +
per asset - gov. debt -
+ + + support -
special rel. regulatory
rugged + purchasing
+ with vendors support
earthquake terrain efficiencies
zone
+
negative time
positive influence
+ correlation - correlation lag
KEY:

Figure 15 Influence diagram of transmission asset management relationships

3.3.2 Feedback loops in the model

In a complex model like this there are a substantial number of feedback
loops. As a result complex system behaviour can arise. A model description

112
of the CLD form does not contain all the detail necessary for a full
understanding of the model’s behaviour, but it is possible to identify
reinforcing and balancing loops. Reinforcing loops are equivalently known as
positive feedback loops. In a reinforcing loop the action of the loop is to
influence the parameter in the same direction as it is already moving (a
destabilising influence). On the other hand, balancing or negative feedback
loops tend to return the parameter to its initial value (a stabilising influence).
Some feed back loops involving available financial resource are:
available financial resource  + replacement of equipment  +investment
spending - available financial resource

available financial resource  + replacement of equipment  +investment

spending +debt level+interest payments - available financial resource

available financial resource  + maintenance spending  - available

financial resource
The first and third of these could be translated as “if you have more you
spend more”. The first and second loops reflect capital funding from revenue
and capital funding from debt. All three are balancing loops.

available financial resource  delay  +use of online CBM  +use of

predictive maintenance  - failure rate  + maintenance spending  -
available financial resource

This is a reinforcing loop. If available financial resource is increased this

gives an opportunity to increase the use of online CBM, thus increasing use
of predictive maintenance and reducing failure rate. Reducing failure rate is
likely to reduce maintenance costs and therefore increase available resource.
This loop is further reinforced by the following:
available financial resource  + replacement of equipment  + proportion of
equipment installed since 1990 +use of online CBM  +use of predictive
maintenance - failure rate + maintenance spending -available financial
resource

113
Similarly a loop involving augmentation of the network instead of replacement
of equipment also reinforces this loop. There is another loop that directly
connects use of online CBM with reducing maintenance spending. This
connection was added because use of online CBM has a stronger negative
affect on maintenance spending than offline CBM, because of the labour and
resources involved in undertaking off-line CBM. The loop thus formed is also
reinforcing:

available financial resource  + replacement of equipment  + proportion of

equipment installed since 1990 +use of online CBM  - maintenance
spending -available financial resource

There are some important loops relating to maintenance and condition of

equipment:
maintenance spending  +maintenance effectiveness  delay - proportion
of poor condition equipment +failure rate +maintenance spending
This is a balancing loop, but has a big delay, relating to the rate of
deterioration of equipment when maintenance is inadequate. In practice, if
maintenance spending and therefore maintenance level is inadequate then
the proportion of equipment in poor condition will gradually increase, but if
maintenance spending is greater than necessary, more maintenance than
necessary is not likely to result in a lower proportion of poor condition
equipment. The skill with which maintenance is applied, including efficiency
of maintenance, for example use of RCM techniques, will have a large
bearing on the actual relationship between maintenance spending and
maintenance effectiveness.

Other loops allow for the possibility of refurbishing poor condition equipment,
rather than allowing it to fail, or refurbishing it after failure (for example):
maintenance spending  +maintenance effectiveness delay  -proportion
of poor condition equipment  +refurbishment of equipment+ maintenance
spending
These, too, are balancing loops. The delay in these loops means a short
term improvement in balance sheets can be obtained by reducing

114
maintenance spending below adequate levels, and there is some evidence
occurring in practice.

It is interesting to note that, because of the cost of maintaining an item in

good condition is different from the cost of repairing/refurbishing it (or
replacing it) on failure, there may be an optimal level of corrective
maintenance. In other words, for some types of equipment the loops may not
balance back to the same point of maintenance spending. In support of this
argument, in the CIGRE study (CIGRE SC 23 and 39, 2000) some
companies report the use of corrective maintenance as a strategy. In
considering such a strategy, note should be taken not only of the effect on
direct maintenance cost, but the effect on reliability and availability and
through these, the effect on available financial resource:

available financial resource  +maintenance spending  +maintenance

effectiveness  -proportion of poor condition equipment + failure rate -
reliability +available financial resource

The relationship between failure rate and reliability is very system-dependent,

being a function of system redundancy and connectivity and flexibility of
substation configuration. Thus in a network with a high level of built-in
redundancy and/or connectivity and/or flexibility might be able to tolerate a
higher level of failure rate on the network without adverse affect on available
financial resource, than a network with low redundancy/flexibility/connectivity,
and therefore might find it cost effective to use corrective maintenance as a
strategy. These issues will be investigated further in Chapter 6.

3.3.3 Discussion on Chapter 3 and the CLD model

Chapter 3 provides background to the research and makes a contribution to
knowledge in the field by
• Analysis and synthesis of information from a wide range of sources
including expert interviews and personal knowledge as well as
published material;

115
• Construction of a generic model of transmission asset management
with broad application to transmission utilities around the world;
• Filling a gap in knowledge in this field by combining in the analysis
aspects of network design, corporate issues and external influences
on transmission asset management; and
• Representing the system structure concisely in the form of a Causal
Loop Diagram which reveals the loop-structures and possible
behaviours of the system.

116
4 Modelling
In Chapter 3 a model of asset management in transmission enterprises was
developed. However the model needs to be taken further, in order to
address the research topic. The first requirement is to validate the model.
The second consideration is how to extract information from it that
• Can demonstrate why regional variations in asset management exist;
and
• Can be of use to managers of transmission organisations wishing to
identify levers for performance enhancement.
To achieve these purposes it is necessary to implement the model in a form
in which data representing particular cases can be used to predict outcomes
that can then be tested against known results. The model also needs to be
constructed in a way that permits sensitivity of outputs to factors of influence
to be examined. To investigate regional variations regional data should be
used, and for performance enhancement individual case studies could be
employed. The set of relationships expressed in the model apply equally to
individual cases and regional studies.

The time-dependent elements (indicated crudely by delays in the CLD model)

hint that the model might exhibit complex dynamic behaviour if implemented
in the time-domain. This suggests a need also for a deeper analysis,
particularly of core parts of the model.

4.1 Methodology
The methodology developed for this research can be summarised in the
following steps:
1. Define the scope and level of the problem
• Define purpose (Section 1.4)
• Set boundaries (Section1.5)
• Set level of abstraction/detail (Section 1.4, 1.5, Chapter 3)
2. Develop the structure and relationships of the model – conceptual model
(Relationships – Chapter 3, Structure – Section 3.3).

117
3. Identify the characteristics of the problem, data and desired outputs
(Sections 4.2, 4.3);
4. Match problem characteristics and purpose of investigation with suitable
methodologies (Section 4.4);
5. Implement model (fuzzy model Sections 5.1.1, 5.1.2 ; system dynamics
model Sections 6.4 to 6.15)
6. Verify model implementation against conceptual model (fuzzy model
Section 5.1.3; system dynamics model Section 6.2)
7. Validate the model (fuzzy model Sections 5.1.4, 0, 5.2.1 to 5.2.7, and
5.3.1 to 5.3.5; system dynamics model Sections, 6.2, 6.6.2, 6.12, 6.13, 6.15);
8. Refine the model according to results; Repeat steps 6 to 8 as necessary
Section 5.1.6).
9. Use model to investigate the problem. Assess outcomes against
reasonable expectations and limitations of the model. If necessary, repeat
steps 4 to 9 (fuzzy model Sections 5.2.8.2, 5.4; system dynamics model
Sections 6.6 to 6.15).

At this stage the first two steps have been completed. Chapter 5 describes
the implementation, verification, validation and use of a multi-level fuzzy
rule-based model to support the aims of the research project. In Chapter 6
the process from implementation to assessment is repeated for a system
dynamics model, encompassing core parts of the conceptual model
developed in Chapter 3.

4.2 Characteristics of the model

The model described in Chapter 3 and depicted in Figure 15 is a complex
one, with multiple feedback loops, and time dependencies. Some of the
parameters of interest are influenced by a large number of factors -
maintenance spending is a prime example of this. Most of the relationships
are not easily represented in mathematical form – their parameters are
imprecisely known.

Potential input data for use in a study are of variable quality, even from the
CIGRE study. CIGRE study data are highly incomplete in some critical

118
parameters. It is also necessary to supplement data in this study with from
other data sources, making the construction of consistent data profiles a
challenge. The input data are most easily represented as a mixture of
quantitative and non-quantitative values.

4.3 Desired outputs

Since the purpose of this research is to explore the performance of
transmission enterprises in the context of their policies and practices the
desirable outputs include measures of performance and indicators of
policies/practices. Examination of Figure 15 suggests that outputs should be
those elements derived from a range of inputs or other outputs, specifically:
• Reliability;
• Availability;
• Maintenance spending;
• Replacement of equipment;
• Refurbishment of equipment;
• Network augmentation;
• Investment in new equipment;
• Investment spending;
• Use of predictive maintenance;
• Use of online CBM;
• Use of offline CBM;
• Failure rate; and
• Available financial resource.

Available financial resource could be considered either an input or an output.

The quality and quantity of data for comparison of model outputs with values
from real systems is limited. The CIGRE study is a useful source, and has
the advantage that input data profiles and outputs derived from the one
source should be reasonably consistent with each other. However, for some
survey questions the data are highly incomplete, especially when
performance measures or spending figures were requested. Hence
supplementary material is also needed for comparison of outputs.

119
4.4 Potential modelling methods

4.4.1 Fuzzy modelling

The concept of fuzzy sets was introduced by Zadeh in 1965 as a way of
dealing with imprecisely defined classes of objects (Zadeh, 1965). Fuzzy
sets are characterised by a membership function which assigns to each
object a grade of membership within a fuzzy set. Zadeh recognised the
relevance of fuzzy logic to human thinking, particularly in the areas of
information communication and abstraction (Zadeh, 1988) – areas relevant
to the present study.

In a recent project (Ammar et al., 2000) addressing the performance of US

state governments, use of fuzzy modelling was motivated by the “desire to
have a consistent means of applying multi-criteria judgement in a manner
that is replicable and explainable”. This reasoning is equally valid for the
current application. Likewise for both applications, the imprecise nature of
inputs and relationships, the mixture of quantitative and qualitative data, and
the need to apply a consistent rule set to all data suggest the use of fuzzy
rule-based modelling. The government performance model was designed for
the purpose of ranking the financial performance of the states against a set of
criteria and rules defined by experts. By contrast, in this asset management
application the intention is to validate the relationships contained in the model
by applying sets of data, and comparing results against known outcomes. In
other words, it is the expertise domain that is being assessed.

The use of a multi-level system allows decomposition of the problem into

smaller more tractable analysis tasks and facilitates construction and
debugging of the model. Data and rule sets can be constructed separately,
allowing different data sets to be tested against the same rules. This also
allows the possibility of performing sensitivity studies on particular input
parameters.

Most input data for this application can be easily represented as fuzzy
parameters with logical outputs eg attitude to new technology (conservative,

120
moderate, aggressive). Output values are also expressed as linguistic
values eg reliability (high, moderate, low), which can be “defuzzified” to a
numeric on a scale established by the modeller (eg 1- 100). Expert system
shells, some with convenient graphical user interfaces are available for
purchase, and suitable for implementing this type of model.

The disadvantages of this methodology are:

• It does not produce very precise outputs, but the nature of available
data, and the impreciseness of relationships within the model would
suggest that a more precise formulation of outputs is not warranted.
• While it is technically feasible to implement feedback loops in a fuzzy
model, the nature of the conceptual model makes this difficult, if not
impossible. However, time dependencies may be crudely represented
by reference to historical data4.

The advantages of the methodology include:

• The fuzzy model concept is able to implement complex relationships
fairly simply.
• The hierarchical structure of the model facilitates extension of the
model. Parameters initially expressed as constants can be converted
to tasks (calculated values), using other parameters or tasks as inputs.
• The separation of data and model in the fuzzy model methodology
makes it simple to apply different test scenarios. This can facilitate
testing of the model.
• Sensitivity of outputs to variation in inputs can be easily investigated.

4.4.2 System dynamics modelling

The representation of the asset management model as a causal loop
diagram in Chapter 3 suggests the possibility of creating a mathematical
simulation of the model or significant parts of it. System dynamics
methodology is well-established in the area of management science – it was

4
The particular software chosen for implementation was also not conducive to the use of
feedback loops in the model.

121
first proposed for application to strategic management problems in the 1960s
(Forrester, 1961), and has since then found application in many fields (see
Chapter 6 for more information). In system dynamics modelling the behaviour
of the model over time can be explored. Several systems dynamics software
packages exist that enable the model to be expressed graphically, as
combinations of symbols. In Stock-Flow Diagrams (SFD) (Lane, 2000) stock
or reservoir symbols, generally shown as rectangles, represent state
variables and some sort of valve symbol is used to represent rate variables.
Ordinary variables may be indicated by a circle. Underlying the graphical
presentation of the model is a set of differential equations. Solving these
using numerical methods such as Euler and Runge-Kutta techniques permits
the time-development of the model to be explored.

In SFD models relationships must be described much more precisely than in

the CLD form of the diagram. SFD models contain more information but
diagrams are also more cluttered. In the Stella program (High Performance
Systems Inc, 2002), one of the commercially available system dynamics
packages, highly non-linear relationships that are not precisely known can be
represented by graphs.

For the purpose of this research the use of system dynamics modelling can
provide the depth of analysis not achievable with the CLD model. The
impreciseness of the input data and of most relationships in the model means
that simulation parameters would have to be synthesised, and outputs
assessed for reasonableness against expected ranges of values.

The methodology facilitates the use of sensitivity studies which may provide
insights into the relationships represented in the model. It can also be used
to understand model’s limits.

Advantages of this methodology:

• It provides an understanding of the development of the system over
time;
• It aids understanding of the structure of the system;

122
 • Verification is greatly simplified by graphical representation of loops;
• It can provide numeric outcomes. However, implementation of the
system dynamics model requires a high degree of parameter
estimation, and results can be only indicative.
• It is easy to test the sensitivity of model outcomes to changes in levels
of inputs.
Disadvantages of this methodology:
• Changes in decision criteria (eg for replacement of equipment)
require the structure of the model to be changed, a time consuming
process.
• Complexity of full mathematical implementation makes models,
potentially, more difficult to implement and validate.
• Data parameters are buried in the model (although with some
additional work this limitation can be overcome).
• Rules that can be easily expressed in terms of linguistic variables can
be represented as approximate mathematical relationships, but in this
form their validity is more difficult to assess.

4.4.3 Markov analysis

Markov analysis has become popular for modelling reliability and condition-
related phenomena in recent times (eg (Brown et al., 1996, Hoskins et al.,
1999)). The technique (Render and Stair, 1994) assumes that there is a
finite number of states in which something (assets, for instance) can exist.
The probability of changing states is estimated and recorded as a matrix of
transition probabilities. In simple formulations the probability of changing
states remains constant over time. The probability of being in a state in the
next period is estimated by multiplying the vector representing current
probabilities of being in each state by the matrix representing transition
probabilities. The model reaches equilibrium conditions after a number of
iterations; the steady state probabilities are not influenced by the initial state
probabilities. The main advantage of this method is that it can be compactly
coded for computation by computer.

123
The model presented in Figure 15 would not be easily adapted to this form of
representation. It would, however, be possible to represent the core parts of
the model (assets degrading, failing, being replaced, being refurbished) using
this method.

4.4.4 Time-based modelling of the decision-making process

A time-based model of the decision-making process was the methodology
initially considered for use with this research. In it a small part of an actual
asset database used as representative of a real system. Decisions regarding
each substation, bay and piece of equipment were to be made for each year
of the simulation, based on expected failure rates (as a function of age),
replacement criteria and maintenance practices based on those used by a
particular utility in Australia. The effect of these practices on the asset base
and revenue of the utility over time could then be calculated using the
Australian regulatory framework.

This approach was abandoned eventually because of the excessive quantity

of data required, and the difficulty of obtaining data, particularly data relating
to financial aspects of the problem. Other disadvantages included the
difficulty of adapting this methodology to apply to other jurisdictions, and the
fact that it was difficult to include in the model’s scope all elements depicted
in Figure 15.

4.5 Selection of model methodology

From the foregoing discussion it can be seen that while the fourth
methodology outlined proved impracticable the fuzzy model, the system
dynamics and the Markov approaches all provide potential benefit to the
research being undertaken. The fuzzy model can provide the breadth of
approach (covering the range parameters suggested in Chapter 3) while the
other two representation can be used to examine key parts of the model
more deeply. The system dynamics approach has been chosen over the
Markov analysis representation because of the greater flexibility allowed by
use of system dynamics modelling, and excellent software available for
modelling.

124
A multi-level rule-based model employing fuzzy modelling concepts has been
created based on the relationships expressed in Chapter 3. A set of regional
data profiles has been developed and applied to the rule set, permitting
predicted outcomes to be assessed against available performance data.

As an extension of this, a set of case studies has been constructed, based on

actual transmission enterprises, to explore the effects of combined external,
internal, business-related, system-related and environmental factors on asset
management. The regional data sets, as average or typical cases represent
less extreme data ranges than can be expected from individual case studies.
As a result the model has been extended, over the period of the research, to
cover expanded data ranges and include factors pertinent to individual cases.

The system-dynamics modelling has been used to explore selected parts of

the model, identified as critical, from analysis of Figure 15 and from the fuzzy
model. In turn, results from the initial system dynamics model have
prompted improvements to the fuzzy model.

125
 5 The multi-level fuzzy model
This chapter contains five sections. The first section describes the
implementation, testing and verification of the multi-level fuzzy model. The
second and third sections present studies by regional case sets, and case
studies on individual companies. In Section 5.4 the use of the fuzzy model
as a strategic management tool is examined, based on the individual case
studies. Findings are summarised in Section 5.5.

5.1 Development of the multi-level fuzzy model

5.1.1 Implementation of fuzzy modelling

For ease of implementation of the fuzzy model, an expert system shell has
been employed in this investigation (Attar Software, 1999). Most data
represented in the model, as described in Chapter 3, are inherently fuzzy
with logical outputs. Some examples are:
climatic factor: (extreme, moderate or mild)
load growth: (high, moderate, low)
Names used in the model are shown in italics here and in subsequent
chapters. Each fuzzy parameter has a numeric variable associated with it
(for example, climatic factor level), interpreted by the program, according to
the membership functions for the values.

1.0
Membership function

Mild Extreme

Moderate
0.5

0
0 20 40 60 80 100
Climatic factor level
Figure 16 Membership functions for climatic factor

126
 Membership functions for the output values are represented either as
‘normalized’ functions on a 0-100 scale, as shown in Figure 16, or as
membership functions reflecting the actual form of the data (Figure 17).
1.0
Membership function

High

0.5
Low Moderate

0
0 5 10 15 20
Load growth recent in %
Figure 17 The membership functions for parameter load growth– an example of
membership functions estimated from real numeric data, in this case percentage
recent load growth.

Considering Figure 17, for instance, a load growth of 8% would have a MV of

0.6 for high, 0.4 for moderate and 0 for low. It would be described in a report
as high – 8%.
The relationships between data and outputs of the model are recorded as
decision trees, and then converted to fuzzy rules. A fuzzy rule has the form:
IF (condition 1) AND (condition 2) THEN outcome is value 1.
An example of a condition might be
Climatic factor is extreme.
The expert system shell employs a common form of fuzzy logic using
standard fuzzy set operations (Klir and Yuan, 1996). Inference of a fuzzy rule
involves looking up the membership value (MV) of the conditions used in the
rule, taking the minimum of all MVs from conditions and assigning it to the

127
outcome of the rule. When several rules with the same outcome are fired,
the outcome is assigned the maximum MV calculated from all the rules.

Fuzzy rule conditions can contain either attributes (constants) or tasks

(executable procedures). A task must be calculated by set of decisions in
order to obtain a value. In the decision tree format, tasks are nested, using a
technique called back-propagation (Luger and Stubblefield, 1999). As an
example, the very simple decision tree for a task called investment in new
equipment is shown in Figure 18.

system augmentation

high high

medium equipment replacement

high high

medium medium

low medium

low equipment replacement

high high

medium medium

low low

Figure 18 The decision tree for the task Investment in new equipment
This task has three possible outcomes (high, medium, low) only two
components - equipment replacement and system augmentation, both of
which are fuzzy tasks. Investment in new equipment generates only seven
fuzzy rules. These are shown in Table 4.

Table 4 Rules generated for the task investment in new equipment

Rule Conditions Outcome
1 system augmentation is high high
equipment replacement is high AND
2 high
system augmentation is medium

128
 Rule Conditions Outcome
equipment replacement is medium AND
3 medium
system augmentation is medium
equipment replacement is low AND
4 medium
system augmentation is medium
equipment replacement is high AND
5 high
system augmentation is low
equipment replacement is medium AND
6 medium
system augmentation is low
equipment replacement is low AND
7 low
system augmentation is low

As an illustration of how the fuzzy logic works consider the case where
system augmentation = 30, and
equipment replacement = 20.
If the membership graph of system augmentation is similar to Figure 16:
MV (30, low) = 0.66
MV (30, medium) = 0.33
MV (30, high) = 0.
Similarly from the membership graph of equipment replacement:
MV (20, low) = 1
MV (20, medium) = 0
MV(20, high) = 0

To calculate the MV for each outcome of the task investment in new

equipment rules are grouped by outcomes. For each rule in the group the
MV is the minimum MV of its conditions:
For the group of rules with outcome “high”
Rule 1 min (0) = 0
Rule 2 min (0, 0.33) = 0
Rule 5 min (0, 0.66) = 0

For the group of rules with outcome “medium”

Rule 3 min (0, 0.33) = 0
Rule 4 min (1, 0.33) = 0.33
Rule 6 min (0, 0.66) = 0.

For the group of rules with outcome “low”

Rule 7 min (1, 0.66) = 0.66

129
For each outcome, the MV is taken as the maximum MV from the group of
rules having that outcome.
MV (high) = 0
MV (medium) = 0.33
MV (low) = 0.66

To translate fuzzy outcomes back into concrete terms, this implementation of

fuzzy logic places the MV for each fuzzy outcome at the point where the
membership function has its highest value. The calculated outcome is then
the average of membership values weighted by the corresponding highest
value of the membership function.

In this example the highest values for outcomes low, medium and high in
investment in new equipment, are 0 to 20 (an average of 10), 50 and 80 to
100 (average 90) respectively. The outcome investment in new equipment
level associated with this task is:
0.66 x10 + 0.33x50 + 0 x90
level = = 23
0.66 + 0.33 + 0.0
The outcome description is determined from the outcome with the highest
MV, in this case “low”. Both a numeric and a descriptive outcome for results
have been recorded in this model, because the combination provides more
information than either descriptive or numeric result alone.

5.1.2 Model structure

Because of the complexity of the model structure, it was decided to treat the
model as a snapshot in time, thus eliminating the need for a large number of
feedback loops, and treating some parameters as inputs. Available financial
resource, for instance, is modelled as a constant revenue sufficiency– in the
diagram of Figure 19 there are no inputs to this parameter, only arrows
starting from it. The proportion of equipment installed since 1990 is likewise
modelled as a constant (named proportion of young equipment) as is the
proportion of aged equipment.

130
To incorporate the delays, for example between maintenance spending and
proportion of poor condition equipment, historical values of the parameters
have been used: for maintenance effectiveness in past years, historical
maintenance adequacy has been used. Similarly, the historical level of
available financial resource is named historical revenue sufficiency.

Some other modifications, mainly resulting from data considerations are:

• Climate and severity is modelled as two parameters – climate (a

descriptive parameter denoting climate type) and climatic factor a
logical parameter with three values (mild moderate or extreme).
• Instead of rate of load growth, load growth recent is used in the model.
This is a mainly a data issue, as it is much easier to obtain historical
data on load growth than projections.
• Loading is modelled as a task depending on losses and pilfering. This
is also a data issue (see Section 0)
• Financial drivers are modelled as two parameters profit as intention
and maintenance spend as KPI.
• The cumulative effect of (over)loading on condition of equipment has
not been modelled, for lack of data. Loading has instead been
modelled as a direct input to failure rate rather than as an indirect
input through proportion poor condition equipment.

131
 maintenance rel to
geographic + manufacturer
spread skill
attitude to + recommended
technology labour
- cost + +
+ use of offline
maintenance
CBM
proportion of + spending +
+ financial
young
drivers
-
equipment + - +
use of predictive
maintenance
refurbishment
+ -
of equip. --
- load
+ + factor
+ -
+ +
+ historical
+ - + maintenance
-
use of online adequacy

+
CBM
+ proportion poor
-
available
failure rate + condit. equip. financial
+ resource
construction -
+
outages
++ +
- availability replacement
as
historical proportion of investment
revenue +
+ aged
sufficiency equipment freq. of failure
+ replacement
combined climate & criterion
maintenance + severity loading
+
activities
- +
live +
working
+ replacement of
equipment
flexibility

reliability
+
+ +
age as
+ + + replacement
criterion +
redundancy
connectivity +
+ ease of
+ system
+ + raising
cost of augmentation capital
land
investment
+ + rate of load
+ spending growth -
-
factors size of external
increasing cost + + company barriers + debt level
per asset -
+ + special rel. regulatory
+ with vendors support
+ purchasing
earthquake rugged terrain efficiencies
zone
+
negative time
positive influence
+ correlation
- correlation lag
KEY:

Figure 19 Effect of translating model as a snapshot

Some outputs have a dependence on a large number of parameters as

Figure 19 illustrates. Maintenance spending, in particular is a very complex
interaction of many parameters. In the decision tree format, used by
Xpertrule it is very difficult to construct a decision tree with more than five or

132
six inputs, depending on the interaction of input parameters. The decision
tree also becomes difficult to debug. The solution to this is to decompose the
problem into logical sub-tasks. In the case of maintenance spending this
task has been decomposed into five subtasks, as illustrated in Figure 20.

maintenance
spending

+ +
+ + +

routine corrective replacement

refurbishment spending
maintenance maintenance costs as
cost limits
cost cost maintenance
+ +
+
+ + + -
+ replacement
an
+ investment +
+
failure rate equipment
financial
labour replacement + incent.
refurbishment costs replacement
of equipment + + +
- - +
replace on replace on
-
- failure contrib. condition age -
- load & climate
+ -
+ + climate
replace on
failures
+ + type

climatic
proportion
factor
loading aged
equipment

system factors financial

reducing use of predictive
proportion poor + driver
condition maintenance
maintenance maintenance historical
equipment
costs + maintenance
- adequacy
- + +
+
+ use of online use of offline + +
+ + CBM CBM
+
+
+
++
redundancy maint.spend revenue
load factor ++ profit as
as KPI sufficiency
intention

connectivity
+
flexibility attitude to
skill Key: positive negative
technology
attribute correlation correlation

+ -
proportion historical
maintenance rel geographic task/
young revenue
to recommended factor subtask
equipment sufficiency

Figure 20 Actual structure of maintenance spending, from fuzzy model. (Note

equipment replacement and loading subtasks have not been elaborated, for reasons
of diagram clarity).
The overall structure of the fuzzy model is summarised in Table 5.

133
 Table 5 Summary of model tasks and their inputs
Task / Subtask Name Parameters & subtasks (subtasks bolded)
construction outages; failure rate; flexibility; maintenance
availability
actions increasing availability
construction outages refurbishment, equipment replacement; system augmentation
corrective maintenance costs labour costs; failure rate
cost of finance credit rating
debt repayments cost of finance; revenue sufficiency; debt level
regulatory support; debt level; revenue sufficiency; credit rating
ease of raising capital
known; credit rating
replacement on condition; replacement on failures;
equipment replacement
replacement on age; financial incentive- replacement
use of online CBM; ruggedness of terrain; cost of land;
factors increasing asset costs
earthquake zone; geographic factor
spending limit; use of predictive maintenance; failure
failure rate contribution of load & climate; proportion poor condition
equipment
failure contribution of load & climate climate (type), climatic factor, loading
financial driver - maintenance maintenance spend as KPI; profit as intention; revenue sufficiency
financial incentive - replacement ease of raising capital; revenue sufficiency, debt repayments
factors increasing asset costs; purchasing efficiencies;
investment cost
investment in new equipment
investment in new equipment equipment replacement; system augmentation
loading pilfering, losses
use of predictive maintenance; use of online; live work;
maintenance actions increasing availability
combined maintenance activities
corrective maintenance costs; refurbishment costs;
maintenance spending replacement costs as maintenance; routine maintenance
costs; spending limit
proportion poor condition equipment historical maintenance adequacy; proportion aged equipment
special relationship with vendor; investment in new equipment;
purchasing efficiencies
size of company
financial driver maintenance; use of predictive maintenance;
refurbishment
proportion poor condition equipment
refurbishment costs labour cost, refurbishment
reliability redundancy; connectivity factor; failure rate; flexibility
replacement on age age as replacement criterion; proportion aged equipment
use of predictive maintenance; proportion poor condition
replacement on condition
equipment; refurbishment
number of failures as replacement criterion; frequency of failures
replacement on failures
as replacement criterion; failure rate; refurbishment
equipment replacement; replacement as investment; financial
replacement cost as maintenance
driver - maintenance
use of online CBM; geographic factor; maintenance relative to
routine maintenance cost recommended; labour cost; system factors reducing
maintenance cost
spending limits financial driver maintenance; revenue sufficiency
ease of raising capital; external barriers; load factor; loading ;
system augmentation
load growth recent; investment cost as KPI; revenue sufficiency
system factors reducing maintenance costs connectivity factor; flexibility; load factor; redundancy
labour skill; profit as intention; maintenance spending as KPI;
use of offline CBM
revenue sufficiency; attitude to new technology; labour cost
proportion young equipment; attitude to new technology;
use of online CBM
geographic factor; historical revenue sufficiency
use of predictive maintenance use of online CBM; use of offline CBM

5.1.3 Verification of the fuzzy model

Verification is the process of assessing that the model adequately represents
the conceptual model from which it has been derived. There are two parts to
verification for this model. The first requirement, verification of the structure
has been readily achieved using the facilities of the software package. The
attributes and sub-tasks potentially contributing to calculation of a task can

134
be displayed in the attributes view for the task. Alternatively all instances in
which a task or attribute is used, can be listed in a usage report. The linkage
of tasks may be examined on the map view.

The second step is to verify the rules associated with each task. This is a far
more time-consuming process. It entails ensuring that the rules in the model
are in accordance with the relationships presented in Chapter 3.
For simple tasks (such as that illustrated in Figure 18) it is possible to check
the decision tree visually. For tasks that have many inputs or for which the
relationships between variables are complex, visual scanning of decision
trees is not practical. Two approaches have been taken to overcome this
problem. One solution has been to structure the problem as a multi-level
task, by decomposing it into more manageable sub-tasks. An example of
this was given in Figure 20, with regard to maintenance spending.

The second, complementary approach has been to write a procedure to test

the value of outputs against two inputs across nominated ranges. For
parameters other than those being tested, the user nominates a base case
from one of the existing data sets. The results are displayed as matrices of
output values, and can be plotted as contour maps (using a spreadsheet
program). Using this visual display it is often possible to detect incorrect
rules by unexpected “bumps” or “troughs” in output values as the value of an
input parameter is increased across its range.

One drawback of this method, arising from a limitation in the Xpertrule

program, is that the facility only works with parameters that are attributes
(constants). For parameters that are tasks the program always executes the
task the first time through. This can make the debugging process more
arduous. For instance, if one wanted to test an output against two inputs that
were tasks, one way around this problem to save a copy of the file, convert
the tasks under investigation back to attributes (losing the substructure of the
model), reconfigure the arrays containing attribute names, and then run the
test, a tedious process. Alternatively, it is possible to change the status of
the task from dynamic to static (so that it does not execute after the first

135
loop), run the program through one iteration of the loop then run the whole
range of values to obtain the matrix of outputs, as required.
The development cycle works most efficiently if the model is constructed and
verified iteratively as follows:
• A task is created first using attributes as inputs.
• The task is tested by varying its attributes over their appropriate
ranges.
• Then the attributes are converted to tasks, and additional decision
trees are added at the deeper level, and so on.
However, in practice it is not always possible to construct the model in
this way, because of the highly interconnected nature of the system being
modelled, and also because one’s understanding of the system changes
gradually over time as more experience is gained with the model.

5.1.4 Validation of the fuzzy model

Validation is the process of proving that a model correctly represents the
system under study. It also encompasses a concept of fitness for purpose.
For Management Science and allied applications it is seldom possible to
validate models fully against real systems, in part because it is impossible to
control all variables to real systems. Validation, therefore, becomes an
exercise in building confidence in the model’s usefulness in explaining
observed behaviour or predicting outcomes.

The process of validation generally falls into one of two categories (Pidd,
1996): black box or transparent box testing. In transparent box techniques
the internal working of the model are examined. In black box testing the
outputs of the model are assessed against a range of inputs, ignoring the
internal structure of the model.

For this application the black box testing has been performed by creating a
set of data profiles for different regions of the world, and comparing the
results against the best available data on performance and policies. The
rationale for this is that all datasets are tested against the same set of rules,

136
and if the model is valid then all datasets should give outputs consistent with
available data. There are drawbacks with this approach:
• Both the input data and the outputs for comparison are less than
perfect - incomplete, inconsistent, not always corresponding to the
same set of companies or countries or the same time period. This
makes it unlikely that a perfect match between inputs and results will
be achieved, even if the model is perfectly correct.
• The regional data tends to be very “average”, because they are
created from averaged or typical data. They therefore tend not to test
the rules at the extreme values of data ranges.
Additional testing with individual case studies engages a larger proportion of
model rules, although some data have also had to be estimated for these
cases. However, sufficient data to create sets for individual case studies
have not come to hand for some of the more extreme cases that could
potentially be applied to this model.

5.1.5 Summary of data used in regional studies for the fuzzy model
Appendix A details the sourcing and treatment of data for regional case
studies. These data inputs are summarised in Table 6 .

137
 Table 6 Input Parameters used in regional studies for fuzzy model
Parameter A AU EE NA SA WE
Age as replacement criterion very significant (100) insignificant (25) very significant (82) significant (42) significant (48) significant (42)
Attitude to technology moderate (40) aggressive (65) moderate (60) moderate (35) moderate (45) moderate (60)
Business acumen moderate (50) moderate (50) low (25) moderate (60) moderate (50) moderate (60)
Climate tropical & wet sub-tropical temperate temperate tropical & wet temperate
Climatic factor moderate (40) moderate (50) extreme (65) mild (30) moderate (60) moderate (60)
Combining of maintenance
medium (44) medium (63) low (33) high (71) low (0) medium (57)
activities (by circuit)
Connectivity factor low (14) moderate (20) moderate (21) moderate (20) low (14) high (32)
Cost of land low (30) low (30) moderate (40) moderate (50) low (20) moderate (50)
Credit rating A (40) A (50) BB (30) A (40) BB (30) AA (70)
Debt level moderate (60) moderate (50) moderate (40) moderate (45) high (75) moderate (55)
Earth quake zone No (30) No (10) No (10) No (20) No (20) No (20)
External barriers low (20) moderate (40) moderate (50) moderate (60) low (30) high (80)
Flexibility high (84) moderate (43 high (67) high (69) high (90) low (31)
Frequency of failure as
significant (57) very significant (77) significant (50) significant (58) significant (60) significant (60)
replacement criterion
Geographic factor moderate (16) moderate (21) low (11) moderate (17) high (39) low (6)
Historical revenue sufficiency adequate (65) not a concern (80) inadequate (40) not a concern (75) inadequate (40) not a concern (80)
Investment cost as KPI significant (55) very significant (75) insignificant (33) insignificant (26) insignificant (0) insignificant (30)
medium $US 19000
Labour cost low $US 9000 pa low $US 2000 pa high $US 29000 pa low $US 5000 pa high $US 26000 pa
pa
Labour skill moderate (50) high (70) high (70) high (70) moderate (50) high 70
Live work low 24% low 20% low 11% low 11% moderate 57% low 3%
Load factor moderate (57%) moderate 62% moderate 63% moderate 61% high 72% high 65%
Load growth (recent) high (9.1%) moderate 3.5% low 1.0% low 2.7% moderate 7.4% low 2.9%
Losses (T&D) moderate (11%) moderate 9% very high 18% low 7% high 15% low 7%
Historical maintenance
maintained (65) maintained (65) under-maintained (30) maintained (65) under-maintained (25) maintained (65)
adequacy
Maintenance relative to
more than at recommended more than less than more than at recommended
manufacturers’ recommended
recommended (85) levels (56) recommended (100) recommended (15) recommended (80) levels (45)
levels
Maintenance spending as KPI significant (55) very significant (75) insignificant (18) very significant (72) insignificant (20) significant (53)
Number of failures as
very significant (67) significant (62) significant (50) significant (58) significant (40) significant (58)
replacement criterion
Pilfering low (20) low (10) high (70) low (10) low (30) low (10)
Profit as intention No (36) No (38) No (0) Yes (58) No (20) No (28)
Proportion of aged equipment low 6% low 10% low 3% moderate 14% low 6% low 10%
Proportion of young
high 65% moderate 37% moderate 56% low 30% moderate 62% moderate 39%
equipment

138
Parameter A AU EE NA SA WE
Redundancy moderate (n-1) (50) moderate (n-1) (50) moderate (n-1) (50) moderate (n-1) (50) moderate (n-1) (50) moderate (n-1) (50)
reasonably supportive reasonably supportive reasonably supportive
Regulatory support highly supportive (70) highly supportive (50) highly supportive (70)
(35) (78) (40)
Replacement considered an
very significant (100) very significant (100) very significant (83) very significant (86) very significant (80) very significant (100)
investment
Revenue sufficiency
adequate (60) not a concern (70) adequate (50) not a concern (70) adequate (55) not a concern (80)
(available financial resource)
Ruggedness of terrain not rugged (30) not rugged (30) not rugged (30) not rugged (30) not rugged (30) not rugged (30)
Size of company medium (40) medium (25) small (30) medium (50) medium (20) medium (50)
Special relationship with
no use (30) no use (35) no use (10) no use (30) no use (20) no use (35)
vendors

139
5.1.6 Refinement of the fuzzy model
The refinement and validation processes are closely linked, as explained in
Section 4.1. During validation, one or more cases might produce an outcome
that is considered too high or too low compared with outputs from other
cases. There are a number of options for correcting such problems:

• The easiest solution is to adjust the importance of a particular

relationship by making minor changes to one or more rules. This
involves tracing through the decision tree for the particular task using
one or more data sets to identify the rule needing adjustment, then
retesting (all affected data sets) to check that the problem is rectified.
Because of the inter-relationships between tasks it is also necessary
to check that other outputs have not been adversely affected. The
process is complicated by the fact that in a fuzzy model multiple rules
will be invoked, so multiple rule paths may have to be considered. It is
common, however, that one path dominates.
• If the problem is one of inadequate resolution, the number of output
levels for the task can be increased. This requires increasing the
number of rules and thus the complexity of the decision tree
associated with the task. Verification of the rules is necessary
following these changes.
• Sometimes the structure of subtasks or the rules combining subtasks
can be altered slightly to change the outputs. These changes also
need to be tested against all affected datasets.
• If an additional factor of importance is identified, this can be added to
the model. Adding parameters will also change the structure of the
conceptual model, as well as the fuzzy model.
• In some cases the correspondence of predictions of outputs with other
data is limited by the quality of either the input data or the data with
which the outputs are compared. Little can be done about this
problem.

140
 • Finally, the inability of the fuzzy model to represent fully the feedback
mechanisms and time delays of the model is recognised as a limitation
of the technique.

The results from the validation and refinement processes are presented in
Sections 5.2 and 5.3 for regional studies and individual case studies
respectively.

5.1.7 Summary of modelling methodology and implementation of the

fuzzy model

In this section the overall methodology for developing, verifying and

validating models of strategic asset management is presented. This
methodology is relevant to a broad range of applications in engineering asset
management. The particular application addressed in this research has been
shown to possess characteristics that make it unsuited to traditional
engineering modelling solutions. It has therefore been necessary to draw on
techniques from Management science and Artificial intelligence for the
implementation of the model. Application of such techniques to the asset
management in electricity transmission enterprises represents a contribution
to knowledge in this field.

The methods used for verification, validation and refinement of the multi-level
fuzzy rule-based model, have been described in detail in this section,
including the development of regional case studies for use as input data.

5.2 Results – Regional Comparisons using Fuzzy Modelling

In this section, results from various model outputs are compared with
available data. The regional cases provide the input data. The model
outputs are compared with data from the CIGRE study and other sources.
Not all model outputs can be tested against other data – data are not
available. However, since there is a significant degree of interrelatedness
between outputs, testing of a subset of significant outputs gives a measure of
confidence in the validity of the model as a whole.

141
5.2.1 Failure rate
Failure rate is a key factor in the transmission asset management model. It
impacts on many other model outputs – including reliability, availability,
maintenance costs and replacement costs. It is therefore important that the
values obtained in the calculation of failure rate accurately reflect observed
values.

Figure 21 shows a comparison of unplanned outages per circuit and per 100
circuit-kilometres from the CIGRE study and the failure rate output parameter
from the model. The differences between the per-circuit and per-100km-of-
circuit values reflect the fact that some failures are related to the length of
circuit (lightning flashovers for example) while some are not related to circuit
length (substation equipment failures for instance). For transmission
systems with, on average, low circuit lengths (WE and EE), the unplanned
outages per circuit will give better measure of forced outage rate. On the
other hand, for systems with greater average line length (SA, notably), the
unplanned outages per 100km circuit length should be a more accurate
reflection of outage rates.

4
low moderate high.

Unplanned
3 outages
per100km
2
Unplanned
1 outages/cct

0 model
A AU EE NA SA WE
Regions

Figure 21 Comparison of unplanned outages per 100km of circuit and unplanned

outages per circuit from CIGRE study with failure rate from model.
In systems with circuits of moderate length, the “true” outage rate should be
somewhere between the two measures. Taking these facts into account the

142
model predicts results similar to actual values, although it underestimates
slightly the failure rate for South America. The reason for this is not entirely
clear – although comparison with the maintenance cost graph suggests that
either the system is under-maintained, which over time would result in higher
failure rates, or that the climatic severity factor in the model is too low, in
which case the model overestimates the additional corrective maintenance
costs associated with a high failure rate.5 Alternatively it may be a matter of
poor data quality (in maintenance costs and/or failure rates).

5.2.2 Reliability of supply

Reliability is an important measure of transmission system performance.
However, comparison of reliability suffers from the wide variety of definitions
used for calculation of transmission reliability – failures per unit time, failures
per unit time with consequence for the customer, unsupplied energy, system
minutes lost, to name a few. Even slight differences in definition can make
comparison difficult. Common sources of variations in definition include
minimum duration of outage counted in the total, the boundary between
transmission and distribution systems, or even between transmission system
and connection assets. At best a rough estimate of the level of reliability can
be made. Fortunately this is similar to the output from the fuzzy model. Table
7 lists some of the definitions and values, and associates with each an
assessment.

Table 7 Reliability measures from various parts of the world

Region Location Value Description Source Assessment
Average interruption duration index
Minutes/customer/year 1987-1996 (Kurihara,
Japan <20 Very high
except 1994 (transmission + 1999)
distribution)
A
Minutes/customer in 1997, down from
(Singh and
Malaysia 300 1000 minutes /customer in 1995
Poh, 1997)
(T,D,G) Moderate
(Mohd Noor,
Malaysia 17 System minutes lost (T) 2001
2002)
System minutes lost (T) 1997-98 (Ministry of
New South
AU 1.21 Energy and High
Wales
Tasmanian result includes equipment Utilities, 1999)
down to 11kV. (Office of the
Tasmanian
Victoria 0 High
Regulator,
2001)

5
Note that there were only three responses in the CIGRE study from South America
providing the required information. These responses may not be typical of the region as a
whole.

143
Region Location Value Description Source Assessment
Queensland 3.8 (Brown, 2001a) High
South
2.4 (SAIIR, 2001) High
Australia
(Office of the
Western Tasmanian
5 High
Australia Regulator,
2001)
(Office of the
Tasmanian Moderate-
Tasmania 17
Regulator, high
2001)
Years 1997/98 and 1998/99. Number
24 moderate-
AU Queensland of unplanned outages causing loss of (ACCC, 2002b)
20 high
supply per year. (T)
5.3 (Plantagie,
AU South Africa System minutes lost in 1998,1999 (T) High
3.8 2000)
Average system reliability from four (CIGRE SC 23
AU unspecified 99.5% High
responses in CIGRE study (T) and 39, 2000)
system reliability 1 respondent from (CIGRE SC 23
EE unspecified 99.7% High
CIGRE study (T) and 39, 2000)
value assumed to be unreliability
(CIGRE SC 23
EE unspecified 10% (target was 5%) – 1 respondent from Low
and 39, 2000)
CIGRE study (T)
min/customer/yr average interruption
Consolidated (Kurihara,
NA <10 duration index (transmission + High
Edison NY 1999)
distribution) in period 1990-1996
1988 – 1994
1995 -1996
Florida Light & 40-60 (Kurihara, high-
NA min/customer/yr average interruption
Power 120 1999) moderate
duration index (transmission +
distribution)
1989-1996
Southern
approx min/customer/yr average interruption (Kurihara, high-
NA California
38-78 duration index (transmission + 1999) moderate
Edison
distribution)
(Hours) customer average interruption (BC Hydro,
NA BC Hydro 2.57 Moderate
duration index (T,D) in 2001 2002)
(Hours) service average interruption (Hydro Quebec
HQ high-
NA 0.51 duration index in 2001 (T). Excludes Transenergie,
Transenergie moderate
exceptional events. 2002)
Data for 97/998 and 98/99.
Southern Percentage of 500kV lines without
57.9%
NA California forced outages. (Note this is equipment (ACCC, 2002b)
52.6%
Edison reliability rather than reliability of
supply)
Data for 97/98 and 98/99. Percentage
Southern
74.6 of 220kV lines without forced outages.
NA California (ACCC, 2002b)
81.70 (Note this is equipment reliability rather
Edison
than reliability of supply)

144
Region Location Value Description Source Assessment
97/98 and 98/99 500kV Annual forced
Southern
0.63 outage frequency (number per line per
NA California (ACCC, 2002b)
1.0 year) (Note this is equipment reliability
Edison
rather than reliability of supply)
97/98 and 98/99 220kV Annual forced
Southern outage frequency (number per circuit
0.7
NA California per year) (Note this is equipment (ACCC, 2002b)
0.48
Edison reliability rather than reliability of
supply)
Transener, High-
SA 0.5-1.5
Argentina Forced outages per 100 circuit-km per moderate
Regional year (Note this is an indicator of
(ENRE, 2001)
transmission Typically equipment reliability rather than supply Moderate-
SA
boards, 4 reliability) low
Argentina
Average value from six responses for (CIGRE SC 23
WE Unspecified 99.4% High
system reliability from CIGRE study and 39, 2000)
14 Figures for 1997/98 and 1998/99.
WE NGC, UK
5 Number of incidents on the
Scottish 6 transmission system resulting in loss of
Power, UK 9
supply to one or more customers from
(OFGEM,
faults on the transmission system. High
2001)
These include faults on connection
Hydro-electric, 8
UK 19 assets where the loss of supply
resulted from customer choice of
connection arrangement.
Unplanned unavailability of circuits
98/99 and 99/00. Forced outages due
to plant breakdown or forced
0.16%
WE NGC maintenance taken with less than one (NGC, 2000) High
0.26%
week notice. (Note this is a measure
of equipment reliability not supply
reliability)

The CIGRE study data also are highly incomplete and, in a number of cases,
figures were reported without explanation of the method of calculation
employed. Six of 28 respondents from Western Europe gave percentage
values for system reliability (not supply reliability), using Equation (5).

Reliability = 1 –
1/(no. ccts) * ∑ (hours of unplanned availability) (5)
(8760 – hours of planned unavailability)

In Australia one favoured measure for supply point reliability is system

minutes lost (Equation 6):

145
 Sys minutes lost =
unsupplied energy *60/peak demand (6)

Unsupplied energy is in MWh and peak demand in MW. Note that supply
point reliability and system reliability defined in Equation (5) are not
equivalent (one does not include impact on customer, the other does), but
are related.

In the model an attempt has been made to predict supply point reliability
based on system characteristics and asset management policies and
practices (including historical ones). The results by region (Figure 22), when
compared with Table 7, are reasonable but more data for comparison,
especially from South America, Eastern Europe and Asia would be beneficial.
Reliability, as seen by customers, is strongly influenced by network
configuration, especially redundancy on the network. Typically transmission
systems are designed to withstand common outages without loss of supply to
the customer (N-1 planning criterion), so loss of supply incidents are
infrequent, even on systems with moderately high equipment outage rates.
Highly interconnected systems will have higher reliability, all else being
equal, because of a higher number of alternative paths for supply. The
cumulative effect of connectivity and redundancy is illustrated in two sets of
figures (Table 8) reported in (ACCC, 2002b).

Table 8 Effect of network configuration on supply reliability

Annual total of unplanned Annual total of unplanned
Year outages outages causing loss of supply
(number per year) (number per year)
1996/7 201 19
1997/8 219 24
1998/9 165 20
1999/2000 173 20

The reliability of supply for this utility is eight to ten times greater than it would
be if each load was supplied radially through a single line.

146
 Reliability of Supply-Model

low medium high

A AU EE NA SA WE

Figure 22 Supply point reliability predicted by the model. The arrows indicate the
range of observed values from Table 15.

5.2.3 Availability of circuits

Comparison of availability also suffers from the variety of definitions used,
although perhaps to a lesser extent than reliability. Availability of supply can
be measured, but availability of circuits is a more common measure. Table 9
collates circuit availability data from a range of sources and provides an
assessment of the values obtained.

Table 9 Circuit availability

Region Location Value Description Source Assessment
(Power Grid
Power Grid
98.85% Circuit availability 2000/01 Corporation, high
Corp. India
2001)
A
99.57%
97% four responses (of possible (CIGRE SC 23 1x moderate
unspecified
100% nine) for circuit availability and 39, 2000) 3 x high
99.47%
99.23%
South Australia 99.25%
99.67%
Queensland 99.60
99.54%
Data for 97/98, 98/99 and
Victoria 99.58%
99/00. Queensland 99/00 only.
99.69%
Tasmania and Snowy Hydro
99.67% (ACCC, 2002b)
Snowy Hydro 98/99 and 99/00 only.
AU 98.63% all high
99.13% Availability of circuits in %.
Tasmania
99.17%
99.54%
New South
99.37%
Wales
99.42%
99.10%
New Zealand 99.10%
99.20%

147
Region Location Value Description Source Assessment
95.5% 1x low
three responses (of possible (CIGRE SC 23
EE Unspecified 60.0% 1x moderate
six) for circuit availability and 39, 2000)
99.0% 1x high
Southern
NA California 99.99% Circuit availability (ACCC, 2002b) high
Edison
97.78%
three responses (of possible (CIGRE SC 23 1 x moderate 2
NA Unspecified 99.9%
seven) for circuit availability and 39, 2000) x high
100%
99%
three responses (of possible (CIGRE SC 23 1 x moderate
SA Unspecified 95%
five) for circuit availability and 39, 2000) 2x high
99.5%
Unavailability of circuits
1997/8. Includes large
NGC 3.7% (OFGEM, 2001) moderate
components from construction
outages
Scottish Power 3.3% moderate
Unavailability/Availability of the
Hydro-Electric 2.5% (Perez-Arriaga et moderate
transmission network
Finland 99.999% al., 2002) high
99.8%
95.84%
99%
WE 98.5%
98%
99.2%
1 x low
90% Fourteen responses (of (CIGRE SC 23
Unspecified 5 x moderate
98.2% possible twenty eight) and 39, 2000)
8 x high
97.48%
97%
98.79%
99.02%
98%
98.99%

The CIGRE study reports circuit availability average of 98% excluding one
respondent for whom there were special reasons for low availability. Table
10 compares the model outputs with average values from CIGRE data and
the range of values from all sources. Predicted results agree well with
reported data for circuit availability across all regions.

Table 10 Availability – Model predictions compared with actual data

Region Model Average (CIGRE data) Range (all data)
A High (70) High Moderate-High
AU High (85) High High-High
EE Moderate (32) Moderate Low-High
NA High (85) High Moderate-High
SA Moderate (50) Moderate Moderate-High
WE Moderate (62) Moderate Low-High

5.2.4 Use of predictive maintenance

The CIGRE study provides some information about the use of different
maintenance strategies: a comparison of level of usage of different
maintenance strategies was presented in Figure 1.

148
As mentioned in Section 2.1.1, the figure suggests that Australasia/South
Africa, and South America have the highest use of online condition-based
maintenance. Australasia/South Africa, Eastern Europe and Western Europe
have relatively higher use of offline CBM; North American respondents report
relatively low usage. High predictive maintenance usage (combining online
and offline CBM) should also be associated with low corrective maintenance
and vice-versa, if the equipment is in similar condition. This relationship
seems to hold true for Asia and North America which have relatively high
corrective and lower predictive maintenance, and for Australasia/ South
Africa with low usage of corrective and higher predictive maintenance. The
South American data are more ambiguous, but perhaps the relatively high
level of corrective maintenance is an indication of historically poor
maintenance quality rather than low level of predictive maintenance in 1998.
The relatively high level of corrective maintenance employed as a strategy
might be a reflection of spending limitations and poor maintenance history.
Table 11 compares the predictions of the model with data from the CIGRE
study.

The predicted level of online CBM in South America is lower than that
suggested by the CIGRE study. The use of a particular strategy can be
considered to be a function of need, inclination and opportunity. The use of
online CBM may be motivated by a need to maintain a high level of circuit
availability. (See Section 6.14.1 for a discussion on the effect of redundancy
on system risk, and the need to maintain high availability.) Note that two
South American respondents to the CIGRE study also report high levels of
live-working; live working is also aimed at increasing circuit availability. The
rapid growth in electricity consumption in South America provides the
opportunity to introduce online CBM, as online condition-monitoring is
cheaper to purchase with new equipment than to retrofit to aged equipment.
However, opportunity may be limited by available financial resource. For this
reason the model does not predict high use of online CBM in South America.
However, it may be that the model overestimates the capital cost of “typical”
online CBM or that the types of online CBM practiced in South America are
less capital intensive than the self-monitoring devices favoured in some other

149
countries (eg Australia). If so, then there may be greater opportunity for use
of online predictive maintenance in South America than suggested by this
model.

Table 11 Comparison of model predictions with reported maintenance strategies

Offline CBM Online CBM
Region Model CIGRE study Model CIGRE study
A moderate (37) low-moderate moderate (32) low
AU moderate (66) moderate moderate (33) low-moderate
EE moderate (50) moderate low (10) low
NA moderate (30) low low (26) low
SA moderate (43) low–moderate low (10) low-moderate
WE moderate (57) moderate low (24) low

The North American entry for Offline CBM is surprisingly low. Inspection of
the CIGRE data reveals that one respondent indicated a time-based
maintenance strategy only for the protection equipment category (a
regulatory requirement), and no entries in any other category. Another two
nominated time-based and time-based plus corrective maintenance as
general strategies. If these results are typical of North American companies,
the level of skill applied to maintenance may be lower than estimated and/or
the attitude to technology more conservative than estimated in input data for
this model.

150
 100
50-60 90
80
40-50
x 70
30-40 60 Labour
20-30 50 skill level
40
10-20
30
0-10 20
10 10

30

50

70

90
Attitude to new
technology

Figure 23 Sensitivity diagram of offline CBM with attitude to new technology and
labour skill for the North American region. X marks the coordinates used for the
North American base case.

Figure 23 is a contour plot of offline CBM showing sensitivity to two

parameters, labour skill and attitude to technology. When the contours are
closely spaced the output is sensitive to changes in input parameters. The
output is displayed as numbers in the range 0 to 60 in this case,
corresponding to low to moderate levels of offline CBM. The graph suggests
that for the North American case use of offline CBM is not very sensitive to
labour skill around the point chosen (no lines are crosses as the point X is
moved vertically). However it is sensitive to attitude to technology – for
instance a decrease in attitude to new technology level from 40 to 30 reduces
the level of use of offline CBM from 34 to 26 (moving horizontally towards the
left on the plot) which is in the low part of the range. (The level of attitude to
new technology in the base case is 35, as the figure illustrates.) This
suggests that some utilities in North America are fairly conservative about
their maintenance practices, preferring to stay with “tried and true” time-
based strategies rather than using condition-based methods. Whether this
represents the typical situation is difficult to prove, as the CIGRE survey
contained only seven responses from North America. Nevertheless,

151
discussions by the author with some utility representatives in the US appear
to support this argument.

5.2.5 Maintenance spending

In the CIGRE study a size parameter was developed for comparison of data.
This parameter comprised Part 1 based on load and Part 2 based on asset
value. For maintenance costs, comparison on the basis of asset value is
more appropriate than comparison on a parameter that includes load, but the
asset value data in the CIGRE survey is very incomplete.
Normalised maintenance

Maintenance spending -
8 median

high
cost / asset value

model
component

moderate

model
4
2
0 low

A AU EE NA SA WE
Region

Figure 24 Maintenance spending predicted by model output compared with CIGRE

study data

As a compromise solution the Part 2 of the size calculation, which

approximates to asset value, was used as a comparator for the CIGRE data
shown in Figure 24. The Part 2 calculation is shown in Equation 7:
Part 2 = 4/7*[(RatioB * Bays(>345kV)) + (RatioT*Tower(>349kV)) + RatioL*OHL(>349kV)+
RatioC * Cable(>345kV))] +
2/7*[(RatioB * Bays(200-349kV) + (RatioT*Tower(200-349kV)) + RatioL*OHL(200-349kV)+
RatioC * Cable(200-349kV))] +

152
1/7*[(RatioB * Bays(50-199kV) + (RatioT*Tower(50-199kV)) + RatioL*OHL(50-199kV)+
RatioC * Cable(50-199kV))] (7)
where
RatioB = 1; Ratio T = 0.27; Ratio L = 0.053; RatioC = 2.22
Bay(n-m kV) is the number of bays in the nominated voltage range n to m kV
Tower (n-m kV) is the number of towers in the nominated voltage range n to m kV
OHL( n-m kV) is the length of overhead line in the nominated voltage range n to m kV and
Cable(n-m kV) is the total length of underground cable in the nominated voltage range n to m
kV.
The ratios were calculated using estimated initial costs of the equipment and
associated maintenance costs discounted over 40 years to give a net present
value(CIGRE SC 23 and 39, 2000 p A2-2). The values calculated using this
base are normalised to a scale of 1 to 100 for comparison purposes.

As Figure 24 shows, there is reasonable correspondence between predicted

values from the model and results from the CIGRE survey. There is a
discrepancy between the model and data for the South American case
However, the South American data consist of only two responses (of a
possible five in the survey, and many more not represented in the survey).
Other data from the CIGRE study suggest that these values may not be very
accurate:
• Outage rates are higher than average – which means that corrective
maintenance costs will be high;
• 4 of 5 respondents report doing more maintenance than
recommended by manufacturers – which is likely to affect routine
maintenance costs;
• Reported numbers of employees engaged in maintenance activities
are high for the size of the utilities.

153
5.2.6 Refurbishment of equipment

Refurbishment costs
maintenance cost x
normalised maint

low medium high

cost/asset value
Refurb % of 6

- model
4

0
A AU EE NA SA WE
Regions

Figure 25 Comparison of refurbishment as a percentage of direct maintenance costs

(from CIGRE study) with model predictions for refurbishment levels

The CIGRE study reports refurbishment as a percentage of maintenance

cost. However maintenance costs vary considerably between regions. In
Figure 25 the percentage of maintenance cost attributed to refurbishment is
multiplied by the normalised maintenance cost per asset value for each
respondent to obtain refurbishment cost per asset value; the results are then
averaged over the region. (The normalised maintenance cost per asset
value is the same as that used in the previous section.) These figures are
compared with the fuzzy model’s estimates of refurbishment costs, based on
refurbishment level and labour cost.

Both CIGRE survey and model results suggest that Eastern Europe has
relatively high levels of refurbishment costs and North America relatively low.
The CIGRE data suggest that Australasia/South Africa and Western Europe
spend more on refurbishment that North America; this is also consistent with
model outputs

154
5.2.7 Replacement of equipment
Figure 26 illustrates model predictions and CIGRE data for replacement rate.
The CIGRE survey asked for a general answer for percentage of equipment
replaced in the past four years, and /or the same information for seven
categories of equipment. To obtain results for comparison it was necessary
to make some assumptions about the data. The rules used are as follows: If
a general answer were given this was used, otherwise an average value over
the seven categories was taken. Some respondents gave no answer at all.
It was assumed when averaging the seven categories that if some categories
had non-zero numbers and others had none that the empty categories were
intended to be zeros. In some cases this may make the averages lower than
true figures. The use of average values across all categories will all produce
some distortion of results, unless all categories contain equal number of
items.

10.0
Median all categories or

Replacement -model
low moderate high
8.0
general (%)

6.0

4.0

2.0

0.0
A AU EE NA SA WE
Median Model

Figure 26 Replacement of equipment - predicted rate from model compared with

CIGRE Study data

155
The model appears to overestimate the level of replacement in South
America, underestimate replacement in Eastern Europe/former Soviet Block
countries and Western Europe.

In the model the estimation for level of replacement is based on the

proportion of poor condition equipment, the percentage of aged equipment
(equipment greater than 40 years of age), replacement criteria used –
condition-based, age-based, failure-rate or number of failures-based together
with level estimated for refurbishment of equipment. Investigation of
replacement of assets by category (Figure 5) shows for Eastern Europe
protection equipment dominates (median replacement level for this category
is nearly three times greater than for all other categories). The expected
lifespan for protection equipment, unlike that of most primary plant, is
probably much less than forty years. If replacement of equipment was low
prior to the collapse of the Soviet Union and in 5-8 years following this period,
as seems to be the case (see Section 0), then much of the protection
equipment in the system has needed replacement by the time of the CIGRE
study. The model uses only equipment greater than forty years for
assessment against age-based criterion for replacement, and will therefore
not predict the extraordinary levels of protection equipment replacement that
occurred in practice. Figure 5 also shows, for Eastern Europe/former Soviet
Block, a higher than average level of replacement of switchgear,
transformers, lines and control equipment than in many other regions. These
combined are sufficient to support the level of replacement predicted by the
model for this region.

The South American case is interesting: despite relatively high failure rates
(shown by model predictions and confirmed by CIGRE study results) and
model predictions of relatively high levels of poor condition equipment, the
reported level of replacement is lower than predicted. Inspection of Figure 25
suggests that refurbishment could be at a slightly higher level than predicted
by the model. Refurbishment can displace replacement (both cannot be
applied to the same equipment at the same time), so this may be part of the
reason for overestimation of replacement by this model. However, inspection

156
of Table 20 below suggests another possibility. This table shows that South
America has a relatively low level of equipment approaching its nominal life.

Table 12 Substation equipment 20 and more years of age

Region Percentage of equipment greater than and equal to 20 years
A 35
AU 63
EE 44
NA 70
SA 38
WE 61

It may be that utilities in South America are choosing to repair and refurbish
rather than replace equipment that is in poor condition (even though relatively
young). If poor-condition equipment is repaired but not restored to good
condition overall, the effect is to increase failure rates (see Section 6.8).

In Western Europe there appears to be a greater consciousness of the

impact of equipment aging on the need for replacement. For example a
paper (CIGRE WG 37.27, 2000) produced by CIGRE (which tends to be
European-dominated) speaks of a “bow wave” of replacement approaching,
which must be addressed immediately. One solution is to prioritise
replacement on the basis of condition and age and attempt to smooth the
replacement rate over time by bringing forward some replacement work.
Comparison of the percentages of equipment twenty or more years of age
(Table 12) for North America and Western Europe (where electricity networks
are both mature) suggests that this “smoothing” process might be already
underway. Discussions with one Western European utility also support this
proposition (Allison and Jay, 2001). The ability to undertake early
replacement of equipment will be strongly affected by available financial
resource: in this respect the utilities of Western Europe appear to be in a
better position than most (see Section 0). Even so, there does seem to be
some concern that the impact of electricity reform will be to constrain
revenues in the future. It is possible that utilities may be choosing to replace
equipment early in anticipation of future revenue reductions.

157
5.2.8 Discussion of results

5.2.8.1 Efficacy of the methodology

The methodology employed has permitted a wide range of numeric and non-
numeric parameters to be modelled, even where precise relationships are not
fully known. The methodology enables the modeller to experiment with the
relationships between different factors. The hierarchical nature of the model
means that an understanding can be gained about the underlying causes of
differences in performance.

In modelling the system there were compromises found necessary between

coarseness of output and the complexity of the rules. Coarseness could be
reduced by introducing more levels of output (for example adding moderately
high and moderately low to the standard {low, moderate, high} output set).
The benefit of this is offset against an increase in the number and complexity
of rules needed to support the enlarged output set. Additional complexity
means greater effort to check and debug, often with little overall gain to the
quality of outputs. This issue is partially addressed by the use of fuzzy logic
to implement the rules.

Similarly there had to be a trade-off between number of factors taken into

account for a particular task and the complexity of writing and debugging the
rule set. Maintenance spending was an especially complex task to write.
The writing / debugging burden in this case was managed by dividing the
task into multiple subtasks. This led to questioning as to whether it was
possible to simplify the tasks by identifying dominant factors. The answer to
this can be assessed by looking at the sensitivity of outcomes to various
input parameters, which will be discussed further in the following sections.
The output functions set up for the modelling exercise have been very useful
for this purpose.

The data to support development of regional case studies proved to be a real

difficulty. The CIGRE study data were often incomplete, of varying degrees
of accuracy, and sometimes inconsistent. The analysis required careful

158
sifting of survey responses, and assumptions about particular answers.
Performance data from this survey were particularly sparse. The CIGRE
study data have had to be supplemented by use of many other sources.
While every effort has been made to ensure that the supplementary material
is compatible with the CIGRE survey results, the data do not relate to the
same set of companies. Even within the CIGRE survey results this
compatibility problem can arise where respondents to one question are not
the same set of respondents to another question (since many respondents
did not answer all questions).

Another data problem arises because many of the circumstances that make
performance, policies and practices different from one enterprise to the next
are taken for granted by those involved. Thus these things are rarely
documented but they can have a significant impact on outcomes. To
understand the differences it is necessary to “read between the lines”. The
interviews conducted with industry representatives have also proven useful
for this purpose. An example is the understanding of reliability, which
appears to vary widely. Corporate culture is another area that can
significantly impact on performance.

The use of regional data to test the model has advantages and
disadvantages. The main advantage was being able to use the CIGRE study
results to access a range of data that is not readily available from other
sources. It was not possible to do this on a country by country basis because
the data were stripped of company and country details (for confidentiality
reasons). The main disadvantage was that in large regions such as Asia and
South America, where there are wide variations in asset management
practices, maturity of networks and financial circumstances, it was difficult to
combine data into a reasonable “typical” regional case. In Asia, for instance,
there are some countries with mature slow growing electricity networks and
many others with rapidly expanding networks and high load growth.
Revenue adequacy and indebtedness also vary significantly over the region.

159
Developing a “typical” profile in such cases tends to produce an “average”
result that masks the more extreme conditions. From a modelling point of
view this means that only middle-of-the-range rule conditions are tested,
rather than the full possible range. For this reason it was decided to
undertake some case studies of individual companies in order to refine the
model.

The main limitation with the fuzzy model is its inability to represent feedback
loops and delays fully. For this reason the system dynamics model has been
developed, and will be used to investigate some of the more important
relationships and feedback mechanisms. On the other hand, a benefit of this
formulation is that it encompasses a much broader range of inputs than is
generally possible with more traditional engineering modelling techniques.

5.2.8.2 Insights to be gained from these studies

The advantage of this type of modelling is not that one can compare results
from different cases – after all, performance data can be compared directly
provided they are in compatible form. The model’s power is that enables
comparison of the underlying reasons for differences and similarities in
outcomes. In this section examples are given of the model’s use to
investigate some questions arising from the results of regional case studies.
The full set of results predicted from the fuzzy model is listed in Table 14.

Examination of the causes of variation of availability and reliability in

Western Europe, North America and Australasia/South Africa
In Table 13, comparison of availability and reliability figures shows that the
model predicts higher availability for Australasia/South Africa and North
America than for Western Europe. However reliability of supply is predicted
to be high for all three cases (slightly higher in WE and AU than NA).

Table 13 Parameters used to assess availability and their values for AU, NA and WE
cases
Parameter AU NA WE
maintenance actions
moderate (50) moderate (52) low (28)
increasing availability
construction outages low (24) low (10) low (15)
failure rate low (10) low (24) low (16)
flexibility moderate (43) high (69) low (31)

160
Examination of this table suggests that the parameters maintenance actions
increasing availability and flexibility may have a role in this result. The
sensitivity of availability to flexibility is plotted in Figure 27.

100
high

80
availability.

60
moderate

AU
40
NA
20 WE
low

0
0 20 40 60 80 100
flexibility %

Figure 27 Availability as a function of flexibility for three cases from fuzzy model

161
 Table 14 Outputs from Fuzzy Model – Regional Studies
Output Parameter A AU EE NA SA WE
Availability high (98-100%) (70) high (98-100%) (85) moderate (95-98%) (32) high (98-100%) (85) moderate(95-98%) (50) moderate (95-98%) (62)
Construction outages moderate (45) low (24) moderate (50) low (10) moderate (60) low (15)
Corrective maintenance
low (27) low (10) low (10) low (23) low (30) low (10)
costs
Cost of finance moderate (30) low (10) moderate (50) moderate (30) moderate (50) low (10)
Debt repayments low (40) low (10) moderate (37) low (23) high (83) low (17)
Equipment replacement low (28) low (24) medium (50) low (16) medium (47) low (24)
Ease of raising capital easy (70) easy (90) moderate (50) easy (70) moderate (50) easy (90)
Factors increasing asset
low (41) low (33) low (24) low (34) moderate (44) low (31)
costs
Failure rate medium (34) low (10) medium (37) low (24) medium (65) low (16)
Failure rate contribution
moderate (50) low (10) moderate (30) low (10) moderate (63) low (23)
of load & climate
Financial drivers for
low (34) moderate (56) high (50) high (58) low (30) low (23)
maintenance spending
Financial incentive
high (77) high (90) moderate (50) high (74) low (10) high (90)
replacement
Investment cost high (63) very low (18) moderate (53) very low (20) very high (86) very low (17)
Investment in new
high (86) low (23) medium (50) low (10) high (76) low (15)
equipment
Loading heavily loaded (50) heavily loaded (40) heavily loaded (51) moderately loaded (30) v. heavily loaded (63) moderately loaded (30)
Maintenance actions
low (35) moderate (50) low (10) moderate (52) moderate (59) low(28)
increasing availability
Maintenance spending low (31) high (47) moderate (55) moderate (35) low (30) high (73)
Proportion poor condition
low (23) low (30) moderate (50) moderate (30) moderate (53) moderate (30)
equipment
Purchasing efficiencies high (60) low (10) low (23) low (10) low (10) low (10)
Refurbishment low (14) low (23) moderate (50) low (12) moderate (34) low (29)
Refurbishment costs low (12) moderately low (21) moderately low (30) low (7) low (10) moderately low (27)
Reliability of supply high:99-100% (72) high: 99-100% (90) moderate: 95-98% (68) high: 99-100% (83) moderate: 95-98% (50) high: 99-100% (90)
Replacement costs
low (10) low (10) low (10) low (10) low (10) low (10)
(maintenance only)
Replacement on age low (22) low (20) low (20) low (20) low (20) low (20)
Replacement on
low (23) low (30) moderate (50) low (24) moderate (40) low (30)
condition
Replacement on failures low (34) low (20) moderate (20) low (24) moderate (57) low (20)
Routine maintenance
moderate (50) moderate (42) moderate (54) moderate (41) moderate (50) moderate (48)
costs

162
Output Parameter A AU EE NA SA WE
Spending limits -
moderate (50) unlimited (70) moderate (30) unlimited (70) moderate (40) unlimited (90)
maintenance
System Augmentation high (77) medium (30) low (10) low (10) medium (69) low (10)
System factors reducing
high (90) medium (42) high (73) high (75) high (90) medium (46)
maintenance costs
Use of offline condition-
moderate (37) moderate (66) moderate (50) moderate (30) moderate (43) moderate (57)
based maintenance:
Use of online condition-
moderate (32) moderate (33) low (10) low (26) low (10) low (24)
based maintenance
Use of predictive
moderate (32) moderate (68) moderate (50) low (23) moderate (40) moderate (56)
maintenance

163
Figure 27 reveals that in this case, although the availability predicted for the
Western European case is sensitive to flexibility, it is not sensitive around the
present level (31%): a large change in flexibility from 31% to 80% would be
required to achieve a significant improvement in availability. The model also
suggests that availability is not as sensitive to flexibility in the other two
cases. In the North American case it can also be observed from Figure 27
that the maximum availability is constrained by a parameter other than
flexibility.

Flexibility is a function of network design, so in this case the historical

development of the network and past design practices have contributed to
current performance in terms of availability. In practice, increasing the
flexibility level is likely to involve significant capital investment; this is unlikely
to be achieved in a short time frame, unless significant system augmentation
is undertaken. Since Western Europe has largely mature transmission
networks with low growth levels this is unlikely to be practicable. It is also
arguable that if availability does not affect consumers significantly, which is
almost certainly the case in Western Europe because of the highly meshed
networks there, the expenditure of additional capital for the purpose of
increasing availability alone is not justified.

The task maintenance activities increasing availability has four inputs: live
working, combining maintenance activities (by circuit), use of online CBM and
use of predictive CBM.

Table 15 does not reveal large differences in values for any of the
parameters contributing to the task maintenance activities increasing
availability but, compared with the Australasian case, the values for the
Western European case for live working, combined maintenance activities
and use of offline CBM are all slightly lower. The North American case
exhibits a higher level of combined maintenance activities and slightly higher

164
use of live working than the Western Europe case, but lower use of online
CBM and predictive maintenance in general.

Table 15 Components of the task maintenance activities increasing availability for

AU, NA and WE
Parameter AU NA WE
live working low 20% low 11% low 3%
combined maintenance
medium (63) high (71) medium (57)
activities
use of online CBM moderate (33) low (26) moderate (33)
use of offline CBM moderate (68) low (23) moderate (56)

Figure 28 illustrates that, for both AU and WE, availability is sensitive to

modification of maintenance activities, but there is less sensitivity to changes
in these maintenance activities in the North American case. The difference
between the NA and WE cases is attributable to the disparity in level of
flexibility. On the other hand, the discrepancy in availability between the AU
and WE cases (where there is less difference in flexibility levels) is more
likely to be related to the combined differences in maintenance strategies
employed, as reflected in the calculated value of maintenance activities
increasing availability.

100
80
availability

AU
60
NA
40
WE
20
0
0 20 40 60 80 100
maintenance actions increasing
availability

Figure 28 Availability as a function of maintenance actions increasing availability for

three regions. Availability for WE is sensitive to this parameter about the current
operating point (28).

165
For reliability the model uses four parameters – redundancy, connectivity,
flexibility and failure rate, three of which are related to system design.
Increases in any of redundancy, connectivity or flexibility will tend to increase
reliability, while failure rate acts in the opposite direction.

Table 16 suggests that the significant differences between regions are in

connectivity and flexibility. Figure 29 illustrates the effect of connectivity on
reliability for these regions: the North American case is insensitive to
connectivity because flexibility is high for this region, whereas for Western
Europe, where flexibility is relatively low, reliability is sensitive to connectivity
for low to moderate levels of connectivity. Figure 30 illustrates the combined
effect of flexibility and connectivity for Western Europe – for high levels of
connectivity the reliability is insensitive to flexibility.

Table 16 Parameters used to determine reliability and their values for AU, NA and WE
Parameter AU NA WE
connectivity moderate (20) moderate (20) high (32)
failure rate low (10) low (24) low (16)
flexibility moderate (43) high (69) low (31)
redundancy moderate (50) moderate (50) moderate (50)

100
high

80
reliability
moderate

60
AU
40
NA
20 WE
low

0
0 10 20 30 40
connectivity (normalised 1-100)

Figure 29 Sensitivity of reliability to connectivity for three regions

166
 90

low moderate high.

flexibility
70
85-95
50
75-85
x 30 65-75
55-65
0
0

12

18

24

30

361
low moderate high
connectivity

Figure 30 Combined effect of connectivity and flexibility on reliability for WE case. X

denotes the existing situation.

In summary, because of the highly meshed networks in WE, high supply

reliability can be achieved at lower levels of circuit availability. In AU, where
the networks are less highly meshed, maintenance strategies needed to
support high reliability of supply also contribute to high circuit availability. On
the other hand, in NA, where flexibility is typically higher, high availability and
reliability of supply can be achieved without resorting to extensive use of live
working or high levels of online CBM.

Investigation into the contributions to failure rate in Asia

compared with South America
Asia and South America both contain developing countries and rapidly
growing networks, while in South America the predicted and actual failure
rates (from the CIGRE study) are higher than in Asia (see Figure 21). In the
fuzzy model there are four factors that contribute to calculation of the task
failure rate. These are proportion of poor condition equipment, use of
predictive maintenance, failure contributions of load and climate, and
spending limit (maintenance spending constraints). The parameters and
their values for these two cases are shown in Table 17. Examination of this
table suggests that the main difference between the two cases is the
proportion of poor condition equipment.

167
Table 17 Parameters used to determine failure rate and their values for the Asian and
South American cases
Parameter Asia South America
Failures load & climate moderate (50) moderate (63)
Proportion poor condition equipment low (23) moderate (53)
Spending limit moderate (50) moderate (50)
Use of predictive maintenance moderate (32) moderate (41)

Note that all the inputs to the task failure rate are themselves tasks. The
factors that contribute to proportion of poor condition equipment are
maintenance history and proportion of aged equipment.
As Table 18 reveals, the difference between the two regions derives from
maintenance history.

Table 18 Parameters contributing to the task proportion poor condition equipment

with values for Asia and South America
Parameter Asia South America
maintenance history well-maintained (70) under-maintained (25)
proportion aged equipment low (6%) low (6%)

40
36
proportion aged equipment
32
high

60-70
28
50-60
low moderate

24
40-50
(%)

20
30-40
16
20-30
12
8
X
4
10

30

50

70

90

under- maintained well-

maintained maintained
maintenance history

Figure 31 Failure rate as a function of maintenance history and proportion aged

equipment for South America. X marks the estimated levels of maintenance history
and proportion of aged equipment.
Figure 31 suggests that increasing maintenance effectiveness can have a
significant effect on failure rate for the South American case, but as the
network matures and the average age of equipment increases there is a
potential for an increase in age-related failures. Note that the result is not

168
very sensitive to the estimated value of maintenance history. Any
maintenance history level less than about 50 would give the same result.

Investigation into maintenance spending in North America compared

with Australasia/South Africa and Western Europe
Table 14 suggests that maintenance costs in North America are lower than in
either Western Europe or Australasia/South Africa, and this is consistent with
data from the CIGRE study. The parameters contributing to maintenance
spending are shown in Table 19. Of the three cases North America has the
lowest routine maintenance and refurbishment costs. Australasia has slightly
higher routine maintenance costs and higher refurbishment costs than North
America, but both are still lower than those of Western Europe.

Table 19 Factors contributing to maintenance spending, with values shown for AU, NA
and WE
AU NA WE
corrective maintenance
low (10) low (23) low (10)
costs
refurbishment costs moderately low (21) low (7) moderately low (27)
replacement costs (treated
low (10) low (10) low (10)
as maintenance)
routine maintenance costs moderate (42) moderate (41) moderate (48)
spending limits unlimited (70) unlimited (70) unlimited (90)

The lower routine maintenance costs from North America are partially offset
by a slightly higher value for corrective maintenance costs, consistent with
higher corrective maintenance costs as a proportion of total maintenance
costs observed in the CIGRE study (reported in Table 1).

The task routine maintenance costs is a function of five factors shown in

Table 20.

Table 20 Factors contributing to routine maintenance costs (for AU, NA and WE)
Parameters AU NA WE
System factors reducing
medium (42) high (75) medium (46)
maintenance costs
Use of online CBM moderate (33) low (26) low (24)
Geographic factor moderate (21) moderate (17) low (6)
Maintenance relative to
at recommended levels less than recommended at recommended levels
manufacturers’
(56) (15) (45)
recommended level
labour costs medium ($19k p.a.) high ($29k p.a.) high ($26k p.a.)

169
Examination of Table 20 reveals two main contributors to the difference in
routine maintenance costs between these cases: The NA case has a higher
value for system factors reducing maintenance costs, and a lower value for
maintenance relative to manufacturers’ recommended levels, both of which
will act to reduce routine maintenance costs. The four factors taken into
account for system factors reducing maintenance costs are redundancy,
flexibility, connectivity and load factor. It is the high flexibility level for North
America that results in the high value for the system factors task: redundancy
and flexibility have greater impact than connectivity on maintenance costs.

The difference in maintenance relative to recommended levels between

regions is an interesting one. There are several possible reasons for this
difference. Perhaps the most likely is that historically there has been a
stronger financial driver to minimise maintenance costs in North America,
than in the other regions, owing to many of the North American utilities being
investor-owned for their whole period of existence. Maintenance practices
develop slowly over time and tend to be a function of corporate memory
(“Our company does it this way”). The slow rate of change is partly a
function of the time lag between the change of a practice and its effect on
failure rates (and ultimately reliability). It may also be a function of reporting
practices and feedback within the organisation (See also Section 2.3.4 on
maintenance period optimisation). Practices that work tend to be reinforced,
but sustained pressure to reduce costs over time may reduce maintenance to
a bare minimum.

Refurbishment costs are a function of the level of refurbishment and labour

costs. (There is an implicit assumption that cost of materials is the same in
all cases, but in practice variations may be caused by factors such as
exchange rate and transport costs.) Refurbishment is a function of
proportion of poor condition equipment, use of predictive maintenance and
financial driver maintenance spending. These may be thought of as relating
to need, opportunity and inclination to undertake refurbishment. Table 21
lists the inputs to the task refurbishment and their values.

170
 Table 21 Inputs to the task refurbishment for three regions
Parameter AU NA WE
proportion poor condition
moderate (30) moderate (30) moderate (30)
equipment
use of predictive
moderate (68) low (23) moderate (56)
maintenance
financial driver
moderate (56) high(58) low (23)
maintenance spending

This table suggests that for North American companies, which use less
predictive maintenance, there is less opportunity to utilise refurbishment of
poor condition assets as a strategy, since the level of detection of poor
condition assets will be lower. The financial driver to keep maintenance
costs down may also discourage the use of refurbishment as a maintenance
strategy, (although it may be considered capital investment). In a
comparison between Australasia and Western Europe, the difference is that
in the latter there is less pressure on maintenance costs than in the former.

There is also a potential trade-off between refurbishment and replacement of

assets. In this fuzzy model the level of refurbishment is set, and then this
used as an input to the task that calculates the level of replacement of
equipment on the basis of condition. In practice the relationships between
maintenance, replacement and refurbishment are more complex than
modelled here. They will be examined in greater detail in Chapter 6.

5.2.8.3 Discussion of results of fuzzy modelling – regional cases

Results from regional studies using the fuzzy model have been shown to
demonstrate predictive power, despite the limitations of data both for inputs
and comparison purposes. Case studies based on individual enterprises,
thus having more consistent input data sets, will further confirm the predictive
power of the model.

However, the real strength of the model is the ability it gives the user to
investigate the causes of differences in results between cases. A question
was posed in Section 5.2.8.1: can the model be simplified by identifying
dominant factors? The answer is that while there are clearly core elements
within the model that play an important role in many outputs, different factors
dominate in the different cases studied.

171
The model shows that performance is limited by factors that vary from one
case to the next. Some of these factors are within the control of the
enterprise; others are outside its control. Hence even with the best intentions
it may be not possible for one enterprise to achieve the same level of
performance as another. A corollary to this is that the revenue needs of
transmission enterprises will vary, between utilities and over time according
to current and historical circumstances; this translates to differences in
transmission prices.

• Within the factors that are controllable, some are able to be modified
within a short time-frame; to effect change in other parameters, such
as flexibility, requires a longer time-frame and considerable capital
expenditure.
• Different aspects of performance, eg reliability and maintenance
spending, share some of the same factors of influence. Therefore
changes in policy will affect multiple aspects of a company’s
performance.
• The sensitivity of outputs to inputs around the present levels of those
inputs is different for different cases. The significance of this is that a
solution that is highly effective for one company may be of little use in
the circumstances of another company.

172
5.3 Case Studies for individual companies/countries
In this section, five case studies will be examined. Two of these are from
Asia: a Japanese case representing companies operating in a mature
network and a Malaysian case where the network is in a phase of rapid
development. The other cases are from Australia, Western Europe and
North America. Although the current status of utilities is described, where
possible, the cases have been constructed using data from 1998, in line with
CIGRE study data.

5.3.1 Australian case study – Powerlink Queensland

The Australian electricity supply industry was historically state-based. The
case study is based on the transmission company operating in the state of
Queensland.

5.3.1.1 Description of the Powerlink Queensland case study

The Queensland Electricity Transmission Corporation, trading as Powerlink
Queensland, was established as the state-owned corporation responsible for
the electricity transmission network in Queensland in 1 July 1997, subject to
the provisions of the Government Owned Corporations Act 1993.

The network of Powerlink is geographically extensive, covering 1700 km from

Cairns in North Queensland to the New South Wales border. It is moderately
interconnected in the southern and central regions, but lightly interconnected
in the north. There are two interconnections with New South Wales – the
Queensland New South Wales Interconnector (QNI), which began operation
in February 2001, and Direct Link, a small privately-owned DC
interconnection.

At mid-1998, Powerlink’s network comprised 4879 circuit km of 275kV line,

3943 circuit km of 132kV line and 441km of 110kV overhead transmission
line; there were only 16 circuit km of underground cable (Powerlink
Queensland, 1999). By mid-2001 this had increased to 505 circuit km of

173
330kV, 6084 circuit km of 275kV, 3984 km of 132kV and 528km of 110kV.
There was no increase in length of cables employed.

A review (Simpson, 2001) undertaken on behalf of the regulator, the ACCC,

reports that
Investment strategies are driven by Code requirements,
ensuring that investment is limited to assets for which a
threshold rate of return will be achieved, and that new assets
satisfy regulatory requirements and that optimum life cycle
costs will be achieved. Powerlink adopts reliability centred
maintenance, whole of life cycle costing and quantitative risk
assessment techniques in developing their various policies,
procedures and standards.

Powerlink has adopted an aggressive attitude toward new technology. At a

CIGRE symposium on Asset Management, Simon Bartlett outlined
Powerlink’s vision for substations (Bartlett, 2000):
• An integrated approach is to be taken for the entire substation.
• Overall operation and maintenance costs are aimed at being less than
0.5% of total asset value, with very high reliability and availability.
• Total capital costs are to be competitive with the Australian industry
standards.
• Routine substation inspections are to be eliminated, and condition
monitoring and remote interrogation facilities will allow maintenance
staff to optimise their site visits.
In the past few years Powerlink has invested heavily in new technology
hybrid-GIS switchgear for new and refurbished substations, including ABB
PASS equipment for 275kV and 330kV substations and Siemens and Alston
modular GIS for 110kV substations. As a pioneer user of the new technology
substations Powerlink has been able to develop close relationships with
suppliers, leading to lower purchase prices. Video facilities have been
installed at a number of substations with wide-area network communications
to enable Brisbane-based staff to inspect substations remotely.

The climate of Queensland varies from tropical in the north to sub-tropical in

the south. Inland areas are dry and coastal areas vary, but some parts are

174
subject to high rainfall and humidity in summer. The south-eastern part of
the state has a high lightning incidence during summer months.

Load growth is moderate, but load factor and plant utilisation are relatively
high compared with other mature networks. Transfer capacity on the grid
varies with different system operating conditions, including which generators
are operating and their level of output. These factors, coupled with
competitive forces and the entry of new generating capacity, have resulted in
considerable uncertainty in planning of network augmentations (Green et al.,
2002). Many sections of the network operate at of near capacity for extended
periods, especially in summer when the grid is more heavily loaded relative
to its capacity (Green et al., 2002). System demand grew by 2.5% from
1997/8 to 1998/9(Powerlink Queensland, 1999), and 4.1% in 2000/1
(Powerlink Queensland, 2001), moderate by world standards. Transformer
utilisation was 32 - 33% for the period, and losses were 4.3 to 4.1 %.
(Transmission losses are higher for Queensland than other Australian utilities
because of Queensland’s long transmission lines.) Load factor for 1998/9
was approximately 69% (ratio energy imported on the grid to maximum ½ hr
generated) (Powerlink Queensland, 1999). Performance of the network in
terms of reliability and availability is high (see Table 7 and Table 9 for
reliability and availability respectively).

Direct maintenance costs in 1998/9 were 1.2% of replacement asset value

including refurbishment costs. In 2000/01 maintenance costs excluding
refurbishment were $AU22.1 million, and refurbishment costs $11 million,
compared with $25 million and $7.4 million in 1998/99 (in 2000/01 dollar
terms) (Simpson, 2001). Powerlink outsources 70% of its maintenance
work. Refurbishment project work includes
• Replacing airblast circuit breakers and circuit breakers with
inadequate fault levels;
• Replacing SVC control systems;
• Replacing substation batteries;
• Replacing some transmission line insulators, conductors and
dampers;

175
 • Decommissioning some substations (Simpson, 2001).

Powerlink reports significant difficulties in taking lines out of service for

maintenance - 35% of circuits are not available for taking out of service
during normal working hours, and another 43% of circuits may not be
available for taking out of service during normal work hours depending on the
generation despatch and network configuration at the time (Simpson, 2001).
Live line practices are used, and in 2002 live substation work practices were
introduced. Additional maintenance costs are incurred as a result of
remoteness of sites. Average travel time to sites in South East Queensland
is 2 hours (Simpson, 2001), and is likely to be higher than this in the northern
and central regions. Helicopter-based line inspections have been used
successfully to reduce maintenance costs for overhead transmission lines.

Powerlink has a well-developed condition-based assessment program for

transformers (Allan, 1991, Allan, 1993) which has been operating
successfully for some years. In recent years the corporation has put
considerable resources into information management including
communications infrastructure, data management (ensuring data are
available to the right people, at appropriate level of granularity, sufficient
quality and in a timely manner). Data are used routinely for manual and
automatic analysis, and the goal is to achieve an integrated management
information system that can access data from different databases within the
organisation (Bell, 2003).

Powerlink’s financial position is sound. It has moderate debt exposure (Table

38), adequate revenue and a reasonably transparent regulatory system
based on a revenue cap. As a government-owned enterprise it enjoys a high
sovereign credit rating (AAA). However, the Competition Principles
Agreement (1995) requires the state government to impose a debt guarantee
fee on Government-owned enterprises to offset the competitive advantages
(particularly interest rate advantages) provided by government guarantees
(Productivity Commission, 2002). Powerlink’s standalone rating is A.

176
Table 22 lists the input parameters for the case data-set, based on this
analysis and a response to the CIGRE survey.

Table 22 Input parameters for the Australian Case study

Input Parameter Value
Age as replacement criterion insignificant (20)
Attitude to new technology aggressive (75)
Business acumen high (65)
Climate sub-tropical
Climatic factor moderate (50)
Combining of maintenance activities medium (63)
Connectivity factor moderate (22)
Cost of land low (30)
Credit rating A (50)
Debt level moderate (50)
Earth quake zone No (10)
External barriers moderate (40)
Flexibility low (34)
Frequency of failure as replacement criterion very significant (80)
Geographic factor low (13)
Historical revenue sufficiency not a concern (80)
Investment cost as KPI insignificant (30)
Labour cost high $US 26 x1000/annum
Labour skill high 70
Live working low 20%
Load factor moderate 63%
Load growth moderate 4.0%
Losses (T&D) moderate 9%
Maintenance history maintained (65)
Maintenance relative to manufacturers' recommended level at recommended levels (50)
Maintenance spend as KPI insignificant (0)
Number of failures as replacement criterion significant (62)
Pilfering: low (10)
Profit as Intention Yes (70)
Proportion of aged equipment low 3%
Proportion of young equipment moderate 47%
Redundancy moderate (n-1) (50)
Regulatory support highly supportive (70)
Replacement considered an investment significant (50)
Revenue sufficiency adequate (65)
Rugged terrain not rugged (30)
Size of company small (20)
Special relationship with vendors some use (60)

5.3.1.2 Results of the Powerlink Queensland case study

Table 23 Outputs for the Australian case study

Output parameter Value
Availability high (98-100%) (81)
Construction outages low (27)
Corrective maintenance costs low (10)
Cost of finance low (10)
Debt repayments low (10)
Equipment replacement low (24)
Ease of raising capital easy (90)
Factors increasing asset costs moderate (45)
Failure rate low (10)
Failure rate contribution of load & climate low (10)
Financial drivers for maintenance spending high (75)
Financial incentive replacement high (90)
Investment cost high (53)

177
Output parameter Value
Investment in new equipment medium (43)
Loading heavily loaded (40)
Maintenance actions increasing availability moderate (57)
Maintenance spending low (31)
Proportion poor condition equipment moderate (30)
Purchasing efficiencies low (23)
Refurbishment low (18)
Refurbishment costs low (15)
Reliability high: 99-100% (90)
Replacement costs as maintenance low (10)
Replacement on age low (20)
Replacement on condition low (30)
Replacement on failures low (20)
Routine maintenance costs moderate (40)
Spending limits (maintenance) moderate (57)
System Augmentation medium (45)
Factors reducing maintenance costs medium (46)
Use of offline condition-based maintenance moderate (66)
Use of online condition-based maintenance moderate (43)
Use of predictive maintenance high (71)

Results from the model generally appear to be in line with observed activities
and performance. The availability is high, and this is consistent with reported
data (see Table 9). The result reported in the table above may be slightly
reduced by the effect of construction outages. In practice Powerlink has
been able to minimise construction outages in some cases by building new
plant adjacent to older plant, avoiding the need to take circuits out of service
during the construction phase. Transmission companies in more urban areas
may not share the luxury of available space to do this. However, the model
does not take into account the available space when calculating construction
outages.

The model prediction of high reliability is also supported by reported data

(see Table 7). This outcome is supported by the low failure rate of
equipment.

The model for ease of raising capital is fairly simplistic, and does not take into
account the size of the local capital market. Hence it seems likely that the
model overstates the ease of raising capital in this case. This may slightly
affect the level of equipment replacement.

5.3.2 Malaysian case study

In Peninsula Malaysia power is supplied by Tenaga Nasional Berhad (TNB).

178
In Sabah the supplier is Sabah Electricity Board (SEB) and in Sarawak the
supplier is Sarawak Electricity Supply Corporation (SESCo) (for generation,
transmission and distribution). The case study will focus on TNB, which has
a high public profile.

5.3.2.1 Description of the Tenaga Nasional Berhad case

TNB was formed from the national electricity board Lembaga Letrik Negara
(LLN) through the Electricity Supply (Successor Company) Act of 1990. TNB
was incorporated in 1990 in Malaysia as a public limited company. It was
listed on the Malaysian stock exchange on 28 May 1992 (KLSE, 2002).

Like many Asian utilities TNB is a vertically integrated company: it has

generation, transmission, distribution and system operator functions. TNB’s
operations are regulated through a licence issued by the Director General of
Electricity Supply (DG). This licence was issued to TNB for an initial period
of 21 years from 1 September 1990.

Total installed generating capacity in Malaysia in 2002 was 14624 MW, of

which 12975 is in Peninsular Malaysia (Trade Partners UK, 2002). In
Peninsula Malaysia TNB Generation Sdn provides 52% of installed capacity,
TNB Hydro Sdn 14%, and five Independent Power Producers (IPPs) provide
34%. TNB holds a monopoly for transmission and distribution on the
peninsula (ICEM A/P, 1997). It is also responsible for construction of
transmission lines and substations for interconnection of IPPs to the National
Grid (Hills, 1997).

TNB is part privately owned, part government owned. In 1997 the Malaysian
government was still the majority stakeholder (Hills, 1997). It has a
sovereign credit rating from Standard & Poor’s of BBB (Standard & Poor's,
2002). It is government-controlled, and its core activities are driven by the
Malaysian government’s industrialisation program (Hills, 1997). The
Malaysian government has a goal of making Malaysia a developed nation by
2020. In January 2001 Asia Week reported that Tenaga was planning to trim
its state ownership from 70% to 51% (Shameen, 2001). The company has

179
been divesting itself of generating plant, and will use the proceeds to retire
most of its debt (Shameen, 2001).

Tenaga has purchased the Sabah Electricity Board, now SESB. Currently a
significant amount of capital is being expended for system rehabilitation work
in Sabah, including upgrading the 66kV network to 132kV and 275kV. This
part of the network had been experiencing significant reliability problems.

The TNB transmission system, known as the National Grid, operates at

mainly 132kV and 275kV. Some 66kV lines are being upgraded to higher
voltages (Hills, 1997). TNB has embarked on construction of a 1024km
500kV transmission line network which will form the (north-south) backbone
of the transmission grid (Hills, 1997). The transmission line has been
completed, but at this stage only the central section is energised at 500kV,
the other two sections being operated at 275kV (Mohd Noor, 2002).

In 1999 Malaysia generated around 59 billion kWh of electricity (EIA, 2001c).

During the 10 years to 1997 demand grew in Malaysia by more than 12% per
annum. Demand weakened by 3-4% from 1997-1999 because of the Asian
financial crisis. TNB forecast for 2001 is 10% growth. Load growth is
expected to be more than 8% per year for 2001- 2005(EIA, 2001c).

Between 1991 and 1995 the transmission and distribution networks were
expanded both to improve coverage and to enhance reliability and stability.
By the end of the Sixth Plan period, 92% of rural households in Malaysia
were served with electricity, compared with 80% in 1990 (Hills, 1997). The
focus of the Seventh Plan (1996-2000) was on expansion and upgrading of
the transmission and distribution infrastructure (Hills, 1997). Of total
expenditure on rural electrification projects by the Federal government, 87%
was for grid extension projects (Hills, 1997). TNB additionally invested RM
100 million in rural electrification projects in the peninsula. The
“unprecedented growth in demand especially in Peninsula Malaysia strained
the TNB system, resulting in supply interruptions” (Hills, 1997).

180
The Malaysian Ministry of Energy, Communications and Multimedia (MECM)
believes that the Malaysian Electricity Market will require investment of the
order of $US 18.5 billion (RM70Billion) over the next ten years: 60-70% in
generation, and the rest in transmission, distribution and retail (Trade
Partners UK, 2002).
In January 1997 Asia Week reported that Tenaga planned to spend some
$US 2 billion a year until 2000 on infrastructure, including $US2.8 billion for
146 substations (Singh and Poh, 1997). Asia Week speculated on whether
TNB would be able to achieve this, given its worsening debt position. Today
construction is continuing but significant parts of the plan have still been
delayed, because the current load growth is slower than predicted at the time
the plan was developed (Mohd Noor, 2002).

Malaysia has currently interconnections to Thailand (single 117MVA, 132kV

line) and Singapore (2 x 250kV transmission lines and submarine cables)
(Hills, 1997). In 1997 it was expected that these would be upgraded, the
Thailand interconnector to 300MW (Hills, 1997). These connections would
form part of the ASEAN grid.

During the 1990s Malaysia suffered a number of serious power outages:

• In 29 September 1992 lightning struck a key substation in the grid,
shutting down much of the Peninsula (Singh and Poh, 1996).
• In June 1995 fire destroyed three cables to Penang, “virtually
crippling” the island for 20 days (Singh and Poh, 1996).
• The whole of Peninsular Malaysia was without electricity for one day in
August 1996 (Hills, 1997, Singh and Poh, 1997).
Singh (Singh and Poh, 1997) says since it was listed TNB has had a “string
of embarrassing incidents that have earned the state-controlled power utility
widespread public ire”.

Planning is undertaken to meet an N-1 criterion. Redundancy has been

maintained at this level, despite high load growth. Equipment is not subject
to significant overloading. There is some pilfering (at distribution levels). The

181
problem is more severe in Sabah, than on the peninsula, but even so
amounts to a significant loss of revenue.

In 1997 Tenaga announced that it had reduced the outage index by nearly
70% over a 19 month period; the index is in minutes lost per customer per
year. At the time of the announcement it was about 300, down from 1000 in
mid-1995. (In Singapore customer-minutes-lost in 1996 was 40-50, but
Singapore’s system is entirely underground. (Singh and Poh, 1997)) The
current (2002) index for transmission system minutes lost stands at 17
minutes. This index includes all incidents occurring on the transmission
system that cause loss of load (Mohd Noor, 2002).

TNB has a huge variety of equipment: policy has been to buy the cheapest
available that meets technical specification. Some equipment from particular
suppliers has caused reliability problems. Currently there is no emphasis on
life cycle costs for transmission.

Circuit breakers are mainly SF6 – both AIS and GIS (outdoor as well as
indoor). Management has been willing to try new technologies. All oil circuit
breakers have been replaced at around 30 years of age because of number
of failures occurring (Mohd Noor, 2002). These CBs were originally from the
UK.

Transformers have tended to be replaced only because of load growth

considerations. Few 275/132 kV transformers have been replaced (except
on failure or for reasons of inadequate capacity). Most are only 30 to 40
years old. The 500kV equipment is very new.

The climate, which is hot and very wet, with a high lightning incidence (200
lightning days per year) (Mohd Noor, 2002), plays a significant role in
equipment failures. The main causes of failure (Mohd Noor, 2002) are
• Lightning (especially on towers with steel crossarms – used mainly in
hilly areas where extra strength is required);
• Encroachment (people, trees);

182
 • Failure of timber cross-arms on towers;
• Degradation of insulators;
• Corrosion on conductors;
• Internal faults on substation equipment;
• Transformer explosive failures;
• Lightning arrestor failures;
• Animals (especially monkeys) shorting out substation equipment.

Time based maintenance is the predominant practice – carried out on all

equipment. This involves shutting down every 18 months, and spending one
working day to undertake routine maintenance including cleaning, painting,
oiling and greasing as appropriate, performing timing checks on circuit
breakers and accuracy checks on relays. Primary and secondary equipment
are checked at the same time. In general more maintenance is done than
recommended by the manufacturers of the equipment (Mohd Noor, 2002).
Maintenance at both ends of a circuit is consistently undertaken at the same
time.

Condition based maintenance, for transformers only, was introduced four to

five years ago. This involves tan-delta testing, dissolved gas analysis,
capacitive testing on bushings. The CBM is almost entirely off-line testing
(Mohd Noor, 2002). (Two online units for dissolved gas analysis are being
trialled). There is strong worker resistance to changing maintenance
practices – this is seen as a threat to jobs, and especially to overtime
payments, as wages are low.

Scrutiny and downward pressure on capital expenditure is very strong as this

is the largest component of spending (Mohd Noor, 2002). By comparison
there is little downward pressure on maintenance costs. Replacement of
equipment is considered part of maintenance, not capital expenditure.

The regulator is part of the government; the regulatory process is not very
transparent. To ensure that there is adequate supply of electricity to meet

183
future needs the government is keen to encourage IPPs and develop a
competitive generation market. It also recognises the need for sufficient
transmission development. The government is sensitive to reliability and
quality of supply issues but is also sensitive to the price of electricity paid by
consumers. Therefore, although the regulator is generally fairly sympathetic
to the needs of TNB, they do not always get the price adjustments that they
seek.

There are now some customers of TNB needing very high power quality eg
integrated circuit chip manufacturers. These large and influential companies
can apply pressure on TNB on supply quality and reliability.

The full data set for the Malaysian case study is listed in Table 24 below.

Table 24 Inputs to fuzzy model for Malaysian case study

Input Parameter Value
Age as replacement criterion Very significant (100)
Attitude to new technology conservative (30)
Business acumen moderate (45)
Climate tropical & wet
Climatic factor extreme (75)
Combining of maintenance activities high (80)
Connectivity factor low (14)
Cost of land low (25)
Credit rating BB (30)
Debt level high (65)
Earth quake zone No (30)
External barriers low (30)
Flexibility low (20)
Frequency of failure as replacement criterion significant (57)
Geographic factor moderate (16)
Historical revenue sufficiency adequate (65)
Investment cost as KPI very significant (70)
Labour cost low $US 3 x1000/annum
Labour skill moderate 50
Live working low 11%
Load factor moderate 60%
Load growth high 11.0%
Losses (T&D) moderate 9%
Maintenance history maintained (60)
Maintenance relative to manufacturers' recommended level more than recommended (85)
Maintenance spend as KPI insignificant (20)
Number of failures as replacement criterion very significant (67)
Pilfering: low (30)
Profit as Intention Yes (50)
Proportion of aged equipment low 6%
Proportion of young equipment high 65%
Redundancy moderate (n-1) (50)
Regulatory support reasonably supportive (60)
Replacement considered an investment insignificant (0)
Revenue sufficiency not a concern (70)
Rugged terrain moderate (60)
Size of company medium (35)
Special relationship with vendors no use (10)

184
5.3.2.2 Results of the Tenaga Nasional Berhad Case Study

Table 25 Outputs from fuzzy model for the TNB case Study
Output parameter Value
Availability low (>95%) (38)
Construction outages moderate (37)
Corrective maintenance costs low (10)
Cost of finance moderate (50)
Debt repayments high (70)
Equipment replacement medium (40)
Ease of raising capital moderate (50)
Factors increasing asset costs moderate (47)
Failure rate medium (43)
Failure rate contribution of load & climate moderate (46)
Financial drivers for maintenance spending high (50)
Financial incentive replacement high (50)
Investment cost high (64)
Investment in new equipment medium (50)
Loading heavily loaded (31)
Maintenance factors increasing availability moderate (58)
Maintenance spending high (69)
Proportion poor condition equipment moderate (37)
Purchasing efficiencies moderate (30)
Refurbishment low (5)
Refurbishment costs low (5)
Reliability moderate: 95-98% (62)
Replacement costs as maintenance moderate (37)
Replacement on age low (22)
Replacement on condition low (20)
Replacement on failures moderate (43)
Routine maintenance costs high (80)
Spending limits (maintenance) unlimited (70)
System Augmentation medium (50)
System factors reducing maintenance costs low (26)
Use of offline condition-based maintenance low (26)
Use of online condition-based maintenance low (26)
Use of predictive maintenance low (18)

These results reflect a company that is focused on meeting augmentation

needs associated with rapidly increasing demand. Predicted moderate
reliability is consistent with the figures reported above. Reliability and
availability are affected by the relatively high failure rate of equipment.
Factors contributing to the reliability result are investigated further in Section
5.4.

Relatively high maintenance costs are predicted partly because of high

routine maintenance costs. Maintenance spending is also influenced by a
tendency to treat replacement costs as maintenance, and a fairly high
predicted level of replacement on failure.

185
5.3.3 Japanese case study
A “typical” Japanese case study has been developed, based on combined
data for a number of Japanese utilities.

5.3.3.1 Description of the Japanese case study

The Japanese power industry is composed of ten vertically integrated
privately-owned utilities. Until recently these companies have enjoyed
monopoly status in their areas of operation. Within each utility’s area the
networks are mature and highly interconnected. The interconnections
between utilities are less well-developed. In general the networks are
(geographically) compact, but serve high load levels, reflecting Japan’s high
population density and per capita electricity usage.

An attempt has been made to identify those respondents to the CIGRE

survey likely to be Japanese utilities, in order to use some of the data from
the survey. The criteria used for filtering were {Region = Asia, Ownership =
Private, Size = not small, Extent = regional, Functions = Generation,
Transmission and Distribution}. On this basis three responses were
identified as likely to be from Japanese utilities.

Load growth has been very low in recent years reflecting low growth in the
Japanese economy. Load factor is also fairly low. This has been identified
as a problem that results in low efficiency in the utilisation of plant
(generation and transmission) (IEA, 1999).

Japanese utilities have enjoyed stable and supportive rates-based regulation.

The Japanese government is gradually trying to introduce competition in the
retail part of electricity sector. As of March 2000 large consumers can
choose their electricity supplier (EIA, 2001b), but so far only a small
percentage of users have chosen to exercise this option. Electricity prices
are among the highest in the world, partly because of the need to import fuel
for generation of electricity (IEA, 1999). As Table 35 illustrates, most of the
Japanese utilities have high debt levels, but relatively high credit ratings
(especially for vertically integrated, privately owned companies), adequate

186
interest coverage owing to access to low interest rates and high income.
Surprisingly only one of the three Japanese respondents from the CIGRE
survey reported profit as the utility’s motive, the other two nominating
meeting fixed financial targets as their intention.

Operations and maintenance costs are high, and utilities have been
attempting to reduce these in the face of introduction of competition between
utilities. This is evident also from the CIGRE survey data, in which all three
respondents have high maintenance costs, and two of the three are among
the highest in the survey. Only one of the three respondents reported on
refurbishment costs as a proportion of maintenance, but for this one the
proportion was quite high for switchgear, transformers, overhead lines and
cables.

Design and operational standards are reported to be very high. Infrastructure

costs are high because designs must cope with a high earthquake incidence,
rugged terrain, and very high land costs. All three respondents report
flexibility (ability to remove a circuit breaker from service without taking the
associated circuit out of service) of 100%. This also supports the view that
Japanese networks are designed for high standards of reliability and
availability.

Reliability of supply appears to be quite high (see Table 7). CIGRE study
reliability and availability data for the particular data sets of interest are
scanty. However, the calculated values for unplanned outages per circuit
and per 100km of line are very low, for the two data sets for which required
information was provided. This suggests that the condition of equipment is
very good, and implies high standards of maintenance. The CIGRE study
results suggest that this is predominantly time-based, and that more
maintenance than recommended by manufacturers is undertaken. All three
responses reported doing more maintenance than recommended. In none of
the three responses was condition-based maintenance reported to be a
major strategy employed. One respondent indicated that corrective
maintenance was also used as a strategy.

187
All three responses indicate that age and condition are used as criteria for
replacing equipment. One respondent also nominated number of failures as
a replacement criterion. Two of the three respondents report moderate levels
of aged equipment (bays > 40 years of age) – 19.8% and 22.6%; the third
utility reports 1.1% of aged equipment. Perhaps the utility with the very low
percentage of equipment over forty years of age uses a lower trigger for
replacement than the other two. The three also report moderate levels of
equipment less than 20 years of age, ranging from 41% to 53%, with the
highest of these being from the utility with the lowest proportion of equipment
over forty years of age.

Table 26 summarises the input data used for this case study. The values
from the CIGRE study have been averaged. For True/False answers one
true answer was translated to a level of 33 (out of 100), two trues to 67 and
so on. For the parameter redundancy, a level slightly above the N-1 level
was used.

Table 26 Input data for the Japanese Case Study

Input Parameter Value
Age as replacement criterion Very significant (100)
Attitude to new technology moderate (40)
Business acumen moderate (50)
Climate temperate
Climatic factor moderate (40)
Combining of maintenance activities medium (44)
Connectivity factor moderate (16)
Cost of land high (90)
Credit rating AA (60)
Debt level high (80)
Earth quake zone Yes (80)
External barriers moderate (60)
Flexibility high (100)
Frequency of failure as replacement criterion insignificant (10)
Geographic factor low (6)
Historical revenue sufficiency not a concern (90)
Investment cost as KPI significant (33)
Labour cost high $US 32 x1000/annum
Labour skill high 80
Live working low 0%
Load factor moderate 58%
Load growth moderate 3.9%
Losses (T&D) very low 4%
Maintenance history well-maintained (80)
Maintenance relative to manufacturers' recommended level more than recommended (100)
Maintenance spend as KPI significant (33)
Number of failures as replacement criterion significant (33)
Pilfering low (10)
Profit as Intention No (33)
Proportion of aged equipment moderate 15%

188
Input Parameter Value
Proportion of young equipment moderate 48%
Redundancy moderate (n-1) (60)
Regulatory support highly supportive (80)
Replacement considered an investment very significant (100)
Revenue sufficiency not a concern (90)
Rugged terrain moderate (60)
Size of company large (65)
Special relationship with vendors no use (30)

5.3.3.2 Results of the Japanese case study

Table 27 Results for the Japanese Case Study

Output parameter Value
Availability high (98-100%) (90)
Construction outages low (10)
Corrective maintenance costs low (10)
Cost of finance low (10)
Debt repayments moderate (50)
Equipment replacement medium (37)
Ease of raising capital easy (90)
Factors increasing asset costs high (90)
Failure rate low (10)
Failure rate contribution of load & climate low (10)
Financial drivers for maintenance spending low (32)
Financial incentive replacement high (90)
Investment cost moderate (57)
Investment in new equipment medium (32)
Loading lightly loaded (11)
Maintenance actions increasing availability low (10)
Maintenance spending moderate (50)
Proportion poor condition equipment low (10)
Purchasing efficiencies moderate (43)
Refurbishment low (5)
Refurbishment costs low (5)
Reliability high: 99-100% (90)
Replacement costs as maintenance low (10)
Replacement on age moderate (40)
Replacement on condition low (20)
Replacement on failures low (20)
Routine maintenance costs moderate (50)
Spending limits (maintenance) unlimited (90)
System Augmentation low (10)
System factors reducing maintenance costs high (90)
Use of offline condition-based maintenance moderate (35)
Use of online condition-based maintenance low (10)
Use of predictive maintenance moderate (30)

Examination of Table 27 suggests the model results are consistent with the
observations noted in the description of this case, with the exception that the
model predicts only moderate levels of maintenance spending. The
elements contributing to the model of maintenance spending are routine
maintenance, refurbishment and corrective maintenance costs, replacement
costs if treated as maintenance, and the spending limits. Maintenance
spending is not constrained by financial considerations in this case. The

189
replacement costs element is low because replacement is always considered
an investment in this case. In this case corrective maintenance cost will be
very low because of the low level of poor condition equipment.

The model also predicts low refurbishment levels, based on the very low level
predicted for poor condition equipment. (The model does not consider the
possibility of refurbishment based on age alone, which is possibly an area for
improvement. However no data are currently available for criteria used for
decisions on refurbishment.) There is some suggestion that the actual level
of refurbishment may be relatively high (see comments in the previous
section). If the predicted level of refurbishment level were higher this would
have pushed the predicted maintenance spending to a higher level also.
This may indicate an unnecessarily high level of spending on maintenance
and refurbishment in particular, which could possibly be reduced significantly
if more condition-based maintenance were used.

Another significant contribution to maintenance spending is from routine

maintenance. The model predicts moderate routine maintenance costs
despite high labour costs, and higher than recommended levels of
maintenance because of the influence of system factors that reduce
maintenance costs. These factors are the high level of flexibility, the high
level of redundancy, moderate connectivity and moderate load factor. These
in combination mean that there is little need to undertake maintenance
outside normal work hours or to use live working techniques. This is also
confirmed in the CIGRE survey answers which show no usage of
maintenance out of hours for the three Japanese utilities contributing to the
survey. Even though the system factors reducing maintenance costs
correctly drives maintenance costs down, it may overestimate the effect on
routine maintenance costs. This is one of those cases in which the model
suffers from a lack of sensitivity, because there are only three levels of output
for routine maintenance spending.

190
5.3.4 US utility case study
This case study is based on a utility not represented in the CIGRE survey. In
building the data set for this case a greater number of assumptions have had
to be made than for some of the other cases. The North American base case
has been used as a starting point for developing this case with adjustments
made based on available information about the practices of this company
compared with typical North American companies.

5.3.4.1 Description of the Duke Energy case study

Duke Energy is the largest utility in the USA. It has generation, transmission
and distribution including retailing functions. The transmission function is
managed by subsidiary company Duke Power. Its network in North and
South Carolina comprises 44kV mainly radial system, 100kV, 220 kV (mainly
double circuit) and 500kV mainly single circuit lines. It is interconnected to
ten other networks and has 250 transmission tie stations.

Operation and maintenance is considered within the framework of the overall

business, as a factor impacting on share value. Reliability of the system is
considered to be acceptable when compared with reliability elsewhere in the
USA, so the strategy is to maintain reliability at current levels (Sherard,
2001). Performance is considered over a five-year window; performance
outside the target level triggers a review and corrective action based on the
causes of the problem.

The Capital and Operation and Maintenance spending programs are

optimised using a tool that maximises net present value given a range of
projects (Sherard, 2001). Some projects are non-discretionary, and some
are discretionary. The value of projects takes into account the cost of
outages to the customer, as well as the cost of the project (Woolweber,
1998). Customer outage costs are determined by detailed surveys and are
classified by customer-type and service-interruption-type (Sullivan et al.,
1996). Failures are funded as a non-discretionary project (Sherard, 2001).

191
Share-price is sensitive to budget over-runs, so spending is monitored
closely throughout the year. If funds run short some projects need to be cut.
Risk is taken into account in decisions associated with curtailing a project,
including the risk of double contingencies. Money can not be tied-over to the
following year, so it is critical that all operations and maintenance money is
spent.

Maintenance comprises a combination of condition assessment and

inspection techniques, condition-based maintenance and corrective
maintenance. Condition/inspection techniques include oil temperature, DGA
on transformers, conducted annually, “Doble testing”, that is, tan delta
insulation measurement (every six years), thermal and voltage system
screening (annual). In 1997 a graded approach to substation preventive
maintenance activities was introduced. This approach, based on a modified
RCM process includes (Woolweber, 1998):
• Identification of critical equipment;
• Targeting specific equipment failure causes to suitable preventive
maintenance tasks;
• Incorporation of predictive maintenance tasks where warranted; and
• Application of a maintenance plan commensurate with identified risks of
failure.
The process comprises the following stages:
• Develop a failure modes/failure causes (FM/FC) library for equipment
types eg SF6 circuit breaker.
• Collect supporting data such as preventive maintenance tasks, predictive
maintenance tasks, work and failure history, staff interviews, vendor
recommendations.
• Identify all maintainable equipment from a transmission tie station to
customers (including transmission and distribution equipment), assigning
to each a criticality level, based on its role in the supply of electricity to
customers, and impact (maintainability, consumer impact, failure
consequence).

192
• Assign to each tie station a criticality value (high or low for transmission,
and high, medium, low for distribution stations) based on an evaluation of
• legal;
• social/political; and
• environmental criteria,
in combination with the equipment failure criticality assessment.
• Develop maintenance tasks, based on collected maintenance information
and the FM/FC library. The equipment’s criticality rating is used to
determine the frequency of the preventive maintenance.

This utility is considered one of the best performing of all US utilities6. Its
system performance and attention to customer requirements have meant that
it has not been subjected to as much regulatory pressure or reform as many
other utilities.

Table 28 Input data for the US Utility Case Study

Input Parameter Value
Age as replacement criterion insignificant (20)
Attitude to new technology moderate (50)
Business acumen high (75)
Climate temperate
Climatic factor mild (30)
Combining of maintenance activities high (71)
Connectivity factor moderate (20)
Cost of land moderate (50)
Credit rating known Yes (90)
Credit rating A (50)
Debt level moderate (45)
Earth quake zone No (10)
External barriers moderate (50)
Flexibility high (69)
Frequency of failure as replacement criterion very significant (70)
Geographic factor moderate (17)
Historical revenue sufficiency not a concern (75)
Investment cost as KPI insignificant (26)
Labour cost high $US 29 x1000/annum
Labour skill high (70)
Live working low 11%
Load factor moderate 61%
Load growth low 2.9%
Losses (T&D) low 6%
Maintenance history well-maintained (75)
Maintenance relative to manufacturers' recommended level less than recommended (15)
Maintenance spend as KPI very significant (80)
Number of failures as replacement criterion significant (58)
Pilfering low (10)
Profit as Intention Yes (80)
Proportion of aged equipment moderate 14%
Proportion of young equipment low 30%

6
Duke Energy has been named in Fortune Magazine as Most Admired Company in the
Energy Sector (2002-3) and Most Admired Company – Gas & Electric Utilities (1999-2001)
based on criteria including (among others) Management and Product Quality, Long-term
Investment Value and Use of Corporate Assets.

193
Input Parameter Value
Redundancy moderate (n-1) (50)
Regulatory support highly supportive (75)
Replacement considered an investment very significant (86)
Revenue sufficiency not a concern (70)
Rugged terrain not rugged (20)
Size of company medium (60)
Special relationship with vendors some use (40)

5.3.4.2 Results of the Duke Energy case study

The results presented in Table 29 are very consistent with the information
known about this company. Interestingly, compared with the North American
case study, the predicted reliability is higher, use of predictive maintenance
higher, and maintenance costs lower than is typical for North America.

Table 29 Results from the US Utility Case Study

Output parameter Value
Availability high (98-100%) (87)
Construction outages low (10)
Corrective maintenance costs low (10)
Cost of finance low (10)
Debt repayments low (10)
Equipment replacement low (10)
Ease of raising capital easy (90)
Factors increasing asset costs low (24)
Failure rate low (10)
Failure rate contribution of load & climate low (10)
Financial drivers for maintenance spending high (90)
Financial incentive replacement high (90)
Investment cost very low (12)
Investment in new equipment low (10)
Loading moderately loaded (24)
Maintenance actions increasing availability moderate (48)
Maintenance spending low (17)
Proportion poor condition equipment low (18)
Purchasing efficiencies low (23)
Refurbishment low (5)
Refurbishment costs low (5)
Reliability high: 99-100% (90)
Replacement costs as maintenance low (10)
Replacement on age low (20)
Replacement on condition low (20)
Replacement on failures low (20)
Routine maintenance costs moderate (45)
Spending limits (maintenance) moderate (50)
System Augmentation low (10)
System factors reducing maintenance costs high (75)
Use of offline condition-based maintenance moderate (50)
Use of online condition-based maintenance low (23)
Use of predictive maintenance moderate (50)

Comparison the two sets of results suggests that, in the Duke Power case,
corrective maintenance costs are slightly lower and the financial drivers on
maintenance spending greater than the average for North America.
Corrective maintenance costs are slightly lower because of a lower failure
rate (10 compared with 24 for the NA case). This results from higher use of

194
predictive maintenance (50 compared with 23) and lower proportion of poor
condition equipment (18 compared with 30). The difference calculated level
of financial drivers for maintenance (90 compared with 58 for NA) is
influenced most strongly by the parameter profit-as-intention, for which the
Duke case has been assigned a value 90, based on an interview with staff,
and the NA case has a value 58, based on 4 of 7 North American
respondents to the CIGRE study indicating that profit was the main goal of
the company.

5.3.5 UK utility case study

Within Great Britain there are four utilities with responsibility for transmission.
In England and Wales the National Grid Company has ownership and
management of the transmission system, and is also system operator. In
Scotland there are two vertically integrated utilities, Scottish Power and
Scottish and Southern Energy. Northern Ireland Electricity operates as a
vertically integrated utility with transmission, distribution and retail functions
(but not generation). In this case study the concentration will be on National
Grid Company, for which considerable published information is available.

5.3.5.1 Description of the National Grid Company, UK case study

The National Grid Company (NGC) was formed when the Central Electricity
Generating Board was split into separate generation, transmission and
distribution companies, following the passage of the 1989 Electricity Act,
effective from 1 April 1990.

Ownership of the National Grid Company was initially by the twelve Regional
Electricity Companies (RECS), but in 1995 the UK government required the
RECS to divest themselves of their shares. NGC became a publicly traded
company in December 1995. As a business NGC has been very successful:
it is part owner of transmission networks in Argentina and Zambia and owner
of a transmission and distribution company in the New England, USA (NGC,
USA). It also has interests in interconnectors in Australia, the USA and the
UK. In 2002 NGC merged with Lattice, a holding company for the England
and Wales gas transmission company Transco, to become NGC Transco.

195
NGC, UK, is the part of this business that owns and operates the electricity
transmission network of England and Wales. It is a regulated monopoly
operating under a price capped incentive-based (RPI – X) regulation. The
regulation scheme has generally been seen to be successful in reducing
operations and maintenance costs (EIA, 1997).

The UK transmission system is geographically compact, but extensively

interconnected. Transmission is at 400kV and 275kV. Land for transmission
development is very difficult to obtain, but transmission loading is largely
moderate, and load growth is fairly low – 2 to 3 % per annum (1.4% at time of
CIGRE study). A large proportion of assets is located on land owned by
others (Jay and Williams, 2001). Maximum demand in 2001 was 52GW
(Allison and Jay, 2001). Generation constraints sometimes occur as a result
of congestion on the five main north-south transmission corridors, but there is
significant spare capacity in other parts of the system. It is estimated that the
network could support 125GW of load if generation were appropriately
located (Allison and Jay, 2001). The difficultly of extending the existing
system has forced NGC to address options for increasing equipment ratings,
and to look for innovative solutions to maximise the utilisation of existing
assets (Jay and Williams, 2001).

The first ten years after electricity deregulation saw the end of centrally
planned generation and the proliferation of privately-owned generators. NGC
is required to connect these generators into the transmission grid within
ninety days(Urwin, 1999). The new generation plant, largely combined-cycle
gas-fired, has been installed with relatively low lead times, sometimes
replacing other more expensive generation. The notice for closures of
generation is even shorter – a minimum of six months. These changes in
generation patterns have resulted in changes to loading and reactive support
requirements on the network. This has led NGC to install significant amounts
of reactive compensation equipment to manage changing voltage support
requirements (Urwin, 1999). To manage the rapidity of changes in voltage
support requirements they have developed transportable Static VAR
Compensators (SVCs). Between 1990 and 1998 some 6000 MVAR of

196
capacitors and 3300 MVAR of SVCs were installed on the network (Urwin,
1999).

The network originally installed by the Central Electricity Generating Board

(CEGB) was conservatively designed, and has been robust in performance.
The age profile of network equipment shows a significant peak in the late
1960s to early 1970s. NGC has considered it necessary to spread out the
replacement program to avoid massive capital expenditure over a short
period (Allison and Jay, 2001). This means that some equipment has been
replaced before it is fully depreciated. There have been problems with
replacing connection assets early – such as transformers supplying the
Regional Electricity Companies: customers have been paying based on the
depreciated value of the asset, and refuse to pay on the basis of a new asset
value when the asset is replaced early, arguing that the equipment was
replaced unnecessarily (Allison and Jay, 2001).

Load is much higher in winter than summer owing to heating loads. Most
outages are planned over the summer period. Very little live-line working,
and no live substation work is done, partly because of health and safety
considerations, and partly because of the extra time required for doing the
job under live conditions. However, it is gradually becoming more difficult to
take circuits out of service for maintenance, because of increased loading on
circuits generally.

While the climate is not usually severe, there are occasionally severe storms,
and high levels of faults are often associated with these. Salt pollution is
probably the worst climatic/geographic issue. Sometimes silicone coatings
are put on insulators to reduce flashover and sometimes substations are built
in-doors to protect from pollution.

Replacement is done based on end of planned life. NGC uses a piece of

software called ALERT to plan the likely capital expenditure program for the
next 25 years, as a result of plant reaching its end of life, based on an
estimate of failure probability density function of all plant on the system

197
(Urwin, 1999). Refurbishment is done in order to enable plant to reach
planned life, and mainly only on pressurised head air blast circuit breakers,
which need a mid-life refurbishment to achieve planned life (Allison and Jay,
2001).

Considerable effort has been put into analysing maintenance activities

including
• Analysis of optimum maintenance periods taking into account failure
rates and outage costs (using more than fifteen years of accumulated
data); and
• Optimising circuit end maintenance (using optimum maintenance
periods for individual equipment).
These activities are time consuming, so tend to be done on items where
significant gains can be expected, eg disconnectors. Nevertheless, at the
time of the CIGRE survey NGC reported doing more maintenance than
recommended over a wide range of equipment types.

In 2001 a program called Rule Based Schedule Evaluation (RBSE) was

introduced to examine the relevance of items within maintenance policies.
These policies are very detailed and exist on each model of equipment. The
RBSE is based on an Access database. It enables each activity in the
maintenance schedule to be examined and linked with specific failure modes,
deterioration rates and impact, to gauge whether its current frequency of
maintenance is appropriate, and whether or not it is needed. It can also be
used to flag work that need not be done in an outage. It is expected that this
will reduce data collection and condition data analysis (eliminating the work
of 22 engineers), reduce the amount of work done during outages, and
reduce protection maintenance by 95%. The current situation is that some
maintenance is not being done because of insufficient staff. It is hoped that
this effort will eliminate unnecessary maintenance activity, enabling all
required maintenance to be done.

There is an equipment health assessment carried out for each type of

equipment each year, broken down to each model. It includes failure rate for

198
the year. The information is used as a reference source to support the
capital replacement plan, including replacement priority and for the
improvement of asset management policy. Work has been undertaken to
improve the efficiency of the asset health review process, and to translate the
knowledge gained into more accessible formats. One initiative to improve
the quality of data is the use of hand-held computers for logging of
inspections; data transfer directly to the maintenance database.

NGC has worked with manufacturers to encourage online condition

monitoring to be built into equipment. Their preferred option is for intelligent
monitoring that informs the user when an action is required, rather than a
system which feeds back a continuous stream of data (Jay and Williams,
2001). Fitting online condition monitoring retrospectively is considered to be
seldom cost-effective. Most of the information gathered by NGC about its
assets is by means of testing and inspection (Jay and Williams, 2001). This
includes such things as helicopter inspection of lines using infrared surveys
and oil analysis of oil-filled equipment. NGC also records information on
faults and defects in a database.

Table 30 summarises the input parameters derived for this case study.

Table 30 Input parameters for the UK utility case study

Input Parameter Value
Age as replacement criterion insignificant (30)
Attitude to new technology aggressive (65)
Business acumen high (70)
Climate temperate
Climatic factor mild (30)
Combining of maintenance activities high (65)
Connectivity factor moderate (17)
Cost of land moderate (60)
Credit rating A (55)
Debt level moderate (55)
Earth quake zone No (10)
External barriers high (80)
Flexibility low (31)
Frequency of failure as replacement criterion very significant (70)
Geographic factor low (3)
Historical revenue sufficiency not a concern (80)
Investment cost as KPI insignificant (30)
Labour cost medium $US 24 x1000/annum
Labour skill high (70)
Live working low 0%
Load factor high 67%
Load growth low 1.4%
Losses (T&D) low 7%
Maintenance history well-maintained (75)
Maintenance relative to manufacturers' recommended level more than recommended (80)
Maintenance spend as KPI significant (53)

199
Input Parameter Value
Number of failures as replacement criterion significant (58)
Pilfering low (10)
Profit as Intention Yes (80)
Proportion of aged equipment low 2%
Proportion of young equipment moderate 44%
Redundancy moderate (n-1) (50)
Regulatory support reasonably supportive (65)
Replacement considered an investment very significant (100)
Revenue sufficiency not a concern (80)
Rugged terrain not rugged (20)
Size of company large (70)
Special relationship with vendors no use (30)

5.3.5.2 Results from the NGC, UK case study

The most noteworthy result from Table 31 is the high level maintenance
spending predicted by the model. A major contributor to this is the high
routine maintenance cost. The level predicted is higher than the typical value
for Western Europe predicted by the model. Inspection of individual results

Table 31 Results of the UK utility case study

Output parameter Value
Availability high (98-100%) (87)
Construction outages low (10)
Corrective maintenance costs low (10)
Cost of finance low (10)
Debt repayments low (17)
Equipment replacement low (10)
Ease of raising capital easy (90)
Factors increasing asset costs low (40)
Failure rate low (10)
Failure rate contribution of load & climate low (10)
Financial drivers for maintenance spending high (90)
Financial incentive replacement high (90)
Investment cost low (22)
Investment in new equipment low (10)
Loading moderately loaded (30)
Maintenance factors increasing availability moderate (48)
Maintenance spending high (80)
Proportion poor condition equipment low (16)
Purchasing efficiencies moderate (36)
Refurbishment low (5)
Refurbishment costs low (5)
Reliability high: 99-100% (77)
Replacement costs as maintenance low (10)
Replacement on age low (20)
Replacement on condition low (20)
Replacement on failures low (20)
Routine maintenance costs high (90)
Spending limits (maintenance) moderate (50)
System Augmentation low (10)
System factors reducing maintenance costs low (25)
Use of offline condition-based maintenance high (70)
Use of online condition-based maintenance moderate (30)
Use of predictive maintenance moderate (63)

from the CIGRE study7, confirms that the maintenance spending (at the time
of the survey) was high compared with the average for Western Europe.

7
filtering on the basis of {region = WE; size = large, ownership = private; functions = transmission only}

200
Note that the case description above suggests that NGC are taking steps to
reduce routine maintenance costs (eg through use of the RBSE tool and
maintenance optimisation studies).

Figure 32 shows that maintenance costs for the NGC are very sensitive to
the efficiency with which routine preventive maintenance is performed. Thus
the measures being implemented to improve maintenance efficiency should
be very effective in reducing costs in this case.

5.4 Use of the model for strategic decision-making

The example from the previous section suggests potential for use of this
model in strategic decision-making. The model cannot be used to assess the
cost of rival proposals, but it can indicate if a proposed strategy is likely to be
effective, and whether there are likely to be negative side effects or additional
benefits associated with the action.

100
80 maintenance
spending
60
40 routine
maintenance
20 cost
0
0 20 40 60 80 100
maintenance relative to manufacturers'
recommended levels

Figure 32 Sensitivity of routine maintenance and total maintenance spending to

maintenance relative to manufacturers' recommended levels

To illustrate this premise reliability will be examined for the TNB case: for this
case study the model suggests that there may be some scope for improving
reliability of supply. Reliability of supply is a function of system design and

201
equipment reliability. In the model there are four attributes for the task
reliability: flexibility, connectivity, redundancy and failure rate. Failure rate is
also a task; it has four subtasks as inputs – failures load and climate (the
contribution of load and climate to failure rate), use of predictive
maintenance, proportion of poor condition equipment and spending limits
(maintenance). This last factor only comes into play if maintenance spending
is severely limited, which not the true in this case. The model calculates that
use of predictive maintenance is low for TNB. If use of predictive
maintenance is increased then this may reduce failure rates and hence
improve reliability of supply. Figure 33 suggests that only a modest increase
in reliability is achievable by increasing use of predictive maintenance in this
case. However, it may be worthwhile to pursue this, if the cost of
implementing predictive maintenance is not great. It is clear that, in this
case, other factors limit the effectiveness of this strategy. Likewise, reducing
proportion of poor condition equipment has little appreciable effect on
reliability.

100
reliability - model

90
80
70
60
50
0 20 40 60 80 100
use of predictive maintenance

Figure 33 Impact of use of predictive maintenance on reliability for TNB case

202
 80
reliability - model
60

40

20

0
0 20 40 60 80 100
failure contribution of load & climate

Figure 34 Sensitivity of reliability to failure rate contribution of load and climate for
TNB case
Figure 34 suggests that the failure rate contribution of load and climate has
more effect on reliability for this case. The current point of operation (failure
load and climate = 46, reliability = 64) has been included as a point on the
graph. For values above the current value the model predicts a rapid drop in
reliability, but reducing this value to 40 or below, increases reliability only
slightly to 66, or the moderate-high level. This analysis suggests that there
are few opportunities to make major improvements in reliability from
strategies that affect equipment performance in this case.

Investigation of the system-design-related parameters shows that both

flexibility and connectivity are low, while design-redundancy is average.

203
 40
36

high
32
28 connectivity
(normalised

moderate
24
value 1 to
20
100)
16
x
12 85-95

low
8 75-85
4 65-75
10

30

50

70

90
55-65
low moderate high
45-55
flexibility (% )

Figure 35 Sensitivity of reliability to flexibility and connectivity for TNB case

Analysis of Figure 35 suggests that a large increase in both flexibility and
connectivity would be required to achieve a substantial improvement in
reliability of supply. Current levels of flexibility and connectivity are shown as
an X on this diagram. Changing levels of flexibility and connectivity are long-
term policies – they are expensive to implement in existing network.
However, connectivity will increase with addition of circuits; this tends to
occur automatically as the network matures. Increasing flexibility, on the
other hand, requires a change in substation design strategy. There is some
evidence to suggest that interconnection between networks can also have a
positive impact on reliability of supply. This has not been factored into the
fuzzy model but will be investigated further in Chapter 6. In summary, the
model shows limited opportunities in the short term for improvements in
reliability of supply for TNB.

Circuit availability is of interest to Australian utilities at present because the

regulator (ACCC) has recently introduced performance incentives based on
circuit availability. The circuit availability for Powerlink (and other Australian
utilities) is already high by world standards. Nevertheless the regulator is
offering incentives to increase availability, and penalties for reduced
availability.

204
The task availability contains four attributes, which are listed in Table 32 with
associated values for the Powerlink case.

Table 32 Factors contributing to availability and their values for the Powerlink Case
Attribute Value
construction outages low (27)
failure rate low (10)
flexibility low (34)
maintenance actions increasing availability moderate (57)
Inspection of this table suggests that there may be some scope for gains by
increasing flexibility and/or by increasing the level of the task maintenance
actions increasing availability. The factors that contribute to the maintenance
actions task and their values are shown below in Table 33.

Table 33 Maintenance activities increasing availability for the Powerlink Qld Case
Attribute Value
combining maintenance activities( by circuit) medium (63)
live working low 20%
use of online CBM moderate (43)
use of predictive maintenance high (71)

90 maintenance activities
low moderate high.

70
by circuit %
combining

x
50

30 87-92

10 82-87
10

30

50

70

90

77-82
low moderate high
72-77
live working %
67-72

Figure 36 Availability as a function of live working and combining maintenance

activity by circuit. Current operating point is shown as X.
Figure 36 demonstrates that either increasing the level of combined
maintenance activities by circuit or live working will have a positive effect on
availability. From the graph, the combining of maintenance activities appears
to provide potentially greater gain for a small increase in parameter level. It
is also probably the less expensive option as live working may be slower and

205
undoubtedly increases risk to workers. As the discussion for this case
explained, Powerlink is now implementing both these strategies.

100
low moderate high.

80
availability

60
40
20
0
0 20 40 60 80 100
low moderate high
use of online CBM

Figure 37 Sensitivity of availability to use of online CBM for Powerlink Queensland

case study. The calculated value of use of online CBM for current parameters is
moderate (43).

Figure 37 illustrates the predicted effect of increasing the use of online CBM
on availability for the Powerlink case8. Online CBM also contributes to
predictive maintenance in general. Increases in online CBM and predictive
maintenance both have potential to increase circuit availability. In this case,
analysis suggests that sensitivity to this parameter is not high around the
current level (calculated from the model as moderate (43)). Hence,
Powerlink would have to increase its use of online CBM significantly to
achieve a modest improvement in availability. Since the predicted value of
online CBM is fairly high by world standards, increasing online CBM
significantly might prove to be an expensive exercise, if it involves retrofitting
condition monitoring to older equipment.

8
Note that the wavy effect of this graph is a function of the use of fuzzy rules. In reality the
graph is likely to be smooth.

206
 100

low moderate high

80
availability
60
40
20
0
0 20 40 60 80 100
flexibility %

Figure 38 Availability as a function of flexibility for the Powerlink case

Figure 38 shows weak sensitivity of availability to flexibility for the Powerlink

case. Powerlink’s new substations, employing hybrid-switchgear, exhibit a
lower level of flexibility than older designs of substations. However, Figure
38 demonstrates that a reduction in flexibility (from the current level of 38%)
is unlikely to have a significant detrimental effect on availability in this case.

5.5 Summary of results from fuzzy modelling

In general the individual case studies have exhibited a higher degree of
consistency between observed and calculated performance than was found
for the regional cases. This is largely because the data used for modelling
are more consistent.

The examples in Section 5.4 illustrate several important features of the fuzzy
AM model:
• The fuzzy model can be used to identify options for performance
improvement.
• It can give an estimate of the likely effectiveness of different options,
given existing conditions. The time-frames for these options will also
range from short- to long-term.
• It can identify barriers to improvements in performance.

207
The model demonstrates that performance among transmission utilities, and
their ability to improve it, may be limited by existing factors such as design of
existing network, and the rate at which the network can be modified.
External factors like climate type and severity may also impact on potential
best performance. The conclusion that can be drawn from these studies is
that solutions cannot necessarily be translated from one utility to another:
options must be analysed in the context of existing conditions for each
enterprise.

208
6 Investigations using system dynamics modelling
In Chapter 4 system dynamics modelling was identified as a potentially useful
tool for investigating, in detail, certain core elements of the conceptual AM
model constructed in Chapter 3. Analysis of the causal loop diagram (Figure
15) showed that there were complex feedback loops involving maintenance,
failure rate, replacement and refurbishment. The relationship between
maintenance and failure rate was found to have a time lag, dependent on
rate of the deterioration of equipment condition. Because of the feedback
loops and time-dependencies, these types of relationships are difficult to
model with a fuzzy rule-based model, but are more easily represented using
a system dynamics model.

System dynamics modelling is a well-established methodology, adapted from

engineering (Pidd, 1998 p229), that has found application across a wide
range of disciplines from business management (Forrester, 1961, Senge et
al., 1994) (Pidd, 1998 pp253-270) to economics (Saeed, 1998), social policy
(Sudhir et al., 1997, Huz et al., 1997) and education (MIT Systems Dynamics
Group, 1990). System dynamics models have been used in practice to
(Senge et al., 1994 p173):
• Show how system structures produce patterns of behaviours;
• Test whether or not a structure replicates observed behaviour;
• Explore how behaviour changes when different aspects of the
structure are altered;
• Identify points of leverage for performance enhancement; and
• Promote understanding of systems thinking, and learning by
experimentation.
In system dynamics modelling the structure of the model is the primary focus
(Pidd, 1998 p230). The main elements of system dynamics models are
levels (also called stocks) and rates (also called flows). The system
dynamics method views feedback systems as interconnected sequences of
levels and flows (Pidd, 1998 p233). Delays are also often present in feedback
systems. Policies (Pidd, 1998 p234) in management science, may be

209
expressed in terms of how rates affect levels and levels affect rates. In
systems dynamics, models tend to be built at an aggregate level with rates
regarded as continuous through time (Pidd, 1998 p230). For example, in this
study we deal with asset populations, their distributions and failure rates
rather than individual assets.

In this chapter, models are developed, using the system dynamics software
package Stella Research (High Performance Systems Inc, 2002), to examine
the relationships between asset population age- and condition-distributions,
and maintenance, refurbishment and replacement policies. The effect of
maintenance and replacement strategies on asset value, revenue and profit
will be examined, assuming straight line depreciation, and revenue cap
regulation in the form applied to Australian transmission utilities (ACCC,
1999a). The impact of asset management policies on network performance
will also be studied in terms of failure rate of equipment.

6.1 Simulation in system dynamics modelling

Forrester (Forrester, 1961) pioneered the simulation of socio-economic
feedback systems in 1961. In system dynamics modelling, the relationships
in the model are represented as a set of difference equations that are solved
by numerical integration. Typically the Euler-Cauchy method of numerical
integration is used (Pidd, 1998 p 244), but Stella (High Performance Systems
Inc, 2002) offers the options of using 2nd order or 4th order Runge-Kutta
numerical integration techniques. Euler’s method is recommended (High
Performance Systems Inc, 2002) for models employing integer arithmetic, but
Runge-Kutta techniques give more accurate integration for continuous
systems, particularly those with oscillatory tendencies.

Models employing numerical integration methods can be sensitive to the

choice of the integration interval dt. It is therefore necessary to check that
the choice of integration interval does not result in dynamic behaviour
patterns that are artefacts of the calculation process, or that result in
significant inaccuracies in calculation. One simple way of doing this is to
compare the model outputs for the current choice of dt and a value that is

210
half the current value of dt; if there is a difference then dt is too large. The
Stella program documentation (High Performance Systems Inc, 2002)
suggests this method, and also that dt should not be larger than 0.5 the
shortest delay in the model.

For the current application the 4th order Runge-Kutta numerical integration
method has been employed. A dt of 0.25 (years) was found to produce
satisfactory results.

6.2 Model verification and validation in system dynamics modelling

As indicated in Section 5.1.3, verification is the process of ensuring that the
implementation of the model is properly realised in the computer program
(Pidd, 1998 p157). In this study it has been greatly facilitated by the use of a
visual interactive modelling system (VIMS) which employs a graphical user
interface (GUI). The user of a VIMS draws the system using predefined
icons, and supplies the necessary parameters for running the simulation. In
Stella the modeller specifies the flow equations, and initial conditions for
stocks, but the program automatically generates the stock equations. In the
present study, verification has been managed by building the model in
sections and testing each section in turn, before linking the models. This
development process has been paralleled in the presentation of the model
and its variants within this chapter.

One problem that has become apparent, during the verification process, is
that of prioritisation of multiple flows from a stock that is not permitted to
become negative. If the model calls for greater outflows from a stock than the
current level plus inflows, then the Stella program assigns outflows according
to its internally generated priority levels. The internally generated priorities
do not necessarily match the priorities required for the model. To overcome
this problem it was found necessary to override the program, by explicitly
apportioning outflows, in cases of limited supply. An example, in which this
measure was necessary, is for retained earnings (Section 6.10). In the
model there are outflows from retained profit of interest repayments,
dividends, capital funding of assets and a flow to adjust the debt level so that

211
the intended gearing ratio is maintained. It was found necessary to force the
program to assign the combined revenue plus retained earnings to cover
interest payments first, capital payments second, adjustments to the debt
level third and dividends last. Without these modifications capital funding, for
example, was given higher priority by the program than interest repayments.

The validity of a model is judged on its suitability for its intended purpose
(Forrester, 1961): a model that is useful for one purpose could be misleading
if used in a manner for which it was not intended. Coyle (Coyle, 1977 p181)
states that establishing model validity is “unquestionably one of the most
difficult areas in management science.” He defines validation (Coyle, 1977 p
181) as the process by which we establish sufficient confidence in a model to
be prepared to use it for some particular purpose.

Ideally in order to validate a model one would compare it directly with the
system that it simulates. However, Coyle contends that, in practice, it is
seldom possible to make a change to a real system and observe its effect: in
management systems it is very difficult to isolate the effect of policy changes
from other things that have happened to the system. Therefore true
validation of systems dynamics applications in management science is not
possible (Coyle, 1977 p182), and the best that can be achieved is a
confidence-building exercise. Coyle proposes that the system dynamics
modeller address the following series of questions (Coyle, 1977 p182):
• Are the system boundaries correct? The model should include those
elements that can alter behaviour, but keep in mind the purpose of the
model.
• Are there any gross errors (for example, negative values of items that
should only be positive or values that are beyond reasonable
bounds)?
• Is there a correspondence between the model structure and the
system? The decisions functions in the model should reasonably
reflect those actually used.
• Are the parameter values correct? Coyne suggests that this may not
be a very important consideration, as the dynamics of a system are

212
 usually not affected much by most of the parameters, provided they
are within a fairly broad range. However there may be a few
parameters that are critical to the dynamics of the system.
• Does the model reproduce system behaviour sufficiently well? Coyle
states that in practice it is seldom possible to test this: data on policies
are scarce and rarely are the policies consistently applied. The
available data will also never cover all model outputs so many
variables must remain untested. Coyle advises that it is seldom
justified to attempt to apply statistical verification of system dynamics
models.
In this particular application, the purpose of the model is to examine the
impact on system and financial performance of different policies for
maintenance and replacement in transmission systems.

To facilitate this investigation, a highly aggregated model is proposed, that

considers the problem in terms of a generic asset population. The asset
population is represented in two forms: the first is by age distribution, and the
second by distribution on the basis of condition. Failure rates are applied in a
manner that approximates the failure characteristics of mechanical
equipment (see Section 6.3). The use of an age distribution of the asset
population allows estimation of the effect of various policies on total
depreciated value of assets, revenue and profit. Representation of the asset
population in terms of condition enables examination of the effect of various
maintenance and replacement policies on the distribution of assets. By tying
these two representations together, it is possible to explore the impact, of
various maintenance and replacement policies, on financial performance of
the enterprise. Therefore, while it is possible to conceive many extensions to
this model, its boundaries are adequate for the purpose of the study. While
the results are numeric, they are merely indicative, and are intended to be
used only for the purpose of comparison of policies. Various approximations
have been made throughout the development of the model and its variants:
the effects of these approximations on the model outputs are discussed, in
their context, in relevant sections of this chapter.

213
Any gross misrepresentations, (such as those arising from the prioritisation of
outflows, as described above) have been corrected during the development
of the model. Testing and verification, as recommended by Pidd (Pidd, 1998
p158), have been undertaken throughout the development of the model.
Likewise, validation of sections of the model has been undertaken as an
integral part of the development process. For example, in Section 6.15,
addition of augmentation at a constant rate of 3% per annum (a value not
untypical of mature networks today) was found to increase the number of
assets on the system, in the longer term, to an impracticable level. This
prompted consideration of the relationships between load growth, system
maturity, size and number of assets. It also led to insights about the effects,
of system augmentation level and system maturity, on asset management
policies employed by transmission enterprises in developing countries
compared with those in developed countries.

The main parameters used in the model are as follows:

• Failure rates for different types of failure mode are included – overall
failure rates are checked against typical data and sensitivity of
outcomes to data are investigated;
• The relationship between (routine) maintenance effectiveness and
degradation rate of assets from good to poor condition is modelled
graphically;
• The relationship between use of predictive maintenance and
proportion of detected poor condition assets is also modelled
graphically;
• Weighted average cost of capital (WACC) – real data has been used,
based on a recent revenue cap decision (see Section 6.10);
• Interest rate – real data has been used based on the calculations used
to determine WACC in the same revenue cap decision (see Section
6.10);
• For gearing ratio, defined as the ratio of debt to debt plus equity the
benchmark gearing ratio from the same revenue cap decision has
been used (see Section 6.10);

214
 • Dividends are modelled, but levels have been set to zero.
Comparison is made on the basis of retained earnings as this avoids
having to choose a level for dividends.
• The relationships between network maturity and load growth has been
estimated, based on published load growth figures and estimated
maturity (modelled graphically) (Section 6.15);
• Relationship between asset average value and network maturity
(modelled graphically) (Section 6.15);
In this model the age-related failure rates are generated based on the
condition-related failures, and rules described in Section 6.6. Failure rates,
as a function of age for different types of transmission equipment, could
potentially be obtained (with some effort) from maintenance databases of
transmission enterprises. However, failure rates as a function of condition are
likely to be more difficult to acquire, unless the utility has developed condition
indices for its equipment (for instance, of the type suggested in (Balzer et al.,
1998)). Actual values will vary with the type of equipment being examined.
Since a generic asset has been used for this study it was not considered
necessary to attempt to find particular values. If this model were applied to a
particular asset class, in an extension to this research, then it would be
appropriate to model the parameters more precisely. Instead, the overall
failure rate predicted by the model is compared with some published
information on transmission equipment (see Section 6.6.1).

Maintenance effectiveness is a function of quality and quantity of

maintenance. Quality of maintenance is determined by the degree to which
routine maintenance is directed towards prevention of degradation and
correction of minor incipient failures. Quality also relates to the skill with
which the maintenance is performed. On the other hand, for maintenance to
be effective it also needs to be performed with sufficient frequency to
minimise degradation – this is the quantity element. Actual maintenance
actions, and the degree to which degradation can be controlled through
maintenance will vary with the type (and design) of equipment. (See also
Section 3.2.1.3 on the relationship between maintenance history and failure
rate)

215
The relationships between maintenance effectiveness and degradation, and
predictive maintenance and detection of poor condition assets are more
difficult to validate, as no data currently exists for these types of relationship.
Justification of the shape of the relationship curves has been attempted by
explaining the reasoning behind choice of shape (see Section 6.5).
Likewise the relationships between network maturity and load growth and
network maturity and average asset value are estimates of parameters for
which no data currently exist.

Does the model reproduce system behaviour sufficiently well? Because of

the nature of the equipment (long-lived and robust) the dynamics of the
system tend to be slow. Policies and practices by contrast tend to change
more quickly, and this has especially been true of the past twenty years when
significant electricity industry reform has occurred. The best evidence that
can be offered, in support of the correspondence between the dynamics of
the model and the real system, is that the model seems to support and
explain some of the observed relationships between asset management
policy and performance discussed in earlier chapters.

6.3 Repairable and non-repairable systems theory – application to this

model
The hazard rate (Davidson, 1988 p10) is measure of probability that a
component will fail in the next time interval, given that it has survived up to
the beginning of that time interval. The term is normally applied to non-
repairable components. The well-known bath-tub curve (Davidson, 1988 p
10) (Moubray, 1992 p12) shows a hazard rate that decreases in the initial
period after the equipment is put into service (wear-in phase) a period of
useful life in which failure rate is constant and a phase of increasing failure
rate at the end of life (wear-out). The actual hazard rate will vary with the
type of equipment: electronic components may exhibit fairly constant hazard
rate over a large period of time, but many mechanical components exhibit a
strong wear-out tendency, with increasing hazard rate throughout their
operating lives (Davidson, 1988 p11). The situation in which hazard rate
increases with time, arises when a component degrades, for example
through corrosion, wear or fatigue.

216
For repairable systems a constant failure rate (rate of occurrence of failure
ROCOF) is often assumed. This assumption is valid if all elements in the
system exhibit a constant failure rate, which is generally not the case for
mechanical equipment, or if there is no dominant failure mode, in which case
the mixing effect of replacements and failures leads to an approximately
constant failure rate (Davidson, 1988 p38-9). The assumption of constant
failure rate also requires that all repairs of equipment restore it to “as good as
new condition” (Davidson, 1988 p39). The constant-failure-rate assumption
may be invalid in cases where equipment is replaced by new technology that
is more reliable than the original equipment; this can result in reliability
growth over time. If equipment is not repaired to “as good as new” condition
there will be an increasing failure rate with time.

For transmission equipment the assumption of constant failure rate may not
be valid. For example in the study of circuit breaker failures (Hastings and
Crisp, 2001) the failure rate illustrated in Figure 14 shows an increasing trend
with age. Since circuit breakers have been the subject of significant
technological development of the past forty years (Knobloch et al., 2000) this
may be a combination of increasing age-related failure rate and reliability
growth resulting from technological improvements. The assumption of
perfect repair is also doubtful because of the likelihood of imperfect
maintenance (see also the discussion on this topic in Section 6.8) and the
existence of some failure modes for which repair is not feasible. In analysis
of electrical power system problems an assumption of constant failure rate is
often used for reliability analysis (for example, (Brown, 2001b)), as this
greatly simplifies the analysis. However, in replacement decisions for major
equipment (Hastings et al., 1999) the increasing risk of failure with age may
be taken into account.

In some transmission equipment there may also be a dominant cause of

ageing that produces a non-constant failure rate with time. For transformers,
insulation degradation is the dominant cause of ageing (Bernstein and
Brancato, 1993). The condition of paper insulation is used as a measure of

217
remaining life for transformers (Allan, 1991). Underground cables also exhibit
failures associated with ageing of insulation. Ageing in insulation is a
common cause of degradation in the condition of transmission system
equipment. All power system equipment relies on insulation for continued
operation. The rate of ageing is dependent on the nature of the combined
external stress and the corresponding inherent resistance of the insulating
materials to these stresses (Bernstein and Brancato, 1993). The external
stresses to which insulation is exposed include mechanical forces, electrical
stresses (kV/cm) and temperature. These external stresses are present to
various degrees throughout the life of electrical equipment, so that the
equipment should be expected to show an increasing hazard rate from
insulation related causes with age. To the extent that the insulation is
integral to the equipment (eg in transformers, reactors, capacitors) the
degradation caused by insulation ageing could be considered non-repairable.
By contrast, in overhead lines, insulators can be replaced individually. This
failure mechanism is therefore avoidable by maintenance, and the effect of
replacing failed insulators, in the course of preventive maintenance, may be
to flatten the overall age-related ROCOF (Davidson, 1988 p38-39) from this
cause, for this class of equipment, to a pseudo-constant value. Corrosion is
another major source of failure in overhead lines (both in the conductor and
insulator strings) (Aggarwal et al., 2000). Degradation of mechanical
properties of transmission lines may also be exhibited, in association with
corrosion or heat related effects. While corrosion is repairable to some
extent, it is difficult to repair to as good as new condition, in which case an
increasing ROCOF could be anticipated (as discussed above).

It is therefore arguable that modelling of non-constant failure rate with time is

appropriate: the treatment of transmission equipment in modelling will
depend on the purpose of the modelling exercise.

For the present study an allowance is made for the possibility of increasing
failure rate (ROCOF) with age. The model assumes that
• There is a relationship between the age and condition of equipment;

218
 • Preventive maintenance can affect the rate of degradation of overall
condition of assets (but cannot eliminate degradation entirely);
• Refurbishment can restore equipment to “as good as new” condition;
• Repair related to a particular failure condition will not alter the overall
condition of the asset.

6.4 Modelling population ageing and asset value

For this study a single asset type is used, with value 1 unit. A population of
1000 units is assumed. Initially the total number of assets is held constant:
there is no system augmentation, and each asset taken out of service is
replaced by an identical one, with no time delay. This allows the impact of
various practices to be assessed without the obscuring effect of additional
assets.

The population of assets is modelled using a structure called the main chain,
which is commonly used for any sort of ageing process (Richmond, 2001
p101). In this simple model (Figure 39) there are four states, corresponding
to young (YA), mid-life (MLA), aged (AA) and very aged assets (VAA). It is a
feature of the main chain structure that, in steady state, the asset population
will distribute itself in proportion to the average residence time associated
with each state (stock)(Richmond, 2001 p101)9.

In Figure 39 large rectangles represent stocks of items, and the solid arrows
indicate the direction of flows. The circles represent ordinary variables (also
called converters), the values of which may be constant or calculated.
Circles with plus signs denote summers. The summer is a short-hand way of
indicating a summation of a number of inputs. Dotted lines show
relationships. For instance, the flow rate between young assets and midlife
assets, is defined as (young assets 0 to 9 years)/(young yr range), where
young yr range is the number of years represented by the range, in this case

9
It is noted that there are strong parallels between the main chain structure in system
dynamics modelling and Markov modelling. The Markov transition probabilities are similar to
the failure rates and rates of transition from one state to another in the system dynamics
formulation.

219
a constant, 10. Similar relationships are defined for the other flows between
states.
+ Sum of
all replaced MLA AA VAA
replaced replaced replaced
YA yr range MLA yr range AA yr range

young assets mid life assets aged assets very aged assets
(0 -9 yrs) (10-29 yrs) (30-39yrs) (40+ yrs)

MTBF
early MTBF MTBF MTBF MTBF
early random random random random
randomly VAA
failing randomly randomly MLA MLA failing failing
randomly
failing failing failing failing
failing
+
Sum all MTBF
MTBF MTBF
failing VAA
MLA AA

Figure 39 Simple main chain showing ageing process

Failures of assets can occur in any stage of an asset’s life. Random failures
can occur at any time. Infant mortality failures (early failing in the diagram)
can occur in the early stages after commissioning. Age-related failures
(wear-out) can occur in later stages. Failure rates are calculated by the
number of assets in a stock (state) divided by the mean time between failures
(MTBF) for each failure type. This version of the model assumes all assets
that fail are replaced (indicated on the diagram as a summer labelled sum of
all replaced). Since the number of assets in each age group becomes
constant in steady state, the overall failure rate is constant when the system
reaches steady state.

Including a calculation of asset value based on straight-line depreciation is

simple if the average age of assets in each age group is assumed to be the
average of the age-range for the group (Equation 8). Straight-line
depreciation is used in determination of revenue for Australian transmission
utilities (ACCC, 2002a, ACCC, 2000). Other depreciation methods could be
applied, but have not been considered in this research; this is a possible

220
avenue for further research. For this study an expected life, for depreciation
purposes, of forty years is assumed.
Total depreciated value = value per asset/life expectancy of assets *
(young_assets__0_to_9_years*(life expectancy of assets- age of young assets) +
mid_life_assets_10_to_29_years*(life expectancy of assets- age of ml assets) +
aged_assets_30_to_39_yrs*(life expectancy of assets – age of aged assets)) (8)
Note that all replacement values are in present-day values. It is assumed for
the purpose of modelling that there is no variation in the real cost of
replacements with time, for replacements which have the same technical
capacity as those assets being replaced.

6.5 Condition deterioration, maintenance effectiveness and predictive

maintenance
In modelling condition it is assumed that an overall measure of condition of a
piece of equipment can be ascertained. Asset management decisions can
then be made on the basis of condition. This approach has been taken in a
number of recent AM research applications including (Balzer et al., 1998,
Hoskins et al., 1998, Hoskins et al., 1999, Marshall, 1997).

For this application will be assumed that assets are in either good condition
(GCA) or poor condition (PCA). These states are connected using a main
chain structure (Figure 40). Equipment can fail randomly and assets in poor
condition can fail from condition-related causes. For example, a transformer
with poor condition paper insulation has a higher probability of failing if a
lightning fault occurs on a connected circuit, than if the same fault occurs
near a transformer in good condition (also depending on the fault current). In
this diagram the flow associated with early failing is shown as a function of
GCA. It is, however more correctly modelled as a function of the age of the
asset. In the following sections the linkage to the age and condition models
is described, and this will be corrected.

The model includes the possibility of replacing or refurbishing poor condition

assets. Depending on the replacement strategy used good condition assets
may also be replaced.

221
 fraction
PCA refurbished

sum all replacing fraction

replacing refurbishing
GCA GCA replaced
PCA
fraction
+ PCA replaced
fraction
deteriorating

good condition poor condition

assets (GCA) assets (PCA)
adding replacing
deteriorating
PCA
+

sum all
failing
GCA randomly MTBF randomly PCA
early failing random failing failing MTBF
failing
PCA
MTBF GCA
early failure

Figure 40 Simple model of deteriorating condition, failure, refurbishment and

replacement
The model can be extended to include preventive maintenance effectiveness
and use of predictive maintenance. Maintenance effectiveness, (also
discussed in Sections 3.2.1.3 and 6.2), refers to both the quantity and quality
of maintenance undertaken. The purpose of preventive maintenance is to
detect and correct incipient causes of failure before failure occurs (Davidson,
1988 p12). It includes such things as lubrication and alignment of moving
parts, and ensuring that oil quality and quantity are kept at correct levels. For
transmission lines it could include detecting and replacing cracked or broken
insulator sheds. These are activities that tend to inhibit the degradation of
equipment condition. The rate of deterioration of condition may therefore be
considered a function of preventive maintenance effectiveness10.

Similarly the rate of refurbishment or replacement of poor condition assets is

dependent on the detection of poor condition in assets. The ability to detect
poor condition assets is a function of level of usage of predictive
maintenance. The proportion of PCA replaced is calculated by the product of
the fraction of PCA detected and the fraction of detected PCA replaced.
These concepts are included in the model in Figure 41.

10
It is also a function of other factors such as loading and duty as discussed in Chapter 3.

222
 fraction detected
PCA refurbished

sum all replacing fraction

replacing refurbishing
GCA GCA replaced
PCA
maintenance
fraction detected
+ effectiveness
PCA replaced
fraction
deteriorating ~
fraction
good condition poor condition PCA
assets (GCA) assets (PCA) detected
adding replacing
deteriorating ~
PCA
+

sum all
failing
GCA randomly MTBF randomly PCA
early failing random failing failing MTBF use of
failing
PCA predictive
MTBF GCA maintenance
early failure

Figure 41 Condition deterioration model with deterioration rate a function of

maintenance effectiveness, and refurbishment and replacement of poor condition
equipment a function of predictive maintenance.
The relationship between use of predictive maintenance and the fraction of
PCA detected has been represented graphically in the model. In Figure 41
relationships expressed graphically are indicated by the tilde (~) symbol. In
the context of this model, the predictive maintenance is assumed to be well-
targeted to the failure modes exhibited by the equipment.

Since no data are available for this type of relationship at present, an

approximate relationship (Figure 42) has been developed by reasoning. The
non-zero value of detected PCA for 0% use of predictive maintenance
reflects the fact that a proportion of the conditions likely to cause failure will
be detected in routine preventive maintenance. The flattening of the curve at
high levels of predictive maintenance usage implies a type of diminishing
returns effect: it is considered unlikely that all conditions leading to failure will
be detected. The actual value of fraction of PCA detected at any particular
use of predictive maintenance level will depend strongly on type of
equipment being examined.

223
 1

fraction of PCA detected

0.8

0.6

0.4

0.2

0
0 20 40 60 80 100
Use of predictive maintenance %

Figure 42 Graphical representation of relationship between PCA detected and use of

predictive maintenance

The relationship between maintenance effectiveness and fraction

deteriorating has also been defined graphically as shown in Figure 43. The
function is designed to give deterioration rate of 10% of assets per annum if
maintenance is totally ineffective (eg maintenance not undertaken), and 2%
per annum for perfectly executed maintenance. (See also section 3.2.1.3 for
further discussion of the relationship between maintenance, condition and
failure rate). The non-zero crossing of the y-axis reflects the fact that some
failure modes are not amenable to preventive maintenance. For example,
depolymerisation of paper insulation (in transformers) is not considered a
reversible process: the remaining life of transformers is often estimated in
terms of the condition of the paper (Allan, 1991, Allan, 1993). By contrast
quality and quantity of oil insulation may be maintained through maintenance.

224
 fraction deteriorating
0.12
0.1
0.08
0.06
0.04
0.02
0
0 0.2 0.4 0.6 0.8 1
1- maintenance effectiveness

Figure 43 Graphical representation of the relationship between maintenance

effectiveness and fraction deteriorating

6.6 Linking the age and condition models

In order to investigate the effect of maintenance on the asset age distribution
and valuation it is necessary to marry the condition-based model with the
age-based model of the assets. In this initial representation, it is assumed
that replacement and refurbishment are undertaken on the basis of condition.
Slightly different model structures will be developed in later sections, based
on other assumptions.

A number of rules have been applied in developing this model:

• The total number of assets is constant: all failed assets are replaced
with no delay.
• Number of good condition assets failing from infant mortality per
annum is the same as the number of young assets (0 – 9 yrs) failing
from infant mortality. This follows from the fact that infant mortality
occurs in new assets, which are generally considered to be in good
condition. (Note that this corrects the problem in the condition model,
identified in the previous section.)
• The total number of assets failing randomly per annum is the same for
the age-based representation and the condition-based representation.
This is achieved by setting MTBF random to the same value in both
age and condition representations.

225
 • The number of PCA failing is conserved in the age-based
representation. It is assumed that poor condition assets are in the
highest age groups (i.e. failed assets are first taken from very-aged
assets group, then remainder from younger groups in order of
decreasing age.) Figure 44, below illustrated these relationships in a
diagram simplified to maintain clarity.
replacement of very aged assets

replacement of aged assets

very aged
young assets mid life assets aged assets
assets
(0 -9 yrs) (10-29 yrs) (30-39yrs)
(40+ yrs)

early randomly randomly MLA randomly AA randomly VAA

failing failing failing failing failing failing failing failing

failed assets all replaced without delay

replacement of poor condition assets

good
poor condition
condition deterioration
assets (PCA)
assets (GCA)

early randomly
randomly PCA
failing failing
failing failing

all failed assets replaced (no delay)

Figure 44 This diagram shows the links between failure of poor condition assets and
age-related failures.

• For very aged assets

1. The mean time between failures (MTBF) for very aged assets is not
less than the MTBF PCA.
2. The number of very-aged assets failing cannot be more than the
number of PCA failing.

226
 • For aged assets there are also two limiting conditions:
1. The MTBF of aged assets is not less that the MTBF PCA.
2. Number of assets failing per year cannot be greater than the
number of assets in poor condition failing per year less the very-aged
assets failing from the same cause.
• The same applies for mid-life assets except that the number of assets
failing per year is limited by number of assets in poor condition failing
per year less the combined very-aged and aged assets failing per
year.

Similar rules are applied for replacement:

• The number of assets replaced is conserved between the condition-
based and aged-based representations.
• When replacing poor condition assets, these are assumed to come
from the very-aged asset group first, followed by the aged assets and
then mid-life assets. These relationships are illustrated in Figure 45.
• Replacement of assets is limited by the number of assets available in
an age group as well as the total assets replaced. Rate of
replacement of any group of assets (owing to condition) cannot
exceed the rate of replacement of poor condition assets.

Refurbishment is assumed to bring the asset back to good condition. It is

also assumed not to affect the age of the asset for the purpose of
depreciation. In practice, treatment of refurbishment for depreciation
purposes varies: some utilities modify the age of refurbished assets for
purposes of depreciation, while others do not (CIGRE SC 23 and 39, 2000 p
70).

Since the model is combined from two main-chain structures and the total
asset population is conserved, it also reaches a steady state condition.

227
 replacement of very aged assets

replacement of aged assets

replacement of mid life assets

young assets mid life assets aged assets very aged assets
(0 -9 yrs) (10-29 yrs) (30-39yrs) (40+ yrs)

early randomly randomly MLA randomly AA randomly VAA

failing failing failing failing failing failing failing failing

failed assets all replaced without delay

replacement of poor condition assets

good condition poor condition

deterioration
assets (GCA) assets (PCA)

early randomly
failing failing randomly PCA
failing failing

all failed assets replaced (no delay)

Figure 45 This simplified diagram illustrates the links between replacement of poor
condition assets and the replacement of assets from different age groups, for
condition-based replacement.
There are some limiting assumptions in this model.
• The effect of remedial maintenance has not been addressed in this
version of the model – it has been assumed that all equipment that
fails will be replaced. The effect of repairing failed poor condition
assets will be considered in Section 6.8.
• It is also assumed that the transmission enterprise is able to claim
revenue for failed assets that have been replaced. This is a
reasonable assumption, except that if revenue is set in advance (say
for five years) the company may or may not be able to recoup the
revenue for a replacement asset until the next revenue reset.

228
A range of tests was performed to check the validity of the model, in addition
to the rules described previously:
• Total failures per annum must be the same in age and condition-
based representations.
• Total failures per annum from each type of failure (infant mortality,
random and wear-out) must be the same for the age-based
representation and the condition-based representation.
• Total number of assets must be conserved (and equal) for age and
condition-based representations.
The model can be used to investigate the sensitivity of failure rate and asset
age-distribution to various parameters and strategies. A number of possible
scenarios are examined in the following sections.

6.6.1 Impact of maintenance effectiveness on asset population and

value for a condition-based replacement/ refurbishment scenario
To study the impact of maintenance effectiveness on age and condition
distributions of assets maintenance effectiveness is varied while other
parameters are maintained at constant values: use of predictive maintenance
50%; fraction PCA detected being replaced, 0.5; fraction PCA detected being
refurbished, 0.2. This study has been carried out for three values of MTBF
PCA 2, 5 and 10 years (PCA failing fraction is the reciprocal of MTBF PCA).

200
Poor Condition

150
Assets

100
50
0
0 0.2 0.4 0.6 0.8 1
Maintenance Effectiveness fraction

PCA failing fraction p.a 0.5 0.2 0.1

Figure 46 Variation in condition of assets with (preventive) maintenance effectiveness

fraction, and PCA failing fraction.

229
As maintenance effectiveness is increased the proportion of assets in good
condition increases (Figure 46). This means that fewer poor-condition assets
fail (Figure 47) and fewer are refurbished or replaced (on the basis of
condition). As a result the number of aged and very aged assets increases
(Figure 48), and the depreciated value of the asset population is reduced
(Figure 49).

7
6
average failure rate

5
4
% pa

3
2
1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
maintenance effectiveness fraction

PCA failing fraction p.a. 0.5 0.2 0.1

Figure 47 Impact of maintenance effectiveness on failure rate

For lower failure rates the number of assets in poor condition is higher for the
same level of maintenance effectiveness.

230
 800

600 ya
assets mla
400
aa
200 vaa
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
maintenance effectiveness fraction

Figure 48 Number of assets in different age groupings as a function of maintenance

effectiveness fraction, for MTBF PCA of 2 years.

1000
depreciated asset

800
600
value

400
200
0
0 0.2 0.4 0.6 0.8 1
maintenance effectiveness fraction

PCA fraction failing p.a. 0.5 0.2 0.1

Figure 49 Total depreciated asset value as a function of maintenance effectiveness

Pursuing effective maintenance policies and replacing and refurbishing

equipment on the basis of condition should represent the most efficient use
of resources, provided maintenance effectiveness is also coupled with
efficiency. However these results imply that an enterprise pursuing policies
that extends the life of equipment will suffer loss of revenue linked to the
depreciation of assets. The overall effect on profit (taking into account
revenue and debt levels) will be examined in Section 6.10, while effect of
pursuing an age-based replacement program on the condition of equipment
will be examined in Section 6.7.

231
The average failure rate for the whole asset population is in the range 6 to
1.8% over the maintenance effectiveness fraction range of 0.1 to 0.9 for
MTBF PCA of 2 years. For MTBF PCA of 10 years, maintenance
effectiveness has less effect: the range of values is 3 to 1% for the range of
maintenance effectiveness from 0.1 to 0.9.

Note that failure rate, indicated in this study, is of the right order for
transmission equipment. A CIGRE study into failure in transformers(CIGRE
WG 12.05, 1983), which considered only transformers not more than 20
years old found an overall failure rate of 2% per annum. If older equipment
had been included in the survey this figure could be slightly higher. Taking
only equipment in the range 0 to 30 years (young, mid-life and aged
equipment) the range of failure rates over the same maintenance
effectiveness range is 6 to 0.6% and 3 to 0.49% for the cases with MTBF
PCA of 2 and 10 years respectively.

A CIGRE international enquiry (Janssen et al., 1996) into reliability of single

pressure SF6 high voltage circuit breakers found, for circuit breakers, placed
in service from January 1983 to December 1991, a major failure rate of 0.72
failures per 100 circuit-breaker years and minor failure rate of 4.49 failures
per 100 circuit-breaker years11. An earlier CIGRE study (Mazza and Michaca,
1981) of circuit breakers placed in service between January 1978 and 1982
reported a major failure rate of 0.56 per 100 circuit breaker years and minor
failure rate of 5.35 per 100 circuit-breaker years. These figures are also of
similar order to those suggested by the model outputs.

6.6.2 Impact of use of predictive maintenance on asset population and

value for a condition-based replacement/refurbishment scenario
In this section, the effect of predictive maintenance on failure rate, the age
and condition distributions of assets and total depreciated asset value will be

11
Major failure – complete failure of a circuit breaker causing loss of one or more
fundamental function. Requires removal of the breaker from service within 30 minutes.
Minor failure – failure (inability to perform a required function), other than a major failure.

232
examined. To achieve this, the use of predictive maintenance is varied from
10 to 90%, which in turn, affects the proportion of PCA detected.

For this study a maintenance effectiveness fraction of 0.7 was used. 50% of
detected poor-condition equipment is replaced and 50% of poor condition
equipment is refurbished. For this example the MTBF PCA is assumed to be
2 years.

1000
800
assets

600 GCA
400 PCA
200
0
0 50 100
use of predictive maintenance %

Figure 50 Variation of the condition of equipment within the asset population with use
of predictive maintenance

Figure 50 shows a slight increase in the proportion of good condition assets

as use of predictive maintenance is increased. However, the effect on failure
rate is quite marked (Figure 51), particularly for higher levels of predictive
maintenance usage. This result reinforces the commonly held belief (CIGRE
WG 13-09, 2000, Bergman, 2001) that the use of (well-targeted) predictive
maintenance, coupled with condition-based replacement and/or
refurbishment is effective for reducing failure rates of equipment.

233
 failure rate % assets
3
2.5
2
p.a. 1.5
1
0.5
0
0 20 40 60 80 100
use of predictive maintenance %

Figure 51 Impact of predictive maintenance on failure rate, if condition-based

replacement/refurbishment criteria are used; MTBF PCA 2 years.

The consequence of increasing predictive maintenance on age distribution

(Figure 52) for this case is to increase the number of very aged assets on the
network, which translates into a reduction in the total depreciated value of
assets (Figure 53).

600
500
young assets
400
assets

mid-life assets
300
aged assets
200
very aged assets
100
0
0 50 100
use of predictive maintenance
%

Figure 52 Impact of varying use of predictive maintenance on asset age distribution

This is a result of refurbishing a portion of the poor condition assets.
Refurbishment increases with increasing use of predictive maintenance
(Figure 54). However, because the model assumes that refurbished assets
are not revalued, the depreciated value of assets that are refurbished

234
remains unchanged while the condition is restored: assets refurbished to
good condition may remain in service for some time beyond when their
depreciated value declines to zero.
depreciated value of

600
500
400
assets

300
200
100
0
0 20 40 60 80 100
use of predictive maintenance %

Figure 53 Depreciated value of assets decreases as use of predictive maintenance

increases
assets being refurbished

14
12
10
8
6
4
2
0
0 20 40 60 80 100
use of predictive maintenance %

Figure 54 Refurbishment increases with increasing use of predictive maintenance

This conclusion is further verified with a comparison illustrated, in Figure 55,

of total depreciated value of assets for two cases each with 50% of detected

235
PCA replaced, but one with 50% of detected PCA refurbished and the other
with 0% refurbished.

asset value (units) 800

total depreciated

600
400
200
0
0 20 40 60 80 100
use of predictive maintenance %

50% 0%

Figure 55 Depreciated asset value for the case of refurbishment of 50% of detected
PCA and no refurbishment of PCA. The replacement of detected PCA is held constant
at 50%
Refurbishment can be either treated as a maintenance expense or
capitalised. If it is expensed costs can be recovered (provided they are
allowed by the regulator). On the other hand, if the refurbishment is
capitalised, a profit may be earned on the investment. If the book value is
modified then some return of investment could also be achieved. In practice
there may be some uncertainty over the additional life that refurbishment of
an asset can achieve. Thus there may be a risk that return on and of
investment may not be achieved if refurbishment is capitalised.

In Australia some utilities have shown a preference for treating

refurbishment as a maintenance expense rather than as capital expenditure
((ACCC, 2002a p64)) 12. The reasoning behind this is that rates of return on
equipment are not high and because optimised depreciated replacement
value is used for revenue determination there is a significant risk that a

12
In the case of Electranet the ACCC quarantined the refurbishments from optimisation for
10 years and allowed depreciation of the refurbishment costs over the same period,
effectively addressing the risk posed by optimisation.

236
company may not be able to obtain full return on investment or return of
investment. For some types of equipment refurbishment is perhaps the most
appropriate type of maintenance – transmission lines, for example. For the
case where refurbishment is necessary, this study highlights the importance
of revaluing refurbished equipment to appropriate levels.

By contrast with the Australian situation, in New Zealand the revenue is

based on optimised replacement value with no depreciation factor.
Therefore, in the New Zealand case there is no reduction in revenue
associated with expensing the refurbishment of assets.

As a final study using this condition-based replacement model, the effect of

replacing 50% of detected poor condition assets and 100% of poor condition
assets is compared. This study uses MTBF PCA of 2 years. As expected,
the failure rate is lower (Figure 56) for the 100% case. The depreciated
value is slightly higher (Figure 57) for the 100% case, as more poor condition
assets are being replaced. For the 50% case, the full benefit of the predictive
maintenance is not obtained. This effect is most noticeable at higher levels
of predictive maintenance.

Note that if one compares the failure rate of the 100% replacement of
detected PCA case and the 50% refurbishment, 50% replacement case,
(Figure 51) the model predicts slightly lower failure rates for the latter. This is
because the infant mortality failure rate of refurbished equipment has not
been modelled.

A scenario where less than all detected PCA are replaced or refurbished
might arise where the utility has limited financial resources. This situation is
quite common. It is likely to occur when the level of augmentation of the
system is high, and debt levels are also high, as has been evident in a
number of developing countries. It may also occur when circumstances lead
to a loss of revenue (such as during the period of transition from communist
to capitalist economy in countries of the former Soviet Block). Even in
developed countries utilities will not always have the capital to undertake all

237
projects simultaneously. The need to smooth capital spending over time is a
factor often taken into account in replacement plans (CIGRE WG 37.27,
2000).

3.5
failure rate % p.a.

3
2.5
2
1.5
1
0.5
0
0 20 40 60 80 100
use of predictive maintenance %

50% 100%

Figure 56 Comparison of failure rate for 50% replacement of detected PCA and 100%
replacement of detected PCA. Refurbishment of PCA is not undertaken in either case.
MTBF PCA is 2 years for both studies.
asset value (units)
total depreciated

600
580
560
540
520
500
0 20 40 60 80 100
use of predictive maintenance %

50% 100%

Figure 57 Total depreciated value compared for cases with 50% and 100%
replacement of detected PCA. Refurbishment of PCA is not undertaken in either case.
MTBF PCA is 2 years for both cases.

238
6.7 Age-based replacement and the impact on asset value and failure
rate
In this model it is assumed that fully depreciated assets are replaced. This
translates to replacing equipment in the very-aged assets group. When age-
based replacement is undertaken, instead of condition-based replacement,
some equipment in poor condition may not be replaced and some equipment
in good condition may be replaced. If a strong relationship between age and
condition exists, a reasonable approach is to assume that the very-aged
assets to be replaced are in poor condition. If more very-aged assets are to
be replaced than the number of items in poor condition then the remainder
are replaced from those in good condition. In practice there is likely to be
more variability in the condition of aged equipment than this simple model
suggests, because of the effect of such things as loading, ambient
temperature and duty.13

The construction of this model is very similar to the previous version, except
that now there is a link from very aged assets being replaced to PCA being
replaced and GCA being replaced (as illustrated in Figure 58). The link from
use of predictive maintenance to PCA being replaced and the links from PCA
being replaced to very aged assets being replaced, aged assets being
replaced and mid-life assets being replaced are all removed. There is no
replacement of assets aged less than forty years because only fully
depreciated assets are replaced. As in previous studies, refurbishment is
undertaken on the basis of condition.

13
For instance the duty for circuit breakers switching reactive loads is more onerous than for
those switching unity power factor loads. Breakers switching reactive devices are also likely
to operate more frequently since they are often switched on a daily cycle based on line
loading. The breakers with reactive loads are therefore likely to suffer wear-out earlier than
those with unity power factor loads.

239
 replacement of very aged assets

young assets mid life assets aged assets very aged assets
(0 -9 yrs) (10-29 yrs) (30-39yrs) (40+ yrs)

early randomly randomly MLA randomly AA randomly VAA

failing failing failing failing failing failing failing failing

failed assets all replaced without delay

refurbishment of poor condition assets

replacement of poor condition assets

replacement of GCA

good condition poor condition

deterioration
assets (GCA) assets (PCA)

early randomly
failing failing randomly PCA
failing failing

all failed assets replaced (no delay)

Figure 58 Simplified age-based replacement model: fully depreciated assets are

replaced. Replaced items are assumed to come from PCA and any remaining assets
are replacements of GCA. Refurbishment of PCA is allowed.

6.7.1 Impact of varying maintenance effectiveness fraction

The impact of varying maintenance effectiveness is now assessed.
For this study use of predictive maintenance is 50% and there is 20%
refurbishment of detected PCA. In Figure 59, failure rate as a function of
maintenance effectiveness is illustrated for three values of MTBF PCA failing
– 2 years, 5 years and 10 years.

As in the condition-based replacement case, there is a significant impact

from maintenance effectiveness on failure rate and the age distribution of
assets (Figure 59 and Figure 60 respectively). At low maintenance

240
effectiveness, the number of assets in poor condition is high, and thus there
will be a high failure rate of poor condition assets. For high failure rate, and,
therefore, high level of replacement of failed assets, the number of assets in
the very-aged assets group is low. However, as maintenance effectiveness
increases the number of very-aged assets increases, while the number of
poor condition assets is reduced; therefore the proportion of poor condition
assets replaced annually is very high. Hence the number of failures of poor
condition assets becomes quite low. For maintenance effectiveness 0.8 and
above, the model predicts that all PCA, and a small number of good condition
assets, are replaced annually. This accounts for the levelling of the failure
rate graph at maintenance effectiveness fraction of 0.8.

10
failure rate (% p.a.)

0
0 0.2 0.4 0.6 0.8 1
maintenance effectiveness
MTBF of PCA:
5 yrs 10 yrs 2 yrs

Figure 59 Failure rate as a function of maintenance effectiveness fraction for an age-

based replacement model.

This age distribution results in a total depreciated value that decreases with
increasing maintenance effectiveness (Figure 61), but levels off at
maintenance effectiveness of 0.8.

241
 1000
800
assets 600
400
200
0
0 0.2 0.4 0.6 0.8 1
maintenance effectiveness

young assets mid-life assets

aged assets very-aged assets

Figure 60 Distribution of asset age as a function of maintenance effectiveness fraction

for an age-based replacement model (shown for MTBF PCA of 2 years).
total depreciated value

1000
(asset value units)

800
600
400
200
0
0 0.2 0.4 0.6 0.8 1

maintenance effectiveness

Figure 61 Total depreciated asset value as a function of maintenance effectiveness for

an age-based replacement scenario (illustrated for a cased with MTBF PCA of 2
years).

6.7.2 Effect of (condition-based) refurbishment for age-based model

The previous study included some refurbishment of poor condition
equipment. Comparison of cases with and without refurbishment (Figure 62)
shows that, with refurbishment, the failure rate is lower, for low to moderate
levels of maintenance effectiveness. The two plots converge when
maintenance effectiveness is greater than or equal to 0.8; at this point all

242
PCA are being replaced.

10
failure rate % p.a 8
6
4
2
0
0 0.2 0.4 0.6 0.8 1
maintenance effectiveness fraction

no refurbishment with refurbishment

Figure 62 Failure rate as a function of maintenance effectiveness for age-based

replacement case with no refurbishment and refurbishment at 20% of detected PCA,
with predictive maintenance level 50%. MTBF of PCA is 2 years in both studies.
total depreciated value

1000
800
(units)

600
400
200
0
0 0.2 0.4 0.6 0.8 1
maintenance effectiveness fraction

no refurbishment with refurbishment

Figure 63 Total depreciated value as a function of maintenance effectiveness for age-

based replacement with no refurbishment and 20% of detected PCA refurbished.
MTBF of PCA is 2 years for both studies.

Compared with the case with refurbishment, the total depreciated value of
assets, for the case with refurbishment, is slightly higher (Figure 63)

243
because, as in previous models, no revaluation of refurbished assets has
been undertaken.

6.7.3 Comparison of age-base and condition based replacement

models with refurbishment for different values of maintenance
effectiveness
In Figure 64 and Figure 65 the age-based replacement scenario is compared
with the condition-based replacement case which had 50% replacement of
detected PCA (used in Section 6.6.1). Both cases have use of predictive
maintenance, 50% and refurbishment of detected PCA, 20% and MTBF PCA
2 years. Use of higher MTBF PCA value results in a reduction in the
difference between replacement strategies at low maintenance effectiveness
levels (see Figure 69). However, age-based replacement still results in worse
failure rates than condition-based replacement at low maintenance
effectiveness levels.

10
failure rate % p.a.

8
6
4
2
0
0 0.2 0.4 0.6 0.8 1
maintenance effectiveness fraction

condition-based replacement age-based replacement

Figure 64 Comparison of failure rate for condition-based replacement 50% of detected

PCA, and 50% predictive maintenance and age-based replacement of assets (assets >
40 yrs).
The model suggests that strategy of condition-based replacement of assets
is superior for reducing failure rate at lower maintenance effectiveness
fractions. This is because the replacement of assets, in the condition-based
strategy, is proportional to the number of poor condition assets, which is
higher for lower values of maintenance effectiveness, whereas the age-
based replacement strategy is proportional to the number of very-aged

244
assets, which is higher for higher maintenance effectiveness fraction. In
practice, because there are likely to be poor-condition assets with age less
than the replacement cut-off, as previously explained, the age-based
replacement strategy is not likely to be quite as successful at high
maintenance effectiveness fractions as this simple model predicts.
Nevertheless it would appear from the results that either age- or condition-
based replacement strategies can produce satisfactory results (in terms of
system performance) provided adequate effective maintenance is
undertaken.
total depreciated value

1000
800
(units)

600
400
200
0
0 0.2 0.4 0.6 0.8 1
maintenance effectiveness fraction

condition-based age-based

Figure 65 Comparison for total depreciated value of condition-based replacement

(20% of detected PCA, 50% predictive maintenance) and age-based replacement
(>40yrs)
The plots for depreciated value for condition-base and aged-based
replacement also show better performance at lower levels of maintenance
effectiveness for a condition-based replacement strategy, and at higher
levels for an age-based strategy. Once again this is a function of the number
of assets replaced: for high levels of maintenance effectiveness there are few
poor condition assets to be replaced (using the condition-based replacement
strategy) but more assets in the 40 + age bracket to be replaced annually (for
the age-based strategy).

245
6.8 Remedial maintenance of poor condition assets
Although, for the purpose of reliability analysis it is often assumed that repair
is to “as good as new” condition, in practice this often not the case
(Davidson, 1988 p12). Davidson explains that this can occur when repair of
one element in a device leaves other parts still in a “wear out” condition. This
means, in reliability terms, that the condition after repair may be “as bad as
old”.

Thus to include remedial maintenance of PCA in the model it has been

assumed that failed PCA are repaired to the same condition they were in
before they failed. Remedial maintenance can also be applied to failed
assets in good condition (random or infant-mortality type failures). In these
cases the repaired assets will also be in good condition. Repair or
replacement of young assets failing from infant mortality will restore the
equipment to the same age grouping. Because the MTBF for infant mortality
and random failures of GCA are low compared with the MTBF PCA including
repair following such failures will have a second-order effect on outcomes,
and has not been included in the current investigation. Another option for
modelling that has not been incorporated in the current investigation is
refurbishment of failed PCA. These options are possible useful extensions to
the model for future research.

To model the remedial maintenance of failed PCA, a solution used is to add a

flow called PCA being repaired, which is dependent on the failure of PCA and
a converter called fraction failing PCA repaired and feeds back into the PCA
stock. The number of PCA being replaced then is reduced to (1- fraction
failing PCA repaired) * PCA failing. The repaired and replaced assets in the
age distribution part of the model then must be reconciled with the condition-
based part of the model. This requires addition of three more flows: VAA
being repaired, AA being repaired and MLA being repaired. Each of these
feeds back into their respective age groups. Each is calculated by the assets
failing (per unit time) from the associated age group multiplied by the fraction
failing PCA repaired. Likewise the assets being repaired for each age group
is reduced by the factor (1- fraction failing PCA repaired).

246
Figure 66 and Figure 67 show the impact of these changes on the model
outcomes: the condition-based model with maintenance effectiveness of 0.7,
use of predictive maintenance 50%, refurbishment of detected PCA 0.2, and
fraction of detected PCA replaced 0.5 is used as a based case (for which
fraction failing PCA repaired is 0.0). The value of fraction failing PCA
repaired is then varied in the range 0.1 to 0.9.
total depreciated asset

540
520
value (units)

500
480
460
440
0 0.2 0.4 0.6 0.8 1
fraction failed PCA repaired

Figure 66 Total depreciated asset value with increasing use of remedial maintenance
of PCA

4
failure rate % p.a.

2
1

0
0 0.2 0.4 0.6 0.8 1
fraction failed PCA repaired

Figure 67 Failure rate as a function of fraction failed PCA repaired

The overall effect, as should be expected, is to increase the average age of
assets, thus decreasing the revenue derived from depreciated value of
assets. However, failure rate also increases with remedial maintenance of
poor condition assets, because of the assumption that repaired poor
condition assets remain overall in poor condition. This result contrasts with

247
(condition-based) refurbishment, where assets are restored to good
condition, and the effect of increasing refurbishment is to reduce failure rate.

The conclusion that can be drawn from this is that a policy of repairing poor
condition assets is detrimental to system performance. Repairing good
condition assets would not have the same performance penalty, because
these are returned to good condition. The enterprise could then have the
benefit of retaining the return on and of asset value for the remaining life of
the asset. This is particularly true for assets failing from infant mortality – if
these can be repaired most of their useful life is still ahead of them.

In practice, avoidance of repairing poor condition assets may not be easily

achieved, for it depends on the ability to ascertain the condition of the assets.
Therefore, unless condition-based assessment techniques (and hence
predictive maintenance) are used, an enterprise will have limited information
on which to decide whether or not an asset should be repaired. In addition to
this, there may be financial circumstances that do not permit replacement of
an asset that has failed, even though it is known to be in poor condition.
Under such circumstances it may be necessary to repair the asset if it is
required for network operation.

6.9 Replacement on failure only

In some utilities, particularly in North America (for example (Lee, 2001)),
there is no replacement policy for ageing assets, and low usage of predictive
maintenance, which in turn argues for little opportunity to pursue a condition-
based replacement policy. Even when there is no use of predictive
maintenance, the model allows for some detection of poor condition in
assets, as part of normal routine maintenance. Thus there could be a small
amount of condition-based refurbishment and/or replacement of poor
condition equipment undertaken. To prevent this from occurring, the fraction
detected PCA refurbished and fraction detected PCA replaced can be set to
zero.

248
The effect of employing a replacement on failure-only policy is now
investigated, using the same parameters as were used for the previous study
(except that the case is studied both with remedial maintenance and without
remedial maintenance.)

Figure 68 illustrates that this policy results in high failure rates for low
maintenance effectiveness. It also reinforces the conclusion, in the previous
section, that repairing failed PCA has a detrimental effect on failure rates
(The 50% remedial maintenance case generates higher failure rates than the
0% case over the full range of maintenance effectiveness).

Comparison of failure rates from a replace-on-failure strategy with those from

the age and condition-based replacement strategies shows (Figure 69) that
this replacement on failure model produces higher failure rates than either of
the other policies, as should be expected. In this figure the MTBF PCA is 5
years.

It is interesting to consider whether this policy is more resource efficient or

cost efficient than other policies. If no remedial maintenance is undertaken
(all failed equipment is replaced) and very low levels of predictive
maintenance and refurbishment are used, then maintenance costs are a
function of routine preventive maintenance costs. If an enterprise is highly
efficient in performing preventive maintenance then this could result in very
low maintenance costs overall. Provided the preventive maintenance is also
highly effective, failure rates could be kept down to a level that is
manageable, especially in a network with high redundancy (see discussion in
Section 6.14 below). Asset replacement rates of 2.3 to 1.8% per annum
(corresponding to failure rates for maintenance effectiveness range 0.7 to
0.9, for the “no remedial maintenance” case in Figure 69 are high, but not
excessive for a network with assets of forty years expected life (The CIGRE
study found 70% of respondents had replaced less than 5% of their network
in the four years prior to the survey). If a small amount extra condition-based
refurbishment is added, then the failure rates, and thus replacement levels,
could be reduced further. A replacement-on-failure-only strategy could

249
therefore explain the very low direct maintenance costs reported for the North
American respondents of the CIGRE study.

failure rate % p.a. 20

15
10
5
0
0 0.2 0.4 0.6 0.8 1
maintenance effectiveness fraction

no remedial maintenance 50% remedial maintenance

Figure 68 Effect of replacement on failure only policy, without remedial maintenance

and with 50% of failed PCA repaired to condition before failure (MTBF PCA 2 years).

6
5
failure rate % p.a.

4
3
2
1
0
0 0.2 0.4 0.6 0.8 1
maintenance effectiveness fraction
condition-based age-based replace on failure only

Figure 69 Comparison of failure rate for three replacement strategies. Other

parameters: MTBF PCA 5 years, use of predictive maintenance 50%, refurbishment of
detected PCA 20%.

The total depreciated value as a function of maintenance effectiveness for

this strategy is shown in Figure 70 below. Comparison with Figure 49 shows
that the impact of this strategy on depreciated value is very similar to that of

250
the condition-based replacement strategy first introduced in Section 6.6. It
can be concluded that, compared with the condition-based replacement
strategy, the replace-on-failure-only strategy is lower cost, similar in effect on
depreciated asset value (at least in cases using straight line depreciation)
but comes equipment failure rate penalty.

1000
Total depreciated value

800
600
(units)

400
200

0
0 0.2 0.4 0.6 0.8 1
maintenance effectiveness fraction

Figure 70 Total depreciated value as a function of maintenance effectiveness for a

replacement-on-failure-only strategy, without remedial maintenance.

6.10 The impact of maintenance and replacement policy on debt and

profit
Maintenance and replacement policies have an impact on revenue in a
number of ways. If depreciated value is used in revenue calculation (as in
the Australian case), asset management policies affect the value of assets
used for revenue calculation purposes. If improperly maintained assets fail
before they are fully depreciated the depreciation over their remaining book
life will be lost. Assets that are replaced also need to be funded either from
revenue or debt. The interest payments associated with debt also come out
of revenue, and reduce profit, and potential for dividends. Utilities usually
aim to maintain their exposure to debt at some target value (or within a
range) in accordance with their financial management policy. The assumption
of a constant (benchmark) gearing ratio is made in the determination of
revenue for Australian transmission enterprises (ACCC, 2002a p107)14.

14
It is noted that the ACCC used a benchmark value of 60% for gearing ratio calculation in
the Electranet case, although the actual gearing ratio for this company was 80% at the time.

251
The model has been developed based on the Australian method of revenue
calculation. The revenue calculation for Australian Transmission Network
Service Providers (TNSPs) has three components (Equation 9):
Revenue = WACC * DORC + Depreciation + O&M (9)

where:
WACC = weighted average cost of capital;
DORC = depreciated optimised replacement costs; and
O&M = operations and maintenance.
Straight line depreciation is used in this calculation. The term optimised
indicates that the return on capital is based on prudent expenditure: the
regulator has the power to disallow return on capital for sub-optimum capital
projects, which is intended to provide a disincentive for over-expenditure.
The O&M component is based on anticipated costs, adjusted for inflation and
an efficiency gain, the value of which is set by the regulator. This is
described as a CPI-X scheme (see also sections 2.2.1.2, and 0 for more
information).

In this implementation a number of simplifying assumptions have been made:

• All values are in today’s dollars and replacement costs are implied to
remain constant in real terms.
• Dividends have been modelled but the value of dividends has been
set to zero, so retained earnings has been used as a measure of
financial well-being.
• Refurbishment (as in previous sections) is not capitalised – it is treated
as O&M.
• It is assumed that there is no net effect from O&M on retained
earnings – all the O&M allowance is spent.
• It is also assumed that the O&M allowance is sufficient, to the extent
that it does not impact on maintenance effectiveness.
The assumption of negligible net effect from O&M on retained earnings is a
reasonable approximation of what occurs in practice: the elements of
revenue based on asset value contribute up to 80% (Jones et al., 2000) of

252
revenue, and the CPI – X scheme is considered by some15 to provide a
disincentive to reducing O&M. The assumption that there is no impact on
maintenance effectiveness is more difficult to defend, in the longer term, if
the regulator continues to apply a minus X adjustment.

As depicted in this diagram, a constant gearing ratio is maintained in the

model by an adjustment (labelled repaying debt) from retained earnings. As
described previously, while the model includes an output for dividends, it was
decided to set this flow to zero, for the purpose of modelling, and use
retained earnings as the output for comparison of strategies.

The modifications to the model to incorporate revenue, retained earnings and

debt level are shown in Figure 71. Two new stocks are added to the diagram
– debt level and retained earnings. The flow into retained earnings is
revenue and since O&M is being ignored, this is calculated by the two
components return on capital (WACC * DORC) and depreciation. Since the
expected life of assets is forty years, assets in the VAA group are fully
depreciated and do not contribute to DORC. The average age of assets in
each age group is assumed to be the average of the range for the age group.
DORC is therefore calculated by (Equation 10):
DORC = value per asset * optimisation factor/ expected life *
[(no. assets in YA)* (expected life – avg age of YA) +
(no. assets in MLA) * (expected life – avg age of MLA) +
(no. assets in AA ) * (expected life – avg age of AA)] (10)
The annual depreciation is the change in value from one year to the next.
Since linear depreciation is used this is equivalent to (Equation 11):
Annual depreciation =
Value per asset/ expected life * optimisation factor *
(no. assets in YA + no. assets in MLA + no. assets in AA) (11)

15
This view was expressed in a CIGRE Asset Management seminar (Murray, 2000). The
disincentive to cost reductio is partly offset by a scheme which allows profits from reductions
in O&M costs to be retained in part over the next regulatory review period.

253
 internal funding ratio target
gearing
ratio
accumulating debt
WACC debt reducing debt
return
on capital

avg age MLA earning revenue repaying debt

retained
earnings
avg age YA paying interest

annual
avg age AA depreciation
DORC interest
rate

capital funding dividends

optimization
of assets
factor

expected life

value MLA replaced AA replaced VAA replaced

per
asset

young assets mid life assets aged assets very aged assets
(0 -9 yrs) (10-29 yrs) (30-39yrs) (40+ yrs)
purchasing
assets

early randomly randomly MLA randomly AA randomly VAA

failing failing failing failing failing failing failing failing

Figure 71 Illustration of model dependencies when retained revenue and debt

calculations are added to the model.

The outflows from retained earnings are dividends (set to zero), capital
funding of assets, interest and repaying debt. In practice a portion of capital
spending is funded from earnings and the remainder from debt. To model
this relationship an internal funding ratio and a target gearing ratio have been
defined. The amount drawn from retained earnings for capital funding of
asset is initially the value of assets replaced multiplied y the internal funding
ratio. The amount funded from debt is the total value of assets replaced less
the amount funded from capital. If funds are available, after interest
payments have been made, they are transferred from retained earnings to

254
maintain gearing at the target level. This is the flow repaying debt. Gearing
ratio has the usual meaning here: ratio debt to (debt plus equity), where debt
plus equity equals retained earnings plus DORC. A target gearing ratio of
60% has been used, based on the benchmark level used by the Australian
regulator for a recent revenue decision (ACCC, 2002a). For consistency the
values of WACC (7.17%) and interest rate (6.39%) have been drawn from
the same source.

An optimisation factor of 1 has been assumed for the initial study (i.e. no
optimisation). The effect of introducing optimisation is considered later in this
section.

Predicted impact of maintenance effectiveness on retained profit is shown in

Figure 72, for a condition-based replacement policy. Parameters for the
condition-based replacement strategy were: 50% use of predictive
maintenance and 50% replacement of detected PCA. The model also
assumed a 20% of detected PCA refurbished, and MTBF PCA 2 years. The
impact of extending MTBF PCA will be examined in the Section 6.12. The
study length was 200 years, allowing the values to settle to steady state.

The model suggests that because of the competing effects of interest

payments (a function of replacement level) and revenue (a function of
depreciated value) maximum revenue is obtainable for a level of
maintenance effectiveness less than perfect (0.73). For higher values of
maintenance effectiveness there is actually a slight decline in profit. The
model also implies that if maintenance effectiveness drops below a certain
level (approximately 0.6 for the parameters chosen for this study) the
company will make a loss each year and debt level will become
unsustainable. The point at which the company starts to make a loss is
sensitive to the initial value of retained earnings. In all the case studies a
fairly large initial value of retained earnings was chosen (121 asset value
units), because no attempt has been made to set the studies to start in
steady state. The studies all tend to steady state because they are based on
two main chain structures with constant population. Hence the values (for

255
maintenance effectiveness values greater than the break-even point), give
results that are independent of initial conditions.

1000

800
(1 = asset value)
TDV
monetary units

600
retained
earnings
400
debt
200

0
0.5 0.6 0.7 0.8 0.9
maintenance effectiveness

Figure 72 Impact on profit, debt and depreciated value of variation of maintenance

effectiveness

Clearly, in practice, the situation of loss-making would not be allowed to

continue indefinitely. What would the likely outcome be? On posting a loss,
one could expect the company to discontinue replacing assets that have not
failed (capital rationing), and possibly repairing rather than replacing the
assets that have failed. This, in turn, would result in an increased proportion
of poor condition assets and an increased failure rate. The organisation
would also probably put an argument to the regulator for greater revenue as
soon as possible, in order to increase their maintenance budget. (Recall that
the CIGRE study found that deteriorating system condition was a major
influence on maintenance budget). Provided that the additional maintenance
budget was channelled towards improving maintenance effectiveness the
company could return to financial viability.

The study was repeated allowing no refurbishment of poor condition assets.

256
 (asset value units)
300

retained earnings
250
200
150
100
50
0
0.5 0.6 0.7 0.8 0.9 1
maintenance effectiveness

with refurbishment no refurbishment

Figure 73 Comparison, for the condition-based replacement policy case, of retained

earnings with and without refurbishment of PCA

As Figure 73 shows, without refurbishment the point of maximum retained

revenue occurs for a slightly higher value of maintenance effectiveness. The
maintenance effectiveness level at which the company just breaks even also
increases.

The effect of optimisation on profitability is shown in Figure 74, for the

previous case with no refurbishment and the same case, but with an
optimisation factor of 0.9 applied.

The model does not take into account possible reductions in replacement
cost from assets that are optimised down in value or completely optimised
out. Figure 74 suggest that a moderate amount of optimisation will have a
significant effect on the profitability of the company. The break-even point for
profitability is increased significantly when optimisation factor of 0.9 is applied
to return on capital and return of capital. Note however, that the point of
maximum retained earning is the same for both cases, although the level of
maximum retained earnings is much lower for the case with optimisation.

257
 300

(asset value units)

retained earnings
250
200
150
100
50
0
0.5 0.6 0.7 0.8 0.9 1
maintenance effectiveness

no optimisation optimisation to 90% of value

Figure 74 Effect on retained earnings of optimising asset value compared with the
case with no optimisation

6.11 Comparison of profit outcomes for various replacement policies

The foregoing analysis can be extended to the age-based replacement and
replace on failure only scenarios. The results, over a range of maintenance
effectiveness values, are shown in Figure 75 below. In each case there is no
refurbishment of assets, the optimisation factor is 0.9 and the retained
earnings are accumulated over a 200 year run. For the condition-based
replacement case, as in previous studies, a 50% use of predictive
maintenance is used with 50% replacement of detected PCA. The analysis
suggests that all three policies give similar results, except that the age-based
replacement case is not sensitive to maintenance effectiveness at levels of
0.8 and above. This occurs because, for these values of maintenance
effectiveness, all poor condition assets are replaced, and some good
condition assets are also replaced. Because the failure rate of good
condition assets is so much lower than it is for poor condition assets, the
number of assets replaced annually for aged-based replacement (including
those assets replaced after failure) is virtually constant for high levels of
maintenance effectiveness.

258
 180

retained earnings (asset

160
140
120
units)
100
80
60
40
20
0
0.6 0.7 0.8 0.9 1
maintenance effectiveness

age-based condition-based replace on failure only

Figure 75 Comparison of retained earnings for three different replacement scenarios.

An optimisation factor of 0.9 is used, and there is no refurbishment in each case.

6.12 Sensitivity of profit to failure rate assumptions

A very interesting result occurs if MTBF PCA is varied. Figure 76 and Figure
77 below show retained earnings as a function of maintenance effectiveness
for three maintenance strategies. In the first case the MTBF PCA is 5 years
and in the second, 10 years. These graphs show that as the value of MTBF
PCA is increased the differences between results from different strategies
become greater. For the higher value of MTBF PCA the range over which a
positive level of retained earnings is achieved is much wider for the replace-
on-failure strategy than for the condition-base replacement strategy. The
replace-on-failure strategy can also produce positive retained earnings for
much lower levels of maintenance effectiveness. In each case the age-
based replacement strategy yields the best retained earnings for the higher
levels of maintenance effectiveness (except over a very narrow range of
maintenance effectiveness levels, where the condition-based replacement
strategy produces a better outcome). Retained earnings appear to be less
sensitive to the level of MTBF PCA for condition-based replacement than it is
for the other two replacement strategies.

259
 300

250
(asset value units)
retained earnings

200

150

100

50

0
0 0.2 0.4 0.6 0.8 1
maintenance effectiveness

condition-based age-based replace on failure

Figure 76 Retained earnings is presented as a function of maintenance effectiveness

for three replacement strategies and MTBF PCA of 5 years. (Other parameters: use of
predictive maintenance 50%, replacement of detected PCA 50% in condition-based
replacement case; refurbishment of detected PCA 20% in all cases; optimisation
factor 1.0)

300
250
(asset value units)
Retained earnings

200
150
100
50
0
0 0.2 0.4 0.6 0.8 1
Maintenance effectiveness fraction

condition-based age-based replace on failure

Figure 77 Retained earnings as a function of maintenance effectiveness. (Parameters:

MBTF PCA of 10 years, all other parameters same as previous case.)

260
6.13 Comments on Maintenance efficiency and the impact of
replacement and maintenance strategies on maintenance
spending
It is evident from the previous studies that replacement and maintenance
strategies have a bearing on maintenance spending. Other factors also
contribute to this, notably maintenance efficiency and labour cost.

Maintenance efficiency can be defined as the resources taken to perform

maintenance to achieve a level of performance outcome. It includes such
things as the person-hours taken to maintain a piece of equipment, the
person-hours taken to diagnose problems, the spares required to support
maintenance activities. The number of hours taken to maintain equipment
also depends on the frequency of maintenance, and the activities involved in
maintenance. Undertaking maintenance more often than necessary will not
increase the performance of assets, and can actually reduce performance if
maintenance is not perfectly performed. Undertaking unnecessary
maintenance, likewise, does not increase performance, and has the potential
to reduce it. The use of such techniques as RCM and maintenance
optimisation (through analysis of failure modes and rates) can help to
increase maintenance efficiency. The use of predictive maintenance and the
effective use of condition monitoring could increase maintenance efficiency
by:
• reducing the hours needed to perform maintenance,
• identifying equipment in need of refurbishment, and
• eliminating (or at least reducing) the counter-productive repair of poor
condition assets.
The increase in maintenance efficiency with use of predictive maintenance is
offset by the cost of undertaking predictive maintenance. It is therefore
important that predictive maintenance is well-targeted toward detecting the
major causes of failure.

A high level of maintenance effectiveness does not necessarily imply a high

level of maintenance efficiency. The CIGRE study did not give sufficient
detail on maintenance practices to estimate maintenance efficiency, but the

261
discrepancy between the maintenance spending predicted by the fuzzy
model and actual maintenance spending reported for the Japanese cases
suggests that maintenance efficiency could be improved in that country,
despite high maintenance effectiveness. The lack of financial drivers to
reduce maintenance costs in Japan, at least until recently, is likely to be a
major cause of persistent inefficiency.

6.14 Impact of failure rate on system performance

In this chapter until now the performance of the network has been considered
in terms of equipment failure rate. However, as discussed in Chapter 3,
system design also plays a large part in network performance.

6.14.1 Effective redundancy

A network comprises a set of parallel and series connected components.
Whether failure of element of the network affects the continuity of connection,
and to what extent, depends on where in the network the failure occurs: parts
of the network may be more tightly meshed than others. It can also depend
on when the failure occurs: the loading of the transmission network is usually
time-dependent. Ultimately these factors condense to an effective overall
redundancy level in the network.

It is proposed that a measure of effective redundancy of a network is the ratio

of circuit outages per annum to the number of loss of supply incidents per
annum, caused by the circuit outages. The fascinating figures of Table 8
suggest that, for a transmission network of very moderate connectivity
designed to an N-1 redundancy level, there is an effective redundancy factor
somewhere in the range of 8 to 10 between number of outages and outages
causing supply loss. This is at first glance surprising, but current can flow
through parallel paths in interconnected networks at transmission, sub-
transmission and even distribution level. If for this particular case, failures
are assumed to cause outages (an approximation), a 3% failure rate per
annum would cause 30 outages per 1000 assets per year and of the order of
3 failures of supply per 1000 assets per year.

262
For this network there are only two points of interconnection with another
transmission network, plus connections to sub-transmission / distribution
networks: for highly intermeshed and interconnected, moderately loaded
networks in Western Europe and North America the effective redundancy
factor is likely to be much higher. This means that significantly higher
equipment failure rates can be sustained for the same supply reliability, or
alternatively that with effective maintenance practices very low levels of
unreliability should be readily achievable. It also supports the commonly held
view that, for networks of low connectivity (eg South America),
interconnection to adjacent networks can be of significant benefit, in terms of
improving reliability of supply to customers.

In terms of replacement strategy, the following observations can be made:

• Profit-driven companies with systems having high levels of effective
redundancy (eg North America) could find replacement-on-failure-only
strategies attractive, because of the low-implied level of maintenance
cost, and the low impact of failure rate on reliability of supply.
Because North American companies predominantly operate under
rate-based regulation, based on historical or replacement costs of
assets and not depreciated value, there will be little incentive to
replace assets that have not failed: assets in service continue to
attract an income regardless of age. These observations are
supported by evidence from interviews of representatives from three
US utilities(Lee, 2001, Sherard, 2001, Teixiera, 2001): companies
tend to keep equipment in service until it fails; if more capacity is
required new assets are connected in parallel with existing ones
(which increases redundancy), and as little as possible is spent on
maintenance. It follows from this argument that in North America
there will be less use of on-line condition-based maintenance, as with
proportionately fewer replacements there is less opportunity to use on-
line CBM, and less inclination to use off-line CBM. From the CIGRE
study the North American respondents also reported results consistent
with this view: the proportion of aged assets is typically in the North
American cases, a little higher than other mature electricity systems,

263
 the level of replacement of equipment is low (see Figure 4) and use of
both offline and online CBM is low compared with other regions (See
Figure 1). The Fuzzy Model also adds some weight to these
arguments – use of a replace-only-on-failure criterion will have the
effect of increasing two inputs proportion of aged equipment and
redundancy. The effect on reliability of supply of different proportions
of aged equipment for three levels of redundancy ranging moderate
(50) to high (70) is illustrated in Figure 78, using the North American
base case. The graph confirms that as redundancy increases, the
impact of aging equipment on reliability of supply is reduced.
• Enterprises that have lower levels of financial drivers on maintenance
spending, and which operate in networks of high effective redundancy
(eg Western Europe, Japan), could achieve very high levels of supply
reliability (and circuit reliability) by employing a moderate amount of
predictive maintenance and condition-based replacement. Availability
of circuits may not be as high, compared with other enterprises,
because higher effective redundancy implies lower risk associated
with taking lines out of service.
• In networks with more moderate levels of effective redundancy (eg
Australasia, South America and parts of Asia, failure rates of
equipment will have a higher impact on reliability of supply. Therefore
in order to achieve high reliability, high maintenance effectiveness is
required. A relatively high level of use of predictive maintenance (eg
in Australia, refer to Table 11) will be effective in maintaining failure
rates at low levels (See Figure 21).
• Networks with very low levels of effective redundancy (eg India), are
likely suffer from low reliability of supply. In such networks the system
risk associated with maintenance is high and hence utilities will be
under pressure to achieve a high level of availability of circuits, as is
reported for India (Kumar et al., 2002).

264
 Reliability of supply
high
moderate
low

0 10 20 30 40
proportion of aged equipment %
Redundancy:
moderate 50 mod-high 60 high 70

Figure 78 For the North American base case this graph shows how increasing
redundancy reduces the impact of aging equipment on the reliability of supply.
The particular data that have been used here for calculating effective
redundancy are seldom published, but should be readily obtainable from the
databases of transmission enterprises. It would be very informative to
compare these parameters to obtain a basis for comparing the likely impact
of different operation and maintenance schemes on performance of the
system. This is an area recommended for further research.

6.15 Impact of system augmentation rate on network composition and

performance
If augmentation of the network is included in the model, the system will no
longer be in steady state. Assets are always added in good condition and to
the young assets group. Therefore, for enterprises with revenue based on
asset value or return on investment, the augmentation of the network is an
opportunity for increased profits as well as improved overall performance with
the introduction of the latest equipment.

If a very simple model of augmentation is implemented which allows a

constant rate of asset growth as a percentage of assets, a 3% growth per
annum results in a very large growth in assets over time, as Figure 79 shows.

265
 400000
350000
300000
250000
assets

200000
150000
100000
50000
0
0 20 40 60 80 100 120 140 160 180
years

GCA PCA

Figure 79 A 3% rate of augmentation of assets per annum – clearly impractical.

Clearly this is not a practical situation. The first point is that it is simply not
practical to expect the number of assets to grow indefinitely, for reasons of
space required for instance. In practice as networks grow the capacity per
asset (and therefore value per asset) grows. The rate of increase of assets
probably decreases gradually with maturity of a network. In the long term it is
unrealistic to expect that the Earth can support ever increasing energy
demand, in which case the rate of increase of assets will probably decline to
zero. If multiple smaller assets are replaced with fewer larger capacity
assets as they need to be replaced then the total number of assets on the
network could decline over time, even if the replacement value of assets
increases.
There are signs that this is already occurring in some places: For example,
in Western Europe there is a strong trend today toward the replacement of
assets with higher capacity and/or more compact designs because of
extreme difficulties in obtaining land for augmentation (Aesbasch et al., 2000,
Lamsoul et al., 2000, Mittard et al., 2002). As networks mature and the

266
population density increases in other parts of the world, such pressures are
also likely to extend to other regions.
replacement of very aged assets

replacement of aged assets

young assets mid life assets aged assets very aged assets
(0 -9 yrs) (10-29 yrs) (30-39yrs) (40+ yrs)

early randomly randomly MLA randomly AA randomly VAA

failing failing failing failing failing failing failing failing

failed assets all replaced without delay

value per asset

refurbishment of poor condition assets

network
replacement of poor condition assets
maturity

good condition deterioration poor condition

assets (GCA) assets (PCA)
augmenting network

maturity
aug. fraction maintenance
effectiveness
early randomly
failing failing randomly PCA
failing failing
load growth
rate %

augmentation
rate all failed assets replaced (no delay)

Figure 80 Simplified model incorporating system augmentation

Figure 80 shows the simplified diagram of a model representing a network

being augmented. The model is based on the same condition and age
distribution model used previously. In addition it has a flow named
augmenting network that is an input to good condition assets. An equal
number of assets is added to the young assets.
The value of augmenting network is a function of load growth rate, and
network maturity. The relationship between network maturity and the
augmentation rate (augmentation fraction) is defined using a graphical
function so that as network maturity increases from 0 to 100% the value of
the multiplier, maturity augmentation factor, decreases from 1.0 to 0.0 as a

267
straight line. This means that for 100% network maturity the rate of increase
of assets from augmentation drops to zero. Network maturity is assumed to
increase at a rate of 2% per annum to a maximum of 100%; this relationship
is implemented through the connector between network maturity and the flow
increasing maturity. This is also modelled as a graphical function, which is
simply constant from 0 to near to 100% maturity and then drops to zero.
Load growth is also assumed to be a function of network maturity.

The load growth has potential to grow most quickly for networks of low
maturity as more consumers are added to the network and the growth of
supply of electricity fosters industrial growth. As the network becomes
mature, the growth rate is reduced to the rate at which the consumption of
users is increased (Khatib, 1998). Khatib reports that the global growth rate
of electricity consumption was about 4% per annum in the past forty years
but is expected to drop to 2.9% per annum in the period 1997 to 2020.
However, during this period the average growth rate in developing countries
is expected to be between 4-6% per annum. The global rate of electricity
demand growth is expected to drop to about 1.7% per annum in the period
2020 to 2030, as the networks in currently developing countries become
more mature. These figures support the argument that load growth is
reduced with increasing network maturity. Some currently developing
countries have experienced significantly higher growth rates than 6% in
recent years: in the period 1980-1997 the average annual growth in
production of electricity (figures from (World Bank, 2000a)) was 13.5% in
Indonesia, 12.1% in the Republic of Korea, 12.5% in Thailand and an
unusually high 33.1% in Paraguay. By contrast, in European Union countries
the growth in production was 2.4% per annum, over the same period.

The graphical relationship assumed is shown in Figure 81.

268
 25

20
load growth %
15

10

0
0 20 40 60 80 100
network maturity %

Figure 81 Modelled relationship between load growth and network maturity

The tendency towards use of higher capacity and higher value assets as the
network matures has also been modelled. The graphical function (illustrated
in Figure 82), assumes that the value of assets will increase by a factor of
three over the full range of network maturity.

3.5
3
value per asset

2.5
2
1.5
1
0.5
0
0 20 40 60 80 100
network maturity

Figure 82 Modelled relationship between network maturity and value per asset
When this model is run a more reasonable asset growth plot is obtained as
shown in Figure 83. The corresponding time-development of network
maturity and load growth are shown in Figure 84.

269
 10,000 8,000

total asset value

8,000 6,000
assets

(units)
6,000
4,000
4,000
2,000 2,000
0 0
0 50 100 150 200
years

total replacement value total assets

Figure 83 The asset growth rate and total replacement value of assets as a function of
time, and with increasing network maturity.

100 20
network maturity (%)

load growth % p.a.

80 15
60
10
40
20 5

0 0
0 50 100 150 200
years

network maturity load growth

Figure 84 Network maturity and load growth as a function of time

It is also possible to examine the projected age distribution of assets as a

function time (Figure 85), and other parameters that are modelled. However
it is clear that with non-constant asset population caused by varying
augmentation rates, the underlying effects of maintenance and replacement
policy are obscured. Hence an enterprise seeking to understand the effects
of policies and practices on their long-term system performance and

270
profitability could potentially benefit from modelling their policies using these
or similar techniques.

4000
3500 MLA
3000
2500
assets

2000 YA
1500
1000 AA
500 VAA
0
0 50 100 150 200
time (years)

Figure 85 Age distribution of assets as a function of time.

6.16 Summary of findings from system dynamics modelling

A simple model combining condition and age of assets has been shown to be
effective for modelling a range of different maintenance and replacement
scenarios. A number of findings have resulted from studies undertaken using
this model.
• Establishing a high level of preventive maintenance effectiveness is
the fundamental way of producing high system performance, through
control of condition degradation.
• Use of predictive maintenance coupled with condition-based is
effective for reducing failure rate of poor condition assets by means of
targeted refurbishment or replacement.
• Either condition-based or age-based replacement policies can result in
high system performance when coupled with effective preventive
maintenance. However, failure rates will be slightly higher for the age-
based replacement strategy for the same level of maintenance
effectiveness.

271
• A policy of refurbishment of poor condition assets is effective in
reducing failure rate but will result in reduced depreciated value of
assets unless refurbishment is capitalised.
• At lower maintenance effectiveness levels an age-based replacement
policy produces a higher failure rate and lower replacement levels
than a condition-based scenario.
• A replacement on failure only policy, with low levels of condition-based
refurbishment can result in very low maintenance costs at small failure
rate penalty if coupled with high preventive maintenance effectiveness
and efficiency.
• For the simplified model of the Australian revenue calculation,
condition-based replacement produces the highest retained earnings,
but age-based replacement produces a high level of retained earnings
over a wider range of maintenance effectiveness levels. This effect is
accentuated for higher levels of MTBF of Poor Condition Assets.
• A replacement-on-failure-only can produce better retained earnings
outcome at lower levels of maintenance effectiveness than either
condition- or age-based replacement strategies, particularly when the
MTBF of Poor Condition Asset is high.
• Repair of poor condition assets increases overall failure rates, all else
being the same, but repair of assets generally in good condition is
worth considering.
• The impact of failure rate on the reliability of supply is dependent on
the effective redundancy of the network. The effective redundancy
depends not only on the configuration and loading of transmission
network but also on interconnected networks including other
transmission networks to which there are interconnections, together
with sub-transmission and distribution networks that provide parallel
paths for power flow. Networks with high effective redundancy will be
more tolerant to failure rates than networks with lower effective
redundancy.
• A measure for determining effective redundancy has been proposed:

272
 the number of forced outages of equipment (in a year)/ number of
forced outages resulting in loss of supply (in a year).
This figure could be averaged over five years to account for the
variation arising from non-uniformity in the connectivity and loading of
networks.
• System augmentation tends to mask the underlying impacts of
maintenance and replacement policies on network performance and
profitability. For networks with high growth rates less attention may be
given to optimising maintenance and replacement policies than that
given in low growth rate networks.

273
7 Discussion of regulation of transmission utilities
supported by modelling results
The results from the studies performed for this research highlight the
important influence of regulation on the performance of the transmission
enterprise. One of the main ways that a transmission enterprise can be
influenced to improve its performance is through the mechanism of
regulation. Since transmission enterprises are generally treated as regulated
monopolies regulation is of fundamental importance to the ongoing viability of
the transmission enterprise and the transmission network it supports. It is
therefore valuable to consider the desirable traits of the transmission
enterprise, and the implications for regulation.

7.1 Desirable traits in electricity transmission enterprises

1. The transmission enterprise must provide transport of electricity from
generators to consumers with adequate reliability, quality and availability of
supply.

The required level of reliability, quality and availability of supply varies

between customers. Some industrial customers have high quality power
needs. Research (Sullivan et al., 1996) has also shown that the cost to
consumers of planned outages (with customers given notice of outages) is
lower than for unplanned outages. Planned outages are reflected in
availability of supply. However, in transmission networks built with N-1
redundancy there should not need to be supply outages for maintenance; in
this case availability of supply becomes equivalent to reliability of supply.
Availability of circuits should not impact on reliability of supply, except in the
event that, during a planned outage, the supply on parallel path(s) for
electricity is disrupted. The risk that this occurs is a function of failure rate,
and other factors such as human error in switching operations, and the
number of alternative paths for supply.

2. The transmission enterprise provides its service at lowest possible cost to

users of the network.

274
Cost to users can be narrowly defined as the price of electricity – which
reflects the operation and maintenance costs, level of augmentation and
equipment replacement, debt servicing, financial management and profit-
making by the transmission enterprise. However, in a broader sense, cost
also includes the impact of network constraints on the generation costs or
market price for generated electricity (depending on whether an electricity
market exists), and any environmental costs associated with the transmission
of electricity. In a merit-order despatch system, transmission constraints can
force out-of-merit-order generation despatch. In a competitive market
transmission constraints can cause very high spot prices for electricity which
may be a reflection of costs or opportunistic bidding by generators. Two
factors that impact on cost therefore are effective redundancy of the network
(defined in chapter 7) and circuit availability. In practice, the impact of circuit
availability will be different for different circuits, and depends on effective
redundancy: If effective redundancy for part of a network is high then the
effect of circuit availability on network constraints for circuits in that part of the
network will be low, and vice versa.

3. The transmission enterprise must provide its services in a way that is

sustainable indefinitely into the future.

This reflects on the operation and maintenance practices, replacement and

augmentation. It depends on revenue sufficiency and regulatory governance
ensuring security and continuity of revenue into the future. Profit should be
commensurate with risk. It is noted that risk has artificially been introduced
into the electricity transmission regulation of Australia and New Zealand by
means of the optimisation process.

4. The transmission enterprise should make efficient use of resources

(allocative efficiency).

The transmission enterprise should not be encouraged to undertake

unnecessary replacement of equipment or augmentation of the network. It

275
should make appropriate use of advances in technology consistent with
efficient operation and maximum benefit to society.

Maintenance and operation should also be undertaken efficiently. The level

of maintenance that needs to be undertaken for effective maintenance will
depend on the nature of the plant, the type of technology employed, the age
and the condition of the equipment.

5. The transmission enterprise must employ appropriate safety standards,

both for employees and the public.

This is consistent with sustainable and low cost operation, because failure to
observe appropriate safety standards exposes an enterprise to significant
risk.

6. The transmission enterprise must operate in a way that has minimal

impact on the environment.
Public expectations are that transmission utilities will protect the environment
from pollution including the eradication of use of PCBs in oil. Best practice
for SF6 management minimises leakage of this greenhouse gas into the
environment. There is also increasing pressure on transmission utilities to
minimise visual pollution. This is reflected in opposition to the building of
transmission lines. Less visible options such as underground cables and
compact substations come at the price of increased capital expenditure.
Other short term solutions are also being employed, such as operating
equipment closer to capacity limits, building additional reactive plant and
FACTS devices to increase transfer capability.

7.2 Implications for regulation of transmission utilities

7.2.1 Overall revenue

Overall revenue must be sufficient to serve the needs of the system in order
to ensure its continued viability (both in terms of maintenance and investment
in new equipment). It must also allow sufficient profit to ensure the viability of
the transmission enterprise.

276
The results in this thesis suggest that the adequacy of revenue is a function
of a range of factors, not simply the network’s size or capacity. Maintenance
and operating costs will depend on the age and condition of the network, the
cost of labour, length of travel times, failure rate (including the impact of
climate and loading), the need to undertake work outside normal hours or to
undertake live work. Deterioration of the transmission network’s condition is
a function of climate (particularly ambient temperature), loading and
effectiveness of past preventive maintenance as well as time. These factors
are either outside the control of the transmission enterprise or a function of
past decisions that have consequences for the present network.

Investment needs depend on current loading levels and load factor as well as
anticipated load growth. Any need to upgrade network performance or to
reduce network constraints by increasing capacity, redundancy or replacing
poor condition equipment will also require capital spending. Rapid rates of
system augmentation need to be funded by commensurate increases in
revenue most probably increased debt.

The cost per asset (of same size and function) may also vary significantly,
depending on factors such as ruggedness of terrain, level of seismic activity,
lightning incidence, the likelihood of high winds or ice storms, the length of
lines and the need to employ expensive technology options like GIS and
underground cables (particularly in highly urban areas). Consumers with
high supply quality needs may need to be provided with a transmission
supply with higher specifications for performance – eg higher redundancy
level and/or over-designed equipment; this is likely to result in a more
expensive network. These special requirements are likely to be reflected in
higher revenue requirements and thus higher transmission prices.

Revenue for transmission utilities generally includes return on investment

(profit), return of investment (depreciation) and an allowance for operation
and maintenance costs.

277
7.2.2 Revenue based on return on investment
Revenue based on return on investment is generally a large component of
the revenue for transmission enterprises. For Australia and New Zealand a
figure of 80% for the capital components of revenue was suggested by Jones
(Jones et al., 2000). This component has a large impact on revenue
sufficiency and it is important that it provides appropriate drivers for the
disposition of income.

The return on investment component of revenue has two main parts – the
rate of return that is deemed to be appropriate and the valuation of assets to
which it applies. Asset value appears to be a source of major difference from
one regulator to the next ((Perez-Arriaga et al., 2002)). In the USA return on
investment is largely based on historical asset values. By contrast, in the
Australia it is based on deprival value, interpreted as Depreciated Optimised
Replacement Costs and in New Zealand the Optimised Replacement Value
is used. The use of optimisation can potentially have a significant effect on
profitability as illustrated in Section 6.11. Likewise the use of depreciated
asset value has a negative effect on revenue that is accentuated as the life of
assets is extended beyond their book-life (by means of effective
maintenance, for example). Although interest repayments are decreased
with increasing asset life, profitability can actually decrease with improved
maintenance effectiveness, when coupled with a condition-based
replacement strategy, if depreciated asset value is used as the basis for
calculation of return on investment. In fact Figure 75 suggests that under the
Australian regulatory framework there is an incentive to replace assets on the
basis of age: not only is there no reduction in the return-on-investment
component of profit with increasing maintenance effectiveness, there is also
an opportunity to reduce maintenance costs and increase reliability of supply
by replacing old maintenance-intensive equipment with new less-
maintenance-intensive plant. However, replacement solely on age is not
consistent with allocative efficiency because equipment of the same age can
have quite different performance because of design, duty and environmental-
factors. This problem highlights the necessity to match book-life of assets
with the physical process of asset deterioration.

278
It is noted that a recent discussion paper by the Australian regulator (ACCC,
2003) proposes to apply the regulatory test to all types of capital expenditure
including replacement and refurbishment. At time of regulatory reset the
regulator will assess whether the regulatory test process has been
appropriately applied, in order to decide if a project should be included in the
asset base for purposes of revenue calculation. It is not yet clear what effect
this will have on the asset management practices of Australian TNSPs. The
regulatory test is itself also under review.

Because the usefulness of assets remains constant over most of their life,
and only decreases in the wear-out phase of condition, it is arguable that
deprival value of transmission assets is more closely related to the
replacement value than to the depreciated replacement value of assets.
Deprival value is most closely related to the condition of the assets.

When depreciation is not a factor in return on investment calculation (as in

the New Zealand regulatory framework), capitalised refurbishment and
maximising the life of assets should have a positive impact on profit.

For the traditional US form of rate-based regulation there appears to be an

incentive to keep equipment on the system until it fails, even to the extent of
paralleling equipment, rather than replacing old equipment with new larger-
capacity equipment, because the old equipment continues to attract an
income. Paralleling equipment is perceived to increase redundancy, which is
seen as the key to reliability of supply (See Section 0). This form of
regulation generally allows the cost of operations and maintenance to be
recovered fully, which is perceived to hinder improvements in efficiency. The
lack of incentive to replace equipment can also explain the low use of online
condition-based maintenance reported for North America in the CIGRE
study.

279
7.2.3 Drivers of Operations and Maintenance
A common theme in recent literature on asset management is the pressure to
reduce operations and maintenance costs. This trend has been experienced
in both locations where electricity reform has been introduced and those in
which it has not been introduced. Ideally the aim is to maximise the efficiency
of O& M without compromising its effectiveness. In Chapter 6 the strong links
between maintenance effectiveness and both equipment failure rate and
financial performance were demonstrated. It is therefore important that
transmission enterprises not be driven to reducing maintenance spending
below levels capable of sustaining effective maintenance. The results of
reducing maintenance below adequate levels would not be immediately
evident in system performance but will be seen several years later.

In practice, reported O&M costs can be distorted by accounting practices.

For instance an “asset” might be defined at the level of a bay or it might be a
circuit-breaker. If replacement of part of an asset is considered an expense
then in one case replacement of a circuit breaker would be a maintenance
expense, in the other capital expenditure. Likewise refurbishment can be
considered an expense or an investment. There is also a trade-off between
capital expenditure and maintenance. Purchasing equipment with online
condition monitoring facilities can reduce maintenance costs (by use of online
CBM) but will increase capital costs. Likewise purchase of equipment with
higher specifications or employing more robust technology is likely to cost
more, but may have long-term benefits in reducing maintenance costs and
failures. Thus it is not appropriate to minimise O&M; minimisation of total life
cycle costs is a far better goal, consistent with providing a transmission
service at minimum overall cost to consumers. Despite the concept of total
life cycle cost being well-documented, it does not appear to have been taken
up by regulators who, almost without exception, separate O&M from CAPEX.

In its recent discussion paper (ACCC, 2003) the Australian regulator has
indicated a preference for benchmarking of O&M costs. This practice has
also been adopted in the UK in respect to RECs (see Section 2.2.1.2). As
noted by the ACCC document, nearly all the regulatory difficulty relating to

280
the use of exogenous cost measures arises from the problem of
distinguishing legitimate cost differences from variations due to controllable
efficiency differences alone. The fuzzy modelling in this research strongly
highlights the diverse causes of differences in costs and performance
between different transmission utilities; as a result there is a risk of
inequitable assessments.

7.2.4 Performance-based incentives in regulation of transmission

enterprises
The recent trend in regulation of transmission towards performance-based
incentives (or penalties) is often related to some definition of reliability or
availability.

Reliability of supply is a key customer requirement, but is a function of

effective redundancy as well as system performance (See Section 6.14).
Redundancy tends to mask the effects of poor equipment performance.
Equipment failure rate is itself a function of previous maintenance practices,
so overall reduction in reliability of supply is a symptom of a serious
deterioration of equipment condition. To counteract this requires an
improvement in maintenance practices or an improvement in redundancy.
An improvement in preventive maintenance is not likely to show immediate
results. Refurbishment or replacement of poor condition equipment may be
required at higher levels in order to restore the equipment to good condition.
Thus reducing revenue through penalties based on reliability of supply may
impose a severe burden in the short to medium term, at a time when
additional funds are required to restore equipment condition.

Equipment reliability, while not directly related to customer needs, is less of a

lagging indicator than reliability of supply. Thus it may be more effective as a
parameter on which to base penalties or incentives. Use of equipment
reliability also relates more directly to performance of maintenance, because
it is not affected by system effective redundancy.

Significant emphasis has recently been placed on performance incentives

based on circuit availability in recommendations for transmission service

281
standards for Australian transmission network service providers (ACCC,
2002b). Availability of circuits in general is not a parameter that should be
maximised: 100% availability indicates that either no maintenance has been
undertaken or that it has been done with the equipment in service. There is a
cost penalty associated with live working in addition to personnel safety
concerns, so it should not be undertaken unless necessary. Likewise a low
availability may be function of construction or refurbishment of lines or
substation equipment rather than poor performance. On the other hand it is
appropriate that transmission enterprises be given incentives to minimise
overall costs to consumers including the effect of transmission constraints,
either by maximising availability on critical circuits or by increasing effective
redundancy of the network for those parts of the network participating in
constraints.

7.3 The effect of changing regulation

When regulation changes transmission enterprises have to adapt to the new
circumstances. It takes an appreciable time for changes to filter through the
whole of an organisation: first of all, managers need to understand the
implications of the changes for their organisation; then they need to assess
the strategies that will maximise benefit for their company under the new
conditions; then the strategies have to be implemented, and finally these
need to be reflected in the practices of the company and its corporate culture.
Clearly the skill levels of managers and their ability to initiate and manage
change will vary from one organisation to another, as it does in the wider
industry community.

As a consequence for companies in the throes of regulatory change (as

many have been in the past twenty years) apparent inconsistencies between
external drivers and reported practices can be observed. For example, in the
CIGRE study Eastern European enterprises reported using age-based
replacement strategy, when a condition-based strategy would seem to be
more appropriate for networks with equipment in degraded condition and with
limited funds for replacement and refurbishment of assets.

282
Tools such as those presented in this thesis can assist managers to
understand the impact of both external and internal drivers on the
performance of their networks and their businesses. As such they may assist
to hasten the process of aligning asset management strategy to the new
circumstances that confront them.

283
8 Conclusions
This thesis has broken new ground in the application of organisational
modelling concepts to transmission asset management. It draws on
techniques from Management Science and Artificial Intelligence to explore
problems intractable by traditional engineering methodologies. In doing so it
opens up opportunities for the development of new decision support tools for
transmission enterprises. Indications from discussions with electricity
industry managers are that there are currently no commercially available
tools directed at the strategic level of asset management in electricity
transmission. Furthermore, evidence from available literature suggests that
systematic research integrating the engineering, regulatory and financial
aspects of transmission asset management is very rare. On the other hand,
the pace of change in the electricity industry throughout many parts of the
world has never been so great. There is an urgent need to reassess
traditional policies and practices of asset management, and realign them with
current external drivers. At stake are the financial viability of the transmission
company and the ongoing performance of the transmission network. The
research contained in this thesis is therefore both timely and valuable.

The purpose of this thesis has been to examine the factors that influence
policies and practices of electricity transmission organisations, and to
investigate their impact on the performance of those enterprises
Since research covering the full scope of the problem is scarce, it has
therefore been necessary, in the knowledge acquisition phase of this
research, to draw on a very broad range of literature, together with expert
interviews from industry representatives from various parts of the world, and
extensive personal knowledge of the electricity industry, in order to construct
a model of asset management in the electricity transmission sector. This
conceptual model of the system, developed in Chapter 3, has been
represented concisely in a causal loop diagram. The development of this
generic model is a significant contribution to a new area of research. It

284
clearly illustrates structure of system, indicates the complex feedback
mechanisms and suggests potential interactions of variables.

Implementation of this model has posed significant challenges because of

the nature of the data (a mixture of numeric and non-numeric data;
imprecise; incomplete; and often approximate) and the relationships between
elements of the model (imprecisely known and containing a large number of
feedback loops and some significant delays). Such a system is not readily
modelled using traditional engineering techniques; hence there is no well-
developed body of knowledge within power system engineering to draw on
for this type of research. It has therefore been necessary to apply modelling
techniques and methodologies from other disciplines. Two very different
implementations have been explored in this research: a multi-level fuzzy
rule-based model, and a system dynamics representation. The two
implementations offer different, but complementary, insights into factors
impacting on transmission enterprise policies, practices and performance.
Each shows potential for development as a strategic decision support tool for
use by management of transmission businesses. The application of these
techniques to the modelling of strategic-level asset management of
transmission utilities constitutes a significant contribution to knowledge in this
field.

The fuzzy rule-based implementation was chosen because of its facility to

handle non-numeric data and its robustness to imprecision both in data and
rules. The main limitation of this representation is its inability to model fully
the feedback loops of the system. The model has been structured in
accordance with the causal loop diagram. Rules have been developed
according to the relationships derived from the knowledge acquisition phase
of the research. Time dependencies have been approximated by including
historical variables. Model outputs are semi-quantitative; this permits
comparison of outcomes, but not direct interpretation of results into numeric
values. Verification has been assisted by the graphical representation of the
model (a function of the expert system shell program) and by procedures
developed for the purpose of this research. Validation, to the extent possible

285
with this type of model, has been successfully achieved by developing data
sets representing typical regional values and individual companies and
testing values predicted by the model against published information
(aggregated regional values and information reported for individual
enterprise). This procedure develops confidence in the model’s predictive
power.

There are some unusual and useful features of the fuzzy rule-based model:
• Unlike most benchmarking techniques this tool can be used to
compare a range of enterprises with markedly different operating
circumstances.
• Because of the hierarchical structure of the model, the underlying
causes of difference in performance can be investigated. This can
provide information not usually accessible from benchmarking studies.
• It has been demonstrated (Chapter 5) that the model can be used to
identify opportunities for and barriers to performance enhancement.
Sensitivity of performance outcomes to changes in inputs have been
shown to vary from one enterprise to the next, suggesting that
solutions effective for one utility may not be appropriate for use by
another transmission enterprise. Application of the model to such
problems may therefore avoid costly mistakes.

For the second implementation explored in the course of this research,

system dynamics techniques have been employed. The model was
constructed at a high level of abstraction, consistent with strategic level
investigations, and has proved to be a flexible and versatile tool for
examining the relationships between asset management strategies and
performance. This technique was chosen because of its ability to represent
time dependencies in the model. The structure was developed from the
causal loop diagram, but only key elements of the CLD were represented.
Relationships not precisely known were approximated by mapping
relationships graphically. Verification was expedited by use of a graphical
user interface for model construction. The model was developed
incrementally, and tested at each stage of development. Validation, as in the

286
previous model, was an exercise in building confidence that the outcomes
generated by the model were explainable and consistent with observed data.

The model, which links the age and condition distributions of an asset
population, has provided valuable insights into the nature of transmission
asset management. This structure enables the financial aspects of asset
management to be linked with maintenance, refurbishment and replacement
strategies, including those based on asset condition. Whereas the fuzzy
model allows for a breadth of analysis, the system dynamics model allows
deeper analysis of the effect of asset management strategies on profitability.
The impact of maintenance effectiveness, and condition-based maintenance
on asset valuation and revenue for age and condition-based replacement
and refurbishment scenarios have been examined for the Australian system
of revenue determination for transmission enterprises. The model can
readily be adapted to other types of regulation.

The system dynamics investigation has highlighted the importance of

maintenance effectiveness – quality and quantity of maintenance applied to
assets – as a contributor to profitability and equipment performance. It has
also revealed the key role that availability of capital can play in limiting asset
management options, particularly for asset replacement decisions. In the
CLD and fuzzy models this was reflected in the importance of the parameter
revenue sufficiency.

Significant impact of system design, configuration, loading and load factor on

network performance is evident in the CLD and fuzzy models, but has not
been implemented in the system dynamics model, although the likely effects
have been discussed. It has been proposed that the aggregate effect of
these system factors on performance may be measured using effective
redundancy, defined in Chapter 6. This could be a useful input parameter for
performance benchmarking studies favoured by regulators (for example in
DEA studies). Verification of this is outside the scope of this research, but
could provide an avenue for further research.

287
Another non-trivial but potentially fruitful avenue for research is to model
maintenance spending using the system dynamics model. As the CLD and
fuzzy models have demonstrated, a complex combination of factors
contributes to maintenance spending: Maintenance effectiveness relates to
routine time- or condition-based maintenance activities, including minor
corrective actions. It impacts on equipment condition, and through this,
failure rate. Failure rate of poor-condition assets may be reduced by
effective use of condition-based refurbishment and/or replacement of
equipment. Routine, corrective, remedial and condition-based maintenance
all contribute to the cost of maintenance. Refurbishment and replacement
costs may also be included in maintenance spending, depending on
accounting practices of the utility. The efficiency with which maintenance is
undertaken (including skill and targeting of maintenance activities to need)
has a bearing on cost, together with the cost of parts and labour. There may
be an optimum level of maintenance spending per asset (of a particular type)
that depends on the method by which revenue is determined and the
effective redundancy of the network. It would be extremely useful for utilities
to be able to explore relationships between profitability, maintenance cost
and maintenance effectiveness for different asset management strategies.
The results of this research suggest that system dynamics modelling is an
appropriate technique for this type of investigation.

In summary, this research has investigated the factors impacting on asset

management and performance of electricity transmission enterprises
throughout the world. Understanding of interactions between these factors
has been developed through extensive investigations of literature, expert
interviews and personal industry knowledge. Since the analysis is not
amenable to solution by traditional engineering modelling techniques, the
relationships have been explored using innovative approaches drawn from
Artificial Intelligence and Management Science. The application of these
techniques to asset management of transmission enterprises represents a
significant contribution to knowledge. The models thus produced provide
insights into the key factors and relationships within the system. They show
potential for development as strategic decision support tools for transmission

288
businesses operating today in a climate of rapid change, against a backdrop
of existing network, traditional practices and limited tools for analysis.

Asset management in transmission enterprises globally will be enhanced by

frank and thoughtful sharing of information among utilities. However for
maximum benefit to be achieved an understanding of the whole asset
management framework is needed. Future opportunities and threats can be
best assessed in the context of a thorough understanding of current situation,
existing relationships and drivers. This research has made a contribution to
knowledge in the area of strategic asset management, but there is great
potential to extend this knowledge further, for the benefit of transmission
utilities, their networks and the people who use them.

289
Summary of Contributions
The main contributions of this thesis are:
• A review of relationships between drivers of asset management in
electricity transmission, network development, system and enterprise
performance;
• Development of a generic model of policy-level asset management in
electricity transmission organisations, applicable across a wide range
of transmission enterprises;
• Application of multi-level fuzzy rule-based modelling techniques to
transmission asset management;
• Demonstration of how this model can be used to identify causes of
difference in performance between utilities and to identify strategies
for and obstacles to performance enhancement within a transmission
enterprise;
• Application of system dynamics modelling to transmission asset
management;
• Development of a model effectively linking the ageing and condition-
degradation process, and through this connecting maintenance and
replacement strategies with equipment performance and financial
performance of a transmission business;
• An assessment and comments on the relationship between equipment
and network performance, relating this to maintenance and
replacement strategies in different parts of the world; and
• A review of the impact of regulation on transmission asset
management and performance.

290
Appendix A Sources of Data

A.1 General Data Sources and treatment of data

The sourcing of usable data has been a major consideration in the
development of cases for evaluating validity of the fuzzy model. Many data
concerns have arisen regarding the completeness, accuracy, and
consistency of data. Compromises and approximations have had to be made
on occasion. An attempt has been made to match the time-frame of the data
as closely as possible to that of the CIGRE study, which was carried out in
1997 - 1998. Unfortunately the Asian Financial Crisis occurred at the end of
this period, so financial data from Asia (especially currency-related items like
debt level and credit rating) are in some cases affected by this event.

Occasionally a factor that is considered significant has had to be estimated,

using other data. For example, loading has been estimated using system
losses, but the losses data is generally available for transmission and
distribution together, rather than transmission alone. Distribution losses are
usually significantly larger than transmission losses. In addition, distribution
suffers from pilfering, especially in some developing countries, and therefore
losses can be significantly inflated by this “non-technical” element. To
compensate for this the extent of pilfering has also been estimated in the
model.

Factors of lesser importance, for which no data were available, have been
omitted from the model. This will have a secondary level effect on availability
and the maintenance spending.

The listing of sources of data is presented in Table 34. As this table

indicates, a significant portion of the data has been taken from the CIGRE
study. As a consequence the regions studied using the fuzzy model are
those represented in the CIGRE study, and in the groupings used by the
CIGRE Working Group in their analyses of this survey. (See Chapter 2 for
further detail on the results of this survey). The survey results are dominated

291
by Western European data, and data from some regions, especially South
America, are quite incomplete. One obvious error in maintenance spending
for an Australian company has been corrected (from figures published in its
Annual Report of that year).

In general, although a few other dubious figures have been detected it has
not been possible to correct these because while the raw data from the
survey is accessible, the company (and country) details have been omitted.
For cases of data that are clearly invalid, these data have been omitted from
averages used in this study. The lack of country information has made it
difficult to match data from the CIGRE study with data from other sources (eg
labour costs, load growth). For example it appears likely that three of the
nine respondents from Asia are from Japan. For Japan the labour rate is
much higher than the average labour cost in Asia, and the load growth rate is
much lower than many countries in Asia. Hence if the make-up of the
sample group from Asia from the CIGRE study were known, the average
labour cost for this group could be significantly higher than the average for
Asia as a whole. This has implications for maintenance spending predicted
for this case.

Furthermore, it is likely that the group of respondents to the CIGRE survey

are biased toward the better performing transmission enterprises generally –
those who have an interest in keeping abreast with current trends in Asset
Management. This is perhaps less a factor in North America, where there
appears to be less interest in Asset Management generally than in other
countries, and perhaps also less involvement in CIGRE than in other
countries.

Interpreting the survey results is not always straight-forward. In many

questions respondents were asked to indicate a positive response to a
question with a cross, but lack of a cross could indicate a negative response
or could simply mean the question has not been answered. Interpretation
has to be based on the context. Similarly many questions asked for either
general and/or equipment specific responses. Some respondents gave only

292
general responses, some only equipment-specific responses, some both and
some none. This complicates the analysis and assumptions on the meaning
of null or general responses need to be made. Examples of these types of
analysis were given in Section 2.1 and its subsections.

Another problem arises from the interpretation of “Yes /No answer”

responses. The importance of various issues was estimated by calculating
the percentage of positive responses (by region). If 100% or 0% of
respondents indicates a positive response then the average value is clearly
representative of sentiment on this issue. However, if 50% give positive
responses this does not necessarily mean that the attitude of respondents is
typically 50% but merely that there is a range of attitudes reflected in the
responses. Clearly this is an approximation, but reporting of much of the
information contained in the CIGRE study is rare, and no better sources have
been found.

Table 34 Data used in fuzzy model and their sources

Attribute Source
Age as replacement criterion (CIGRE SC 23 and 39, 2000)
Interpreted from various sources including interviews
(Allison and Jay, 2001, Bower, 2001, Lee, 2001, Mohd
Noor, 2002, Sherard, 2001, Stewart, 2001, Teixiera,
Attitude to technology 2001), comments in technical papers (Davies et al.,
1998, Allison et al., 1995, Urwin, 1999, Clippinger and
McDermott, 2001, Jones et al., 2000, CIGRE AP13,
2001, Bell, 2003)
Financial Analyses: (Standard & Poor's, 1999, Richer,
2000b, Standard & Poor's, 2001, Standard & Poor's,
2002)
Business acumen
Unless specifically known, private companies are
assumed to have a slightly higher level than public
companies.
Climate Encarta (Microsoft, 2001)
Climatic factor Encarta (Microsoft, 2001)
Combining of maintenance activities
(CIGRE SC 23 and 39, 2000)
(by circuit)
Connectivity factor (CIGRE SC 23 and 39, 2000)
USA: (Lee, 2001)
Japan: (IEA, 1999)
Cost of land
Unless otherwise known, a higher value is assumed for
more highly urbanised areas
Standards & Poor’s credit information (Flintoff, 2001,
Flintoff, 2002, Mock and Bodek, 2001, Hauret, 2001,
Credit rating Zsiga, 2002, Plantagie, 2000, Duke Energy, 2002)
(Richer, 2000b, Standard & Poor's, 2001, Standard &
Poor's, 2002)
Standard & Poor’s financial information (Richer, 2000b,
Flintoff, 2002, Standard & Poor's, 2002, Standard &
Poor's, 1999)
Asia Week (Shameen, 2001, Singh and Poh, 1996,
Debt level
Singh and Poh, 1997)
EIA Country Analysis Briefs, including (EIA, 2001b, EIA,
2001c, EIA, 2001d, EIA, 2001a, EIA, 2002b, EIA,
2002a)
Earth quake zone Encarta (Microsoft, 2001)

293
Attribute Source
USA: (Hirst, 2000, Kirby, 1999)
External barriers Western Europe: UK (Jay and Williams, 2001, Allison
and Jay, 2001), Switzerland: (Buckner, 2002)
Flexibility (CIGRE SC 23 and 39, 2000)
Frequency of failure as replacement criterion (CIGRE SC 23 and 39, 2000)
Geographic factor (CIGRE SC 23 and 39, 2000)
Ukraine: (Lovei, 1998)
Russia: (Wilson, 1998)
Lithuania: (Standard & Poor's, 1999)
Historical revenue sufficiency
Argentina &Latin America: (EIA, 1997)
Japan: (IEA, 1999)
Australia (ACCC, 2002a, ACCC, 2000)
World Bank data labour cost taken as same as
Labour cost
manufacturing sector (World Bank, 1999b)
Unless known an moderate value has been assumed
Labour skill Malaysia :(Mohd Noor, 2002)
Western Europe:(Davies et al., 1998)
Load factor (CIGRE SC 23 and 39, 2000)
Load growth (recent) (World Bank, 2000a)
Losses (T&D) (World Bank, 2001c)
Various sources including interviews (Allison and Jay,
2001, Bower, 2001, Lee, 2001, Mohd Noor, 2002,
Sherard, 2001, Stewart, 2001, Teixiera, 2001),
technical and economics papers:
Maintenance history (historical maintenance Ukraine: (Lovei, 1998)
effectiveness) Kazakhstan: (World Bank, 1999a, Ametov et al., 1999,
Standard & Poor's, 1999)
Argentina &Latin America: (EIA, 1997)
Australia & New Zealand: (McMahon, 1995);(Allan et al.,
1992, Allan, 1993) (ACCC, 2002a)
Maintenance relative to manufacturers’ recommended
(CIGRE SC 23 and 39, 2000)
levels
Maintenance spending as KPI (CIGRE SC 23 and 39, 2000)
Number of failures as replacement criterion (CIGRE SC 23 and 39, 2000)
India: (Audinet, 2002)
Latin America: (Rudnick and Zolezzi, 2001)
Pilfering
Malaysia: (Mohd Noor, 2002)
Value assumed to be low unless specifically known
Profit as intention (CIGRE SC 23 and 39, 2000)
Proportion of aged equipment (CIGRE SC 23 and 39, 2000)
Proportion of young equipment (CIGRE SC 23 and 39, 2000)
(CIGRE SC 23 and 39, 2000)
Redundancy
India: (Audinet, 2002)
(Standard & Poor's, 1999) (Standard & Poor's, 2001,
Standard & Poor's, 2002)
Regulatory support
South Africa: (Plantagie, 2000)
Japan: (IEA, 1999)
Replacement considered an investment (CIGRE SC 23 and 39, 2000)
USA: (Teixiera, 2001, Sherard, 2001)
Revenue sufficiency (available financial resource)
Russia, Croatia, Lithuania : (Standard & Poor's, 1999)
Ruggedness of terrain Encarta (Microsoft, 2001)
Size of company (CIGRE SC 23 and 39, 2000)
Generally no special relationship except for case studies
Special relationship with vendors of individual companies where a special relationship is
known to exist

A.1 Development of Regional Case Data Sets

In the following subsections data sets are developed for each of the regions
represented in the CIGRE study. They provide input data for regional case
studies, as described in Chapter 5. These data sets, together with data sets
for individual case studies (Section 5.3) have been used to adjust the rules of
the fuzzy model, with the aim of producing sets of results from the model that
are consistent with those anticipated for each region, and also consistent with

294
known outcomes for individual case studies. It is therefore important that the
regional data sets developed in the following subsections reflect, as closely
as possible, typical regional data profiles. Some items within the regional
case sets are also used for the individual cases, where case-specific data
have not been available.

Asia
The electricity industry in Asia is still largely comprised of vertically integrated
utilities, mainly state owned, except in Japan where there are ten privately
owned utilities. Regulation is largely rate-based and fairly supportive,
reflecting the strategic importance of the industry to the economic
development of the region. However there is a driver, especially in cases
where the government is the regulator, to hold prices as low as possible
particularly for domestic consumers, and this can impact on revenue
sufficiency. In the worst cases, tariffs for electricity do not enable companies
to cover costs (Audinet, 2002). Table 35 summarises financial appraisals
from (Standard & Poor's, 1999) for the Asian region. Note that many
countries in the region were at the time experiencing an economic downturn
as a result of the Asian financial crisis. This exposure to currency
fluctuations is reflected in the effect on debt coverage and consequently in
credit ratings (foreign currency ratings). However, it can be observed from
the table that the debt level in Asian electricity enterprises is typically fairly
high, and interest coverage from revenue fairly low, except in Japan, where
utilities have historically enjoyed generous regulatory frameworks, and
access to low interest capital (Standard & Poor's, 1999). In the case of a
government-owned enterprise the debt incurred by the electric company is
typically guaranteed by the government, and therefore attracts a higher credit
rating than would a stand-alone company

295
 Table 35 Enterprises in Asia – Business parameters at 1998

sufficiency
Regulatory
Ownership

Debt level

Business
Function

currency

Revenue
Country

acumen

support
foreign
Credit
rating
Name

local/
##
#

*
Japan Chubu P GTD AA 82.9% high well managed v. supportive 3.5x int. cov.
Japan Chugoku P GTD AA- 84.4% v. supportive 3.5 x int. cov.
Japan Hokkaido P GTD AA 77% well managed v. supportive 4.1x
Japan Kansai P GTD AA 81% v. supportive 3.5 x int. cov.
Japan Kyushu P GTD AA 82% v. supportive 3.6 x int. cov.
Japan Shikoku P GTD AA-- 72.8% v. supportive 3.1x int. cov.
Japan Tohoku P GTD AA- 82% v. supportive 3.3 x int. cov.
Japan Tokyo P GTD AA 87.8% v. supportive 3.3 x int. cov.
A
Thailand Egat G GTD 65% well managed supportive 2.8x int. cov.
BBB-
A+
Hong Kong Hong Kong Electric Co. P GTD <40% well managed; very profitable v. supportive 7.1x int. cov.
A
BBB-
Korea KEPCO G GTD 56% supportive
(for. cur)
2.1 x int. cov.
uncertain
Philippines National Power Corp. G GTD BB+ 80% (1997)
support
1.7x (1998)
moderate
Taiwan Taiwan Power Co. G GTD AA 37.9% conservative 2.5x int. cov.
support
1.5 x int. cov.
Tenaga Nasional BBB low prices, adequate
Malaysia 78% G GTD (1998)
Berhad BBB- operational performance
5.5 x in (1995)
# Ownership: G – government; P private
## Functions G – generation, T– transmission, D – distribution
* int. cov.: interest coverage from earnings (presented as a number of times)

296
The companies listed in Table 35 are some of the better performing
companies in the region. There are others with far worse financial situations,
and poorer electricity supply. In India for instance many of the state-based
transmission and distribution companies suffer from under-funding, partly
from tariffs that are too low, partly from chronic levels of pilfering (Audinet,
2002). Consequently the networks in these areas suffer from inadequate
capital and maintenance expenditure, resulting in poor operational
performance. In Cambodia an Asian Development Bank news release from
2000 (Asia Development Bank, 2000) states that “the country lacks adequate
power everywhere except Kompong Cham, Phnom Penh, Siem Reap, and
Sihanoukville. In the needy provinces, private sector operators have neither
the capital nor the creditworthiness to fund rehabilitation or expansion of their
systems.

The high growth in electricity sales in the region on average has contributed
to the debt exposure of companies by necessitating the augmentation of
electricity networks. In some cases this has led to a purchasing regime
based on minimum capital cost rather than minimal life cycle costs for
equipment (Mohd Noor, 2002). In Japan, however, high level (conservative)
specification of equipment has apparently resulted in lower failure rates (in
GIS equipment at least (Chan et al., 1998)), compared with other countries
as well as higher capital expenditure. Because of the high growth rate the
average age of equipment is typically fairly low in Asia.

In Asia there is also a wide range in the maturity of electricity networks.

Japan, Singapore and Hong Kong have mature electricity networks, while in
many other areas rural electrification is far from complete: In 1998
UNESCAP reported, for example, that 13 % of the rural population in
Bangladesh had access to electric power, 25% in the Lao People’s
Democratic Republic and 38% in Pakistan (UNESCO, 1999).

In developing countries in Asia there was strong growth in electricity

consumption in the period prior to the Asian Financial Crisis, and strong
projected load growth at the time. In developing countries of Asia this strong

297
growth has been coupled with the increase in the number of Independent
Power Producers in the region (EIA, 2002d). These typically operate in
parallel with the existing state-owned electric company, and transmission
function is carried out by the state-owned enterprises. In some countries
electricity growth has been supplemented by a significant rural electrification
program: Bangladesh, India, Pakistan, Sri Lanka, Thailand, Viet Nam and the
Lao PDR (UNESCO, 1999). By 1998 almost all villages in Thailand had
access to electricity.

Table 36 shows the growth in energy production from electricity in Asia,

during the period up to 1998. The average annual growth in energy
production is quite high by world standards in most of Asia during this period.
The rate of load growth is approximately proportional to energy growth if
power factor is assumed to be the same. However, this is not a valid
assumption for cases (such as India (Audinet, 2002)) where peak load is
constrained by available generation. In such cases the power factor of the
load is likely to be quite high.

Table 36 Production of electricity in billion kWh from (World Bank, 2001b)

Country 1980 1998 average% annual growth 1980-1997
Bangladesh 2.4 12.9 10.1
China 300.6 1,166.20 8.6
India 119.3 494 8.6
Indonesia 8.4 77.9 13.5
Japan 572.5 1,036.20 3.9
Korea,
37.2 235.3 12.1
Republic of
Malaysia 10 60.7 10.9
Myanmar 1.5 4.6 6.7
Nepal 0.2 1.3 11.1
Pakistan 15 62.2 8.9
Philippines 18 41.2 4.1
Singapore 7 28.6 8.3
Sri Lanka 1.7 5.7 6.5
Thailand 14.4 90.1 12.5
Vietnam 3.6 21.7 10.4

Just as electricity system maturity varies significantly throughout Asia, there

is a wide range of labour costs in the region (Table 37). Japan has one of
the highest labour rates in the world, but some countries in the region have
labour rates almost fifty times lower.

298
 Table 37 Labour cost in $US (World Bank, 2000b)
Country Labour cost (manufacturing) in $US 1995-1999
Bangladesh 671
China 729
Hong Kong 13539
India 1192
Indonesia 1008
Japan 31687
Korea, Republic of 10743
Malaysia 3429
Pakistan 6214
Philippines 2450
Sri Lanka 604
Singapore 21317
Thailand 2705
Vietnam 711

The average of these values is approximately $US7000, relatively low by

world standards.

Climate is also diverse across this large region, ranging from tropical and wet
in equatorial regions through to subtropical and temperate further north.
Some areas such as Malaysia have a high lightning incidence and the
southern parts of Asia experience tropical storms that contribute to failure
rates of equipment. Some countries in the region are prone to earthquakes
and volcanoes, notably Japan, Indonesia and the Philippines.

Australia, New Zealand and South Africa

The electricity industry in Australia is state based, and in New Zealand and
South Africa, national. In the Eastern and Southern states of Australia, the
vertically integrated state-owned electricity enterprises have been separated
into generation, transmission, retail and distribution, with generation
separated further into competing generation businesses, and retailers
competing for customers. The transmission systems in Australia are
relatively small in terms of energy transferred, but cover large areas. The
transmission enterprises are regulated monopolies operating under a
revenue cap regulatory regime. Two of the transmission companies – those
operating in Victoria, and South Australia – have been privatised, whereas
the remainder are still government-owned enterprises.

299
The revenue cap of Australian transmission companies is calculated based
on the Depreciated Optimised Replacement Costs (DORC) of assets
multiplied by the weighted average cost of capital (WACC) plus an allowance
for depreciation and operation and maintenance (O&M) costs. The operation
and maintenance allowance is adjusted for inflation less a factor X, an
incentive for reducing O& M costs. A new incentive-based scheme has been
recently introduced for service level performance targets, particularly
availability and reliability.

New Zealand operates under a “light-handed” regulatory framework where

the state-owned transmission company sets the tariffs (Jones et al., 2000),
based on Optimised Replacement Costs of assets.

In South Africa Eskom is the state-owned vertically integrated utility, the fifth-
largest utility in the world (Plantagie, 2000). It operates in a supportive rates-
based regulatory environment. Eskom has been involved in a major rural
electrification program providing electricity to an additional 1.75 million
homes in the period from 1994 to 2000 (Plantagie, 2000). This has required
additional transmission infrastructure to maintain a high level of supply
reliability (system minutes lost 3.1 in 1999, 5.3 in 1998 (Plantagie, 2000)).
Eskom’s operation is considered sound, its local currency credit rating is A-
and its foreign currency rating was BBB- in September 2000. In 1998 its debt
level (total debt/total capital) was 58% (fairly high) and its funds from
operations interest coverage was 2.4 x (all financial figures from Standard
and Poor’s (Plantagie, 2000)).

Table 38 shows the debt and funds from operations coverage of interest for
four publicly-owned electricity companies in Australia. Western Power is
vertically integrated, but the others are transmission only. The Tasmanian
utility is very lightly leveraged compared with the others, which are
moderately leveraged.

300
 Table 38 Debt and interest coverage for government owned enterprises in Australia
1997-1998 and 1998-99 (Productivity Commission, 2002)
Powerlink Transgrid (New Transend Western Power
(Queensland) South Wales) (Tasmania) (Western Australia)
Year 97-98 98-99 97-98 98-99 97-98 98-99 97-98 98-99
Debt to total assets
38.4 41.1 37.4 35.2 n.a. 3.5 65.6 64.7
(%)
Interest coverage
2.7 2.5 2.0 2.0 n.a. 91.7 x 2.4 2.3
(x)

Among the privately owned transmission companies in Australia, GPU

Powernet owned the Victorian transmission network in 1998 (it is now owned
by SPI Powernet.) GPU Powernet was a wholly owned subsidiary of Austran
Holdings. This company had a credit rating A at 1999 based on GPU
Powernet’s stable, predictable regulatory regime, monopoly status and sound
operational record (Standard & Poor's, 1999). The company had at that time
a high debt level (total debt/total capital) of 73% and funds from operations to
interest coverage of 1.7 times.

The South Australian government sold (200 year lease) its transmission
network to Electranet Pty Ltd in 2000. Standard and assigned a credit rating
of BBB+ to Electranet (Standard & Poor's, 2001 p150) The ACCC considers
Electranet’s business profile to be in the range above average to excellent
(ACCC, 2002a p102). However, it does have an aggressive financial policy
(gearing ratio of 80%)(ACCC, 2002a p115).

Typically the electricity systems in Australia are mature, with nearly all the
population having access to electricity supply through an electricity network
or, in remote locations, a local supply. The augmentation of the network
therefore is in line with loading and load-growth, which varies from one state
to another, but is generally moderate.

Eastern Europe, Russia and former Soviet states in Central Asia

During the 1990s the states of the former Soviet Union have been
undergoing a period of transition from a communist to a capitalist economy,
following the collapse of the Soviet Union. Transition period has been
categorised by the following phenomena:
• There was a period of sharp decline in output from these economies.
In Central and South Eastern Europe and the Balkan states (CSB) this

301
 decline occurred in 1990 to 1993 followed by a period of gradual
recovery. In the Commonwealth of Independent states (CIS) the
decline continued until about 1998 (World Bank, 2001a).
• The industrial sector shrank, and the service industry grew. The
output from industry and agriculture in 2001 was 45% of 1990 levels
(World Bank, 2001a). This had an inevitable flow-on to electricity
usage, and demand from industrial loads dropped significantly in the
period (Dobozi and Pohl, 1995).
• Private sector output grew substantially during the 1990s, especially in
some of the CSB countries (notably the Czech Republic, Hungary,
Estonia and Romania).
• Poverty increased sharply during the period, because of falling output
and an increase in income inequality.
In the electricity industry in this region the following effects have been
observed:
• There has been a reduction in electric energy consumption and peak
load during the period. Power intensive industries were among the
hardest hit by industrial restructuring (Dobozi and Pohl, 1995).
• Electricity prices rose substantially during the period, but there was
widespread non-payment problems (Dobozi and Pohl, 1995).
• There was some technological regress due to reduced investment and
lack of funding for basic maintenance and repair work (Dobozi and
Pohl, 1995). The European Commission has estimated that some
tens of billions of USD in investment funds are needed to upgrade
Eastern European power systems to international standards (von
Hirschhausen and Opitz, 2001)
• The lack of a well-functioning judicial system and poor regulatory
governance, substantial institutional instability and regulatory risk
make investment in power system infrastructure in transition countries
risky (von Hirschhausen and Opitz, 2001).
• The development of competitive electricity markets is hampered by
inadequate metering and communications systems in some countries
(von Hirschhausen and Opitz, 2001, World Bank, 1999a).

302
 Prior to the 1990s the electricity industries in Soviet Block countries were
typically state-owned vertically integrated monopolies, subject to some
form of rate-based regulation (von Hirschhausen and Opitz, 2001).
Power consumption in the Soviet Union was among the highest in the
world. Capacity utilisation was below international standards. The
connectivity of the network was relatively high (von Hirschhausen and
Opitz, 2001).

By the end of the twentieth century separation of generation, transmission

and distribution had been effected in some countries. There was still very
little competition in generation or retail sectors. Transmission was almost
all state-owned and controlled. The regulatory authority was mostly under
the direct control of the government, with associated lack of transparency.
Price reform has been very slow, and at 1999 none of the Eastern
European countries had achieved cost-reflective pricing (von
Hirschhausen and Opitz, 2001). A small number of companies involved in
transmission have been assigned credit ratings from Standard and Poor’s
(Table 39).

The Hungarian government has pursued modernisation of it electricity

system by foreign investment, in generation and distribution, but at 2000
the transmission system was state-owned (von Hirschhausen and Opitz,
2001).

Table 39 Credit ratings from some former Soviet Block companies

Country Company Functions Credit rating Financial status
low collection rates 85%
Russian Fed. Irkutskenergo, AO ETE GTD CCC-
(25% in cash) in 1998
Kazakhstan Electricity BB- local 4 x debt coverage
Kazakshtan T
Grid Operating Co. B+ foreign 13% debt leverage **
9 x interest coverage *
Russian Fed. Monsensergo GTD CCC payment collection
problems, currency risk
## Functions G – generation, T– transmission, D – distribution
* Interest coverage: interest coverage from earnings, expressed as a multiplier
** debt / debt plus equity as a percentage.

Data from the CIGRE study suggests that only a small percentage of
equipment is over forty years old, the typical life expectancy for equipment in

303
Western Europe. This is at odds with documents suggesting that a large
proportion of equipment is past its design life and in need of replacement.
However, it would appear that the design life of (the predominantly air-blast
technology) transmission switchgear in the USSR was 25 years (Batyayev et
al., 1994).

North America (USA and Canada)

In the USA there are a number of different categories of electricity enterprise.
Utilities are privately-owned companies or public agencies engaged in
generation, transmission and/ or distribution of electric power for public use.
In 1998 there were 3170 utilities in the USA, of which 239 were investor
owned (EIA, 2000). These investor-owned utilities, however, accounted for
more than 75 percent of utility sales to ultimate customers and total utility
generation (EIA, 1996, p9). There are ten utilities owned by the federal
government. In 1995 there were over two thousand other publicly owned,
not-for-profit utilities including Municipals, Public Power Districts, State
Authorities, Irrigation districts and other state organisations. There were, in
1995, also 931 cooperatively owned utilities in the USA, which are owned by
members, and mainly serve members (EIA, 1996).

Non-utilities are privately-owned enterprises that generate power for their

own use and/or for sale to others. The number and nameplate capacity of
non-utilities has been increasing steadily since they first became legal in
1978 after the passage of the Public Utility Regulatory Policies Act (PURPA).
PURPA was designed to encourage co-generation and renewable energy
projects. Utilities were required to connect to and purchase from qualifying
generators. In 1992 the Energy Policy Act (EPACT) permitted the entry of
non-utility generators not qualifying under the special conditions stipulated by
PURPA.

Regulation was traditionally undertaken by the states, using rate-based

regulation. Utilities traditionally operated as monopolies. Large variations in
retail electricity prices from state to state have been motivation for electricity
reform in the U.S. (EIA, 1996) p35.

304
The transmission grids of the United States have evolved into three large
networks.
• The Eastern Interconnected System covers the eastern two-thirds of
the U.S and strongly interconnected with the Canadian system.
• The Western Interconnected System includes states in the South
West, and areas west of the Rocky Mountains, and is also strongly
integrated with the Canadian system.
• The Texas Interconnect has interconnections with Mexico.
The transmission networks are largely owned and operated by some of the
larger utilities (EIA, 2000). Wholesale trade in electricity (reselling) has been
increasing gradually, and is being encouraged by the federal government.
Specifically, in 1996 the Federal Energy Regulatory Commission (FERC)
issued Order 888 which opened (non-discriminatory) transmission access to
non-utilities, and Order 889 which required utilities to develop electronic
systems to share information about available transmission capacity.

Because regulation of electricity is largely state-based, the extent and timing

of electricity reform has varied from state to state. At December 1999,
twenty-four states had enacted legislation permitting consumers to choose
their electricity retailer (EIA, 2000). Generally, the states having highest retail
electricity prices have been quickest to legislate to introduce competition.
Overall there has been a recent trend for utilities to sell generation to non-
utilities. In some cases this has been driven by state legislation requiring
transmission to be separated from generation and distribution (EIA, 2000).

The changes taking place in the North American electricity industry have had
a number of effects on transmission asset management:
• In the uncertain regulatory environment utilities have been reluctant to
commit to development of transmission augmentations, when
adequate remuneration has not been assured (Hirst, 2000).
• Increases in wholesale electricity transactions have changed power
flow patterns, highlighting transmission constraints.

305
 • Transmission constraints have been further highlighted by the market
failure in California in 2000.
• As a result there have been calls for regulators to encourage the
building of transmission (since it represents only a small part of total
cost, but has an important role market operation.)
• Utilities have been reducing operations and maintenance costs in a
bid to be more competitive.
• Regulators have been more conscientious about keeping operations
and maintenance costs down, and therefore have applied pressure on
utilities to reduce costs.

Apart from uncertainty arising from the ongoing restructuring of the electricity
industry some other factors have also been identified as contributing to a low
level of network augmentation (especially interconnections) (Hirst, 2000). In
the US there is strong public opposition in some parts of the country to the
building of electricity infrastructure, including issues of visual pollution,
degradation of land prices, benefits accruing to other places rather than
where the line is to be situated. This is further complicated by the need to
deal with multiple sets of regulations and public authorities when proposed
transmission lines cross state boundaries or federally-owned land.

Despite concerns about lack of transmission investment reliability of supply

has apparently not suffered degradation. Hirst (Hirst, 2000) suggests that
industry has found substitutes for transmission augmentation:
• System operators are making use of improved data collection and
dynamic ratings to run the networks closer to operating limits.
• New gas-fired generators are being located close to load centres,
thus reducing need for transmission.
• Small-scale investments (in reactive plant, and line upgrades)
increase capacity sufficiently in the short term.

Because of the dominance of large, usually privately-owned, vertically-

integrated utilities in the transmission sector, the transmission network has

306
developed certain characteristics. Networks are mature and highly
interconnected. Network expansion is viewed favourably by investors and
regulators, but there is pressure to maintain low operation and maintenance
costs. This is also confirmed by the CIGRE maintenance survey in which all
respondents from North America reported doing less maintenance than
recommended by manufacturers. The CIGRE study also suggests only low
overall usage of condition-based maintenance in this region. Offline
condition-based maintenance appears to be common for expensive
equipment like substation transformers (Lee, 2001, Sherard, 2001).
Equipment is generally kept in service until it has failed (Lee, 2001, Teixiera,
2001). Replacement of failed equipment is favoured over refurbishment.
Age does not seem to be considered a criterion for replacement of equipment
(Sherard, 2001, Teixiera, 2001).

The focus of reliability is on redundancy in network design rather than

maximising the reliability of individual pieces of equipment. This attitude
finds expression strongly in industry-sponsored documents. For example in
the Edison Institutes response (Owens, 2001) to the US Department of
Energy’s National Transmission Grid Study the author states:
Redundancy is the best defense for both reducing congestion (and
making markets work) and keeping the system secure.
When additional capacity is required on the network, equipment is added in
parallel with existing plant. This is perceived to increase redundancy on the
network (Lee, 2001).

Credit ratings of US utilities are typically around the BBB to low A levels, with
the median level BBB+ in 2001 (after the Californian electricity market crisis)
(Standard & Poor's, 2001). The overall trend between 1999 and 2001 was
negative.

South America
Industry restructuring has been a feature of the electricity industry in South
America. Chile in 1982 was one of the first countries in the world to set up an
electricity market. Argentina in 1992 adopted extensive market-based reform
of its industry. This was followed by reform in Peru in 1993, Bolivia and

307
Colombia in 1994 and Central American countries Panama, El Salvador,
Guatemala, Nicaragua, Costa Rica and Honduras in 1997. Brazil, Venezuela
and Ecuador have also started an electricity reform process. Extensive
privatisation within the electricity industry has also occurred in Chile,
Argentina, Bolivia and Peru (CIER et al., 2001).

While electricity consumption per capita is still relatively low there has been a
rapid growth in demand for electricity.

Prior to the reform process the electricity industry was categorised by state-
owned vertically integrated utilities, suffering from inadequate investment in
generation, transmission and distribution capacity (Rudnick and Zolezzi,
2001). By the 1980s electricity infrastructure was “antiquated” (EIA, 1997
p64) and suffering from poor maintenance practices (EIA, 1997 p64).
Electricity utilities were described as inefficiently operated (EIA, 1997)
(Rudnick and Zolezzi, 2001). Pilfering was a major concern (EIA, 1997,
Rudnick and Zolezzi, 2001) and electricity prices high.

Since restructuring and privatisation of the industry, substantial foreign

investment has occurred in all sectors of the industry. Efficiency of operation
and reliability of supply has increased. However, the improvements in
reliability have not been uniform: For instance, in some regions of Argentina
lack of transmission investments has led to outages, and more than 90 % of
outages in the system have their origins in transmission problems (Abdala
and Chambottleyron, 1999).

Credit levels across the region vary significantly with ownership, local
economy performance, risk (especially those companies involved in
generation) and debt level. Table 40 summarises data from (Standard &
Poor's, 1999) relating to the credit worthiness of companies in 1998.

308
 Table 40 Credit rating of South American electricity companies from (Standard &
Poor's, 1999)
Credit rating
Ownership Functions (local/
Country Company Financial status
# ## foreign
currency)
Compania de
int. cov 3.1 x *
Transporte de Energia
Argentina P (25% G) T BBB- FFO to debt 15.2%
Electrica en Alta
**
Tension (Transener)
Eletrobras-Centrais
BB- moderate debt, govt
Brazil Electricas Brasileras G GT
B+ guaranteed
SA
Empresa Electrica del rel. high debt; high
Chile P GT BB-
Norte Grande SA risk generation
strong management
Empresa Nacional de & operations
Chile P GT A-
Electricidat (Endesa) Int. cov. 2.5 x *
FFO to debt 13%
Empresa de Energia BBB T&D parts of
Columbia P GTD
del Pacificon SA, ESP BBB- company low risk
Interconexion 76% G A- 9% return on assets;
Columbia T
Electrica SA ESP (in 1998) BBB- int. cov. 4 x
Electricidad de
Venezuela P ? B+
Caracas
# Ownership: G – government; P private
## Functions G – generation, T– transmission, D – distribution
FFO – funds from operations
Int. cov. – interest coverage from earnings, expressed as a multiplier

An ESMAP study identified an inadequate level of standardisation on

reliability levels as a problem in Argentina and several other South American
countries (ESMAP, 2001). In Argentina, Bolivia and Peru electricity quality
requirements have been introduced with penalties for non-compliance
(Rudnick and Zolezzi, 2001, ENRE, 2001). Pilfering has been “drastically”
reduced (Rudnick and Zolezzi, 2001).

For those countries with private investment system augmentation has been
proceeding rapidly. The length of medium and high-voltage lines in
Argentina, for instance, increased by 42% in 1991-1997(Abdala and
Chambottleyron, 1999).

In all Latin American countries transmission is considered a natural monopoly

(Fischer and Serra, 2000). Countries privatising their electricity industry have
generally adopted non-discriminatory access rules for transmission. In each
case the common grid is financed by all users, but the means by which this is
achieved varies from one country to the next. Although ownership of the
transmission grid varies from one country to the next, all countries in South

309
America grant concessions to private investors for the construction of new
lines, even when the main grid is publicly owned (Fischer and Serra, 2000).

In general transmission companies receive payments based on the

annualised replacement costs of “economically adapted” (optimised)
networks, together with an allowance for operations and maintenance
(Fischer and Serra, 2000). Argentina is an exception, where investment
costs are not remunerated. Often some form of incentive regulation is
applied to operation and maintenance costs.

Western Europe
Western Europe is characterised by highly meshed networks with varying
degrees of interconnection and transfer capability across state boundaries.
Performance standards are generally high. Electricity prices vary
significantly from one state to the next, and transmission prices also vary
significantly from country to country.

The EU Directive 96/92/EC which was passed in 1997 and which became
effective 19 February 199916 has been a catalyst for change in the electricity
markets of Western Europe. The EU Directive was designed to facilitate
market liberalisation by requiring non-discriminatory network access to all
users and increased transparency and separate accounting for generation,
transmission and distribution (Richer, 2000a). Under the Directive’s
schedule, 33% of demand, equating to customers with energy usage of 9
GWh or more, would be open to competition (choice of supplier) by 2003.

Competition and comparison of electricity charges within Europe is putting

downward pressure on tariffs and therefore revenues, with attendant
pressure to reduce costs. However, generally the revenue level of Western
European countries has been adequate or better, with reasonably supportive
regulation. Business parameters from a number of European enterprises are
shown in Table 41.

16
Member countries were given two years to implement the directive.

310
Where regulatory reform has occurred, in the UK, the Nordic region and
Germany the vertically integrated electricity industry has been split into
separate parts. Other countries have so far retained their vertically
integrated utilities.

Table 41 Enterprises in Western Europe – Business parameters at 1998 (Standard &

Poor's, 1999)
Owner- Func. Credit Debt Business Regulatory Revenue
Country Name
ship # ## rating level acumen support sufficiency
cost
Verbungesel high moderate
Austria 51% G GT AA- reduction
lschaft 70% 3 x int. cov.
program
4.7 x int.
sound
Electricité de cov.
France G GTD AA+ financial supportive
France “robust cash
profile
flow”
Italy * Enel G GTD A+ *
Spain Endesa,SA GTD A+ supportive comfortable
1.5-2.2 x
Malta Enemalta G GTD A
int.cov.
2.5 x int.
81% supportive, cov.;
Finland Fingrid Oyj 88% P T AA-
high flexible sufficient
revenue
weak profit
P but
high 1.8 x int.
Iceland Landsvirkjun govt GTD A+ high
augm. cov.
guar.
needs
5.8 x int.
cov.; 3 x
strong
National int. cov.
UK P T AA+ business
Grid Co. after
position
acquisition
of NEES
high
Public (from consistent reasonably 2-2.5 x int
Greece G GTD BBB
Power Corp. high perf. supportive .cov.
aug.)
moderat
Scottish & strong fin. some reg. 4.6 x int.
UK P GTD A+ e
Southern profit risk cov.
46%
revenue
Norway Statnett G T AAA 55%
cap
# Ownership: G – government; P private
## Functions G – generation, T– transmission, D – distribution
* Rating at August 2002 (Flintoff, 2002)
Int. cov. – interest coverage to earnings given as a multiplier

With some exceptions, electricity networks in Western Europe are mature,

growth is relatively low and augmentation is expensive and difficult to
achieve. Utilities in Western Europe have shown a strong interest in online
condition monitoring, and condition-based maintenance (Davies et al., 1998,
Jay and Williams, 2001, Jay, 2001). Aging of equipment is becoming an
important consideration.

311
In some parts of Europe environmental factors are becoming significant
drivers of asset management:
• There is a growing use of distributed generation, particularly wind-
powered and combined heat and power. This can change the use of
and loading patterns on the transmission network.
• In some countries there is a move to reduce the use of SF6 as an
electrical insulator, because of its greenhouse effects. A much greater
effort is being put into elimination of SF6 gas leaks and reclaiming SF6
to avoid environmental damage.

312
Bibliography

Abdala, M. A. and Chambottleyron, A. (1999) Transmission investment in

competitive power systems: decentralizing decisions in Argentina
Public Policy for the Private sector, Note No. 192 The World Bank
Group Accessed on 19/9/00 www.worldbank.org/html/fpd/notes/

ACCC (1999a) Draft statement of principles for the regulation of transmission

revenues CG 98/25, Australian Competition and Consumer
Commission, 27 May 1999

ACCC (1999b) Regulatory test for new interconnectors and network

augmentations 15 December 1999

ACCC (2000) Decision. NSW and ACT Transmission Network Revenue

Caps 1999/00-2003/04 CG98/118, Australian Competition and
Consumer Commission, 25 January 2000

ACCC (2002a) Decision. South Australian transmission network revenue cap

2003-2007/08 ACCC, 11 December 2002

ACCC (2002b) Transmission Network Service Provider (TNSP) Service

Standards ACCC, November 2002

ACCC (2003) Discussion Paper. 2003 Review of the Draft Statement of

Principles for the Regulation of Transmission Revenues Australian
Competition and Consumer Commission, 2003

Aesbasch, H., Mikes, E., Ponchon, P. and Gallon, F. (2000) 'Influence of

electrical arrangement and rated voltages on substation space
requirements and total costs for various gas- and hybrid-insulation
solutions' in CIGRE 2000 Session papers August 2000 Paris Paper
23-201

Aggarwal, R. K., Johns, A. T., Jayasinghe, J. A. S. B. and Su, W. (2000) 'An

overview of the condition monitoring of overhead lines' Electric Power
Systems Research, 53 1 15-22.

Albon, R. (2000) Incentive regulation benchmarking and utility performance.

Utility regulators forum discussion paper ACCC Accessed on 22 July
2003 www.accc.gov.au

Allan, D., Boyd, K. and Lang, P. (1992) 'Condition monitoring techniques for
high voltage insulation effectiveness and economics' in Electric
Energy Conference 1992. EECON '92. 'Driving Systems Harder'.
Preprints of Papers. 19-21 Oct. 1992 Brisbane, Qld., Australia 293-8

Allan, D. M. (1991) 'Transformer life assessment aids system planning'

Power International Edition., June 1991 39-40.

313
Allan, D. M. (1993) 'Practical life-assessment technique for aged transformer
insulation' IEE Proceedings A (Science, Measurement and
Technology), 140 5 404-8.

Allison, M. and Jay, P. (2001) Asset management at National Grid Company

UK, Semi-structured interview (Expert Interview), Recipient Crisp, J.,
12 October 2001

Allison, M. R., Lewis, K. G. and Winfield, M. L. (1995) 'An integrated

approach to reliability assessment, maintenance and life cycle costs in
the National Grid Company' in Second International Conference on
'The Reliability of Transmission and Distribution Equipment' 29-31
March 1995 Coventry, UK 180-5

Amato, F., Peelo, D. and Winkler, M. (1999) 'Life cycle cost management - A
modern instrument for investment decisions' in Electricity '99
Conference and Exposition "Technology for a changing industry".
March 29 to March 31 1999.Transactions. 29-31 March 1999
Vancouver, BC 1-5

Ametov, I., Kasymova, V., Borisovsky, V., Zyryanov, A., Petrov, G. and
Preigel, A. (1999) Status of the electricity pool in Central Asia and the
problems of efficient joint use of water and energy resources in the Syr
Darya Basin Task Order No. 813 Contract PCE-I-00-96-00002-00,
Central Asia Mission, U.S. Agency for International Development,
June 1999

Ammar, S., Wright, R. and Selden, S. (2000) 'Ranking state financial

management: A multilevel fuzzy rule-based system' Decision
Sciences, 31 2 449-481.

Anders, G., Endrenyi, J. and Yung, C. (2001) 'Risk-based planner for asset
management [of electric utilities]' IEEE Computer Applications in
Power, 14 4 20-26.

Anderson, D. (2002) Reducing the cost of preventive maintenance Accessed

on 21 November 2002 www.reliabilityweb.com

Anonymous (2002) Multisectoral initiative on potent industrial greenhouse

gases Accessed on 31/7/02 http://www.migiggs.org/mipiggstxt.html

Asia Development Bank (2000) Improving electricity supply and reliability in

Cambodia. News Release 138/00 Asian Development Bank Accessed
on 13 November 2002

Association of Local Gov. Engineering NZ and National Asset Management

Steering Group Aus. (2002) International infrastructure management
manual, Association of Local Government Engineering New Zealand,
National Asset Management Steering Group Australia, Thames, NZ.

314
Atanackovic, D., McGillis, D. T. and Galiana, F. D. (1999) 'Reliability
comparison of substation designs' Power Delivery, IEEE Transactions
on, 14 0885-8977 903-910.

Attar Software (1999) Xpertrule 3.9.3 Attar Software Ltd, UK

Audinet, P., Verneyre, F. (2002) Electricity in India. Providing power for the
millions., International Energy Agency, Paris.

Australasian CIGRE Asset Management WG (2001) Discussion paper on

asset management issues in transmission utilities in Australia and
New Zealand. An information paper on certain work in progress on
Asset Management issues by the Australasian CIGRE Asset
Management Working Group August 2001

Bak-Jensen, B. B., J.; Bjerregaard, C.G.; Jensen, P.R. (1999) 'Models for
probabilistic power transmission system reliability calculation' Power
Systems, IEEE Transactions on, 14 0885-8950 1166-1171.

Balzer, G., Brandl, M., Strnad, A., Rohsler, H. and Schnettler, A. (1998) 'A
computer aided, reliability-centered maintenance strategy for power
networks' Plant Engineer, 42 1 8-11.

Balzer, G., Scmitt, O., Halfmann, M., Neumann, C. and Strnad, A. (2000)
'Reliability centred maintenance of substation and overhead lines: a
tool for optimised management' in CEPSI 2000

Bartlett, S. (2000) 'Asset Management and Regulatory Drivers' in CIGRE

Technology Update Seminar. Asset Management November 2000
Brisbane, Australia

Basille, C., Aupied, J. and Sanchis, G. (1995) 'Application of RCM to High

Voltage Substations' in The reliability of transmission and distribution
equipment 29-31 March 1995 186-191

Batyayev, Y. V., Dementyev, Y. A., Dorf, G. A., Lavrentyev, V. M. and

Shleifman, I. L. (1994) 'Analysis of the equipment damages and
operation maintenance of classical type EHV and UHV substations
with an air insulation over 35-year operation' in Proceedings :
International Conference on Large High Voltage Electrical Systems 28
Aug.- 3 Sep. 1994 Paris, France Paper 23-105/1-5

BC Hydro (2002) BC Hydro Business Performance as at March 31,2002 BC

Hydro Accessed on 11 November 2002
www.bchydro.com/info/reports848.html

Beardow, M. (2003) 'Economics of asset management: Drawing it together' in

ESAA 2003 Residential School in Electrical Power Engineering
February 2003 Brisbane, Australia

Bell, M. (2002) 'Performance-Based Regulation: A View from the Other Side

of the Pond' The Electricity Journal, 15 1 66-73.

315
Bell, S. (2003) 'Asset Monitoring' in ESAA Study School on Asset
Management February 2003 University of Queensland, Brisbane,
Australia

Bergman, W. J. (1999) 'Reliability centered maintenance (RCM) applied to

electrical switchgear' in 1999 IEEE Power Engineering Society
Summer Meeting Conference Proceedings. Edmonton, Alta., Canada
1164-7

Bergman, W. J. (2001) 'Selecting Substation Monitoring' in 2001 IEEE/PES

Transmission and Distribution Conference October 28 - November 2
2001 Atlanta, Georgia, USA

Bernstein, B. S. and Brancato, E. L. (1993) 'Aging of Equipment in Electric

Utilities' IEEE Transactions on Electrical Insulation, 28 5 866-875.

Bower, A. (2001) Asset Management at YEG, Semi-structured interview

(Expert interview), Recipient Crisp, J., 16 October 2001

Brown, G. (2001a) Powerlink Queensland. Review of network service

standards. Prepared for Australian Competition and Consumer
Commission PB Associates, 20 April 2001

Brown, R. E., Gupta, S., Christie, R. D., Venkata, S. S. and Fletcher, R.

(1996) 'Distribution system reliability assessment using hierarchical
Markov modeling' Power Delivery, IEEE Transactions on, 11 0885-
8977 1929-1934.

Brown, R. E. H., A.P.; Willis, H.L.; Luedtke, F.A.; Born, M.F. (2001b)
'Assessing the reliability of distribution systems' IEEE Computer
Applications in Power, 14 0895-0156 44-49.

Buckner, W. F. (2002) 'The electricity supply industry and its impact on

transmission line technology - economical aspects' CIGRE 2002.
Session Papers, Paper 22-204.

Butler, G. B. (2000) 'Federal incentives are needed for transmission and

distribution' The Electricity Journal, 13 7 78-79.

Camfield, R. J. and Schuster, A. G. (2000) 'Pricing transmission services

efficiently' The Electricity Journal, 13 9 13-32.

Chan, T. M., Heil, F., Kopejtkova, D., O'Connell, P. O., Taillebois, J.-P. and
Welch, I. (1998) 'Report on the second international survey on high
voltage gas insulated substations (GIS) service experience' in
International conference on large high voltage electric systems 30 Aug
- 2 Sep 1998 Paris, France Paper 23-102

CIER, ESMAP and World Bank (2001) Regional electricity markets in South
America. Phase 1. Identification of issues for the development of
regional power markets 01601, CIER, ESMAP World Bank, December
2001

316
CIGRE AP13 (2001) Circuit Breaker Replacement: Transmission company
practices Australia and New Zealand Unpublished discussion paper,
August 2001

CIGRE JWG 23/39.14 (2002) 'Maintenance outsourcing guidelines' Electra,

201 37-47.

CIGRE SC 23 and 39 (2000) An international survey of maintenance policies

and trends Brochure 152, CIGRE, February 2000

CIGRE WG 12.05 (1983) 'An international survey on failures in large power

transformers in service' Electra, 88 21-48.

CIGRE WG 12.09 (1993) 'Lifetime evaluation of transformers' Electra, 150

39-51.

CIGRE WG 13-09 (2000) User Guide for the Application of monitoring and
diagnostic techniques for switching equipment for rated voltages of
72.5kV and above. Brochure No. 167, CIGRE, August 2000

CIGRE WG 37.27 (2000) 'Ageing of the system - impact on planning' in

CIGRE 2000. Session papers August 2000 Paris Paper 22-303

CIGRE WG B2.13 (2002) 'Management of existing overhead transmission

lines' Electra, 205 39-49.

Clippinger, C. K. and McDermott, B. A. (2001) 'Substation integration utilising

microprocessor protection' in 2001 Western Power Delivery
Automation Conference Pullmar 1-31

Cope, D. (Ed.) (2000) Energy sector deregulation in the Apec region, Asia
Pacific Energy Research Centre, Tokyo, Japan.

Coyle, R. G. (1977) Management Systems Dynamics, John Wiley & Sons,

London.

Crisp, J. (2001) 'Asset Management Drivers for Transmission Businesses,

and their Impact on the System' in Distribution 2001. 6th International
Transmission and Distribution Conference and Exhibition 11-14
November 2001 Brisbane, Australia

Dalton, J. G., III, Garrison, D. L. and Fallon, C. M. (1996) 'Value-based

reliability transmission planning' IEEE Transactions on Power
Systems, 11 3 1400-8.

Damstra, R. D., Baylis, J. A., Hoekstra, H. E., Aupied, J. A., Greeneklee, P.,
Tremblay, P., Simoens, L. and Allison, M. R. (2000) 'A review of
methods for the management of transmission outage risk.' in CIGRE
2000. Session papers. August 2000 Paris Paper 23/39-07

Davidson, J. (Ed.) (1988) The reliability of mechanical systems, Institution of

mechanical engineers, London.

317
Davies, T. C., Wester, P., Renaud, P., Slofstra, J. W., Prieto, J., Sundklakk,
H., Baylis, J. A. and Allison, M. R. (1998) 'Drivers for change in
transmission maintenance policy and the role of computer based
information and decision support systems' in Proceedings -
International conference on large high voltage systems August 1998
Paris, France Paper 23/37/39-02

de Parny, R. (1993) 'Helicopter airborne services live line work' in

Transmission and Distribution Construction and Live Line
Maintenance, 1993. Proceedings from ESMO-93., Sixth International
Conference on 74-87

Dekker, R. (1996) 'Applications of maintenance optimisation models: a

review and analysis' Reliability Engineering and System Safety, 51 3
229-240.

Dekker, R. and Scarf, P. A. (1998) 'On the impact of optimisation models in

maintenance decision making: the state of the art' Reliability
Engineering & System Safety, 60 2 111-119.

Dialynas, E. N., Koskolos, N. C. and Agoris, D. (1996) 'Reliability assessment

of autonomous power systems incorporating HVDC interconnection
links' Power Delivery, IEEE Transactions on, 11 0885-8977 519-525.

Dismukes, D., Cope, R. F. I. and Mesyanzhinov, D. (1998) 'Capacity and

economies of scale in electric power transmission' Utilities Policy, 7
155-162.

Dobozi, I. and Pohl, G. (1995) Real output decline in transition economies --

Forget GDP, Try power consumption The World Bank Group.
Transition Newsletter. Accessed on 19/03/2002
http://www.worldbank.org/html/prddr/trans/janfeb95/pg17-18.htm

Duke Energy (2002) Westcoast Energy and Union Gas corporate credit
ratings lowered by Standard and Poor's Duke Energy Accessed on 20
Sep. 2002 www.duke-
energy.com/decorp/content/newscenter/newsreleases/2002

Dwivedi, P. K., Khurana, A. M. and Kumar, A. (1988) 'Gas insulated

switchgear v/s conventional air insulated switchgear techno
economical considerations' in SWICON 88: 2nd International Seminar
on Switchgear and Controlgear 5/59-69

ECAR Electric Equipment Panel (1999) How the aging of major equipment
impacts reliability 99-EEP-61, ECAR, May 1999

EIA (1996) The changing structure of the Electric Power Industry: An update
DOE/EIA- 0562(96), Energy Information Administration, U.S.
Department of Energy, December 1996

318
EIA (1997) Electricity Reform Abroad and U.S. Investment Energy
Information Administration, U.S. Department of Energy, September
1997

EIA (2000) The restructuring of the electric power industry. A capsule of

issues and events DOE/EIA-X037, Energy Information Administration,
U.S. Dept of Energy, January 2000

EIA (2001a) Country Analysis Briefs Energy Information Administration,

Department of Energy Accessed on 2/7/02
www.eia.doe.gov/emeu/cabs/

EIA (2001b) Country Analysis Briefs. Japan Energy Information

Administration, Department of Energy Accessed on 2/7/02
www.eia.doe.gov/emeu/cabs/Japan.html

EIA (2001c) Country Analysis Briefs. Malaysia Energy Information

Administration, Department of Energy Accessed on 2/7/02
www.eia.doe.gov/emeu/cabs/Malaysia.html

EIA (2001d) Country Analysis Briefs. South Korea Energy Information

Administration, Department of Energy Accessed on 2/7/02
www.eia.doe.gov/emeu/cabs/skorea.html

EIA (2002a) Country Analysis Briefs. India Energy Information

Administration, Department of Energy Accessed on 2/7/02
www.eia.doe.gov/emeu/cabs/india.html

EIA (2002b) Country Analysis Briefs. Indonesia Energy Information

Administration, Department of Energy Accessed on 2/7/02
www.eia.doe.gov/emeu/cabs/indonesia.html

EIA (2002c) Country Analysis Briefs. Philippines Energy Information

Administration. Department of Energy, USA Accessed on 2 July 2002
www.eia.doe.gov/emeu/cabs/philippines.html

EIA (2002d) International Energy Outlook 2002 EIA, Dept of Energy, USA
Accessed on 7 April 2003 www.iea.doe.gov/oiaf/eio/pdf/electricity.pfd

Endrenyi, J., Aboresheid, S., Allan, R. N., Anders, G., Asgarpoor, S., Billin ,
Chowdhury, N., Dialynas, E. N., Fipper, M., Fletcher, R. H., Grigg, C.,
Mc Calley, J., Meliopoulos, s., Mielnik, T. C., Nitu, P., Rau, N.,
Reppen, N. D., Salvaderi, L., Schneider, A. and Singh, C. (2001) 'The
present status of maintenance strategies and the impact of
maintenance on reliability' IEEE transactions on power systems, 16 4
638-646.

Endrenyi, J. and Wellssow, W. H. (2001) 'Power system reliability in terms of

the system's operating states' in 2001 IEEE Porto Power Tech
Conference 10-13 September 2001 Porto, Portugal 386-391

319
Energy Policy Group, M. o. C. (1995) Light-Handed Regulation of New
Zealand's Electricity and Gas Industries Energy Policy Group Ministry
of Commerce, Accessed on 19/09/2000 www.med.govt.nz/ers/electric

ENRE (2001) Informe Anual 2000 ENRE,

Eriksson, A., Pettersson, K. G., Krenicky, A., Baker, R., Ochoa, J. R. and
Leibold, A. (1994) 'Experience with gas insulated substations in the
USA' in 1994 IEEE Power Engineering Society Transmission and
Distribution Conference (Conf No. 94CH3428-0) 369-75

ESMAP (2001) Regional electricity market interconnections- Phase 1:

Identification of issues for the development of regional power markets
in South America 016/01, Energy Sector Management Assistance
Programme (ESMAP), Dec 2001

Fischer, R. and Serra, P. (2000) 'Regulating the electricity sector in Latin

America' Economia, Fall 2000 155-218.

Flatt, R. N. (2000) Through-Life Cost (TLC) management: The key to

reducing OPEX Accessed on 2 July 2000 www.poweronline.com

Flintoff, A. (2001) International utility ratings and ratios Standard and Poor's
Accessed on 20/9/02
www.standardandpoors.com/Forum/Ratings/Commentaries/Corporate
Finance

Flintoff, A. (2002) Industry report card: European electricity utilities Standard

& Poor's Accessed on 20/9/02
www.standardandpoors.com/europe/deutch/content/020807reportcard
1032255977.htm

Forrester, J. W. (1961) Industrial Dynamics, MIT Press, Cambridge,

Massechusetts.

Giles, R. L. (1970) Layout of EHV substations, Cambridge University Press,

Cambridge, UK.

Goel, L. (2000) 'Monte Carlo simulation-based reliability studies of a

distribution test system' Electric Power Systems Research, 54 1 55-
65.

Green, M., Waddell, R., McCarthy, P., Kumble, C., Southwell, P. and
Manglick, A. (2002) 'Trends and approaches to planning standards,
service standards and new investment in a deregulated market -
Australian and New Zealand experience.' in CIGRE Session 39. Set of
papers presented to the CIGRE 2002 session 25-30 August 2002
Paris Paper 37-302

Greer, I. (2001) Rating methodology for global power utilities Standard &
Poors Accessed on 20/9/02

320
 www.standardandpoors.com/australiaNZ/resourcecentre/methodology
/corporateinfrastructure

Hammons, T., Taher, F., Voropai, N. I., Belyaev, L. S., Podkovalnikov, S. V.,
Shutov, G. V., Povh, D., Woodford, A., Vallee, G. J., Doucet, G. J.,
Meisen, P. and Weimers, L. (1998) 'International high-voltage grids
and environmental implications' IEEE Power Engineering Review,
August 1998 1-42.

Hastings, N. and Crisp, J. (2001) Powerlink Circuit Breakers Report

Unpublished report, QUT, January 2001

Hastings, N. A., Sharp, B. and Thompson, P. (1999) 'Optimised Replacement

Policy for Assets in Electricity Transmission Networks' in 5th
International Transmission & Distribution Conference and Exhibition.
Conference Papers. Distribution 2000. November 1999 Brisbane,
Australia

Hastings, N. A. J. (2000) Asset management and maintenance, Queensland

University of Technology.

Hauret, C. (2001) Despite strength, credit quality for U.S. Electric Utilities is
trending negatively Standard & Poor's Accessed on 20/9/2002
www.standardandpoors.com/Forum/RatingsCommentaries/Corporate
Finance

Heidari, G. O. and Heidari, M. (2002) 'The effect of land cost on transmission

line design' CIGRE 2002. Session Papers, Paper 22-301.

High Performance Systems Inc (2002) Stella Research 7.03 High

Performance Systems Inc.

Hills, J. (1997) Transmission/Distribution electric eq. Tradeport trade

directory, U.S. & foreign commercial service and U.S Department of
State Accessed on 30 September 2002

Hilmer, F. (1993) National Competition Policy, Australian Government

Publishing Service, Canberra.

Hirst, E. (2000) 'Do we need more transmission capacity?' The Electricity

Journal, 13 9 78-89.

Hogan, W. (1998) 'Getting the prices right in PJM: What the data teaches us'
The Electricity Journal, 11 7 (August / September) 61.

Hoskins, R. P., Brint, A. T. and Strbac, G. (1997) 'The use of condition

information and the physical processes of failure as an aid to asset
management' in 32nd Universities Power Engineering Conference
UPEC '97 vol.2. 10-12 Sept. 1997 Manchester, UK 783-6

321
Hoskins, R. P., Brint, A. T. and Strbac, G. (1998) 'A structured approach to
Asset Management within the electricity industry' Utilities Policy, 7 4
221-232.

Hoskins, R. P., Strbac, G. and Brint, A. T. (1999) 'Modelling the degradation

of condition indices' IEE Proceedings. Generation,Transmission and
Distribution, 146 4 386-392.

Huz, S., Anderson, D. I., Richardson, G. P. and Boothroyd, R. (1997) 'A

framework for evaluating system thinking interventions: an
experimental approach to mental health system change' System
Dynamics Review, 13 2 149-169.

Hydro Quebec Transenergie (2002) Hydro Quebec Transenergie. Report of

activities 2001 Hydro Quebec,

ICEM A/P (1997) Country Report - Malaysia. Energy industry and trade
union strategy in Asia ICEM A/P Accessed on 30 September 2002

IEA (1999) Energy Policies of IEA Countries. Japan 1999 Review

International Energy Administration, 1999

IEEE (2000) IEEE Guide for Selection of Monitoring for Circuit Breakers IEEE
Std C37.10.10-2000, IEEE, March 2000

IEEE Transformers Committee (1991) IEEE Guide for Loading Mineral-Oil-

Immersed Power Transformers rated in excess of 100MVA (65 Deg C
Winding Rise) C57.115-1991, IEEE (sponsored by Transformers
Committee of the IEEE Power Engineering Society), 21 March 1991

Izaguirre, A. K. (1998) Private Participation in the Electricity Sector - Recent

Trends Public Policy for the Private sector Note No. 154, The World
Bank Group Accessed on 19/9/2000
www.worldbank.org/fpd/notes/noteslist/html

Janssen, A. L. J., Brunke, J. H., Heising, C. R. and Lanz, W. (1996) 'CIGRE

WG 13.06 studies on the reliability of single pressure SF/sub 6/-gas
high-voltage circuit-breakers' IEEE Transactions on Power Delivery,
11 1 274-82.

Janssen, A. L. J., Degen, W., Heising, C. R., Bruvik, H., Colombo, E., Lanz,
W., Fletcher, P. and Sanchis, G. (1995) 'A summary of the final results
and conclusions of the second international enquiry on the reliability of
high voltage circuit breakers' in Proceedings of the 35th Session.
International Conference on Large High Voltage Electric Systems.
Aug.-3 Sept. 1994 Paris France Paper 13-202

Janssen, A. L. J., Degen, W., Tudrej, M. and Ikeda, S. (1993) 'Application of

diagnostic techniques for high voltage circuit-breakers' in CIGRE
Proceedings of the 34th Session 1992 Paris, France Paper 13-101

322
Jasamb, T. and Pollitt, M. (2001) 'Benchmarking and regulation: international
electricity experience' Utilities Policy, 9 107-130.

Jay, P. (2001) Operations and Maintenance Philosophy and Strategy.

Presentation made to Saudi Electric Company. National Grid
Company

Jay, P. and Williams, A. (2001) 'Understanding and managing assets in

National Grid' in Engineering Asset Management 2001 2-3 October
2001 London, UK

Jefferies, D. (1997) 'Developing the highways of power' Power Engineering

Journal, IEE, February 1997 5-9.

Jones, P. A. (1997) 'Aligning decision making processes with an asset

management mission' in Engineering Asset Management for Utilities,
Industry and Commerce 7-8 October 1997 London, UK 1.2.1-13

Jones, S., Gardner, K. and Brennan, G. (2000) 'Switchgear issues in

deregulated electricity industries in Australia and New Zealand' in
CIGRE Session 2000. International Conference on Large High
Voltage Electric Systems August 2000 Paris

Karlsson, D., Wallin, L., Olovsson, H.-E. and Solver, C.-E. (1997) 'Reliability
and life cycle cost estimates of 400 kV substation layouts' Power
Delivery, IEEE Transactions on, 12 4 1486-1492.

Karlsson, D. N., P. (2001) 'Earthing of 130 kV power lines-effects on

dependability and line cost' in Transmission and Distribution
Conference and Exposition, 2001 IEEE/PES 609-614 vol.2

Khatib, H. (1998) 'Electrical power in developing countries' Power

Engineering Journal, 12 5 239-247.

Kirby, B., Hirst, E (1999) 'Maintaining transmission adequacy in the future'

The Electricity Journal, 12 9 62-72.

Klir, G. J. and Yuan, B. (Eds.) (1996) Fuzzy sets, fuzzy logic and fuzzy
systems. Selected papers by Lofti A. Zadeh, World Scientific,
Singapore.

KLSE (2002) Tenaga Nasional Berhard (TNB) Kuala Lumpa Stock Exchange
Accessed on 30/9/2002 www.klse.com.my/website/listing/lc/tnb.htm

Knobloch, H., Marin , H., Schramm, H. H., Kerchesch, P., Schiemann, A., de
Hesselle, M., Moller, K., von Starck, R. and Aachen, R. W. T. H.
(2000) 'Technological trends in high-voltage circuit -breakers' in
CIGRE Session 2000. International Conference on Large High Voltage
Systems August 2000 Paris Paper 13-106

323
Kogan, V. I., Fleeman, J. A., Provanzana, J. H. and Shih, C. H. (1988)
'Failure analysis of EHV transformers' IEEE Trans. on Power Delivery,
3 2 672-83.

Kumar, P., Bahl, R. and Vyas, H. (2002) 'Use of new maintenance

techniques for increasing availability of overhead transmission lines -
experiments conducted in India' in CIGRE Session 39. Set of papers
presented to the CIGRE 2002 session. 25-30 August 2002 Paris
Paper 22-106

Kurihara, I. (1999) 'Deregulation of electric power industry in Japan and

present practice of securing reliability in transmission planning stage'
in Power Engineering Society Summer Meeting, 1999. Conference
Proceedings. Edmonton, Canada 143-146

Lalor, R. P. and Garcia, H. (1996) Reshaping Power Markets- Lessons from

Chile and Argentina Pubic Policy for the Private Sector, Note No. 85,
World Bank Accessed on 19/9/2000
www.worldbank.org/fpd/notes/noteslist/html

Lamsoul, J., Rogier, J., Couneson, P. and Van Overmeere, A. (2000)

'Belgian experience on initiatives to improve the capability of existing
overhead lines' in CIGRE 2000. Session papers. August 2000 Paris
Paper 22-206

Lane, D. C. (2000) 'Diagramming conventions in system dynamics' Journal of

the Operational Research Society, 51 241-245.

Larsen, E. R. and Bunn, D. W. (1999) 'Deregulation in electricity:

understanding strategic and regulatory risk' Journal of the Operational
Research Society, 50 337-344.

Laycock, W. J., Gardiner, I. P., McDonald, J. R. and Booth, C. (1998) 'The

effect of deregulation on the performance of protection systems' in
International conference on large high voltage electric systems August
1998 Paris Paper 34-204

Lee, C. (1997) 'RCM approach to ageing insulators' in Proceedings of the

electricity transmission and distribution RCM symposium August 18-9
1997 Brisbane Australia 77-80

Lee, P. (2001) Asset management at San Diego Gas and Electric, Semi-
structured interview: (Expert Interview), Recipient Crisp, J., 30 October
2001

Lee, S. T. and Hoffman, S. (2001) 'Power delivery reliability initiative bears

fruit' IEEE Computer Applications in Power, 14 0895-0156 56-63.

Leonhard, W. and Muller, K. (2002) 'Balancing fluctuating wind energy with

fossil power stations. Where are the limits?' Electra, 204 12-18.

324
Lovei, L. (1998) Electricity Reform in Ukraine: The impact of weak
governance and budget crises Public Policy for the Private sector Note
No. 168, World Bank Group Accessed on 19/9/00
www.worldbank.org/html/fpd/notes/

Luger, G. F. and Stubblefield, W. A. (1999) Artificial Intelligence: Structures

and Strategies for Complex Problem Solving, Addison Wesley
Longman Inc, Reading Ma, USA.

Marshall, C. E. and Visser, H. H. (1998) 'Live work in Eskom Transmission' in

Transmission & Distribution Construction, Operation & Live-Line
Maintenance Proceedings, 1998. ESMO '98. 1998 IEEE 8th
International Conference on 11-18

Marshall, W. (1997) 'Software predicts future maintenance cost.'

Transmission & Distribution World, March 66-78.

Mazza, G. and Michaca, R. (1981) 'The first international enquiry on circuit-

breaker failures and defects in service' Electra, 79 December 1981 21-
91.

Mc Taggart, D., Findlay, C. and Parkin, M. (1996) Economics, Addison

Wesley, Sydney.

McMahon, B. (1995) 'Reliability and maintenance practices for Australian and

New Zealand HV transmission lines' in Second International
Conference on 'The Reliability of Transmission and Distribution
Equipment' (Conf. Publ. No.406) 29-31 March 1995 Coventry, UK
198-203

Meeuwsen, J. J. and Kling, W. L. (1997a) 'Effects of preventive maintenance

on circuit breakers and protection systems upon substation reliability'
Electric Power Systems Research, 40 3 181-8.

Meeuwsen, J. J. and Kling, W. L. (1997b) 'Substation reliability evaluation

including switching actions with redundant components' Power
Delivery, IEEE Transactions on, 12 0885-8977 1472-1479.

Microsoft (2001) Encarta Interactive World Atlas 10.00.00.0808 Microsoft

Corporation

Ministry of Economic Development, N. (2000) Chronology of New Zealand

Electricity Reform New Zealand Government - Ministry of Economic
Development Accessed on 19/09/2000
www.med.govt.nz/ers/electric/chronology/index.html

Ministry of Energy and Utilities, N. (1999) 1998-99 NSW electricity network

management report Ministry of Energy and Utilities, NSW
Government, November 1999

325
MIT Systems Dynamics Group (1990) Systems Dynamics Education Project
MIT, Sloan Business School Accessed on 4 August 2003
sysdyn.clexchange.org

Mittard, P., Bachiller, J. A. and Salamanca, F. (2002) 'Spanish solutions on

substation compaction' in CIGRE session 39. Set of papers presented
to the CIGRE 2002 session 25-30 August 2002 Paris Paper 23-203

Mock, K. and Bodek, D. (2001) Challenges mount for public power in the
U.S. Standard & Poor's Accessed on 20/9/2002
www.standardandpoors.com/Forum/RatingsCommentaries/Corporate
Finance

Mohd Noor, M. J. (2002) Transmission in Malaysia, Semi-structured interview

(Expert Interview), Recipient Crisp, J., 1 October 2002

Moubray, J. (1992) Reliability-Centered Maintenance, Industrial Press Inc.,

New York.

Murray, K. (2000) 'Regulatory Issues. NSW Experience.' in CIGRE

Technology Update Seminar. Asset Management November 2000
Brisbane, Australia

Nakajima, S. (1986) 'TPM--Challenge to the improvement of productivity by

small group activities' Maintenance Management International, 6 73-
83.

Nasser, T. (1999) 'The hidden value of transmission assets' The Electricity

Journal, 69-78.

NEMMCO (1999) An Introduction to Australia's National Electricity Market

NEMMCO Accessed on 14/02/2001 www.nemmco.com.au

Newbery, D. M. (1994) 'Competition and regulation in the electricity industry'

in OECD/World Bank Conference on Competition and Regulation of
Network Infrastructure Industries June 28-July 1 1994 Budapest

NGC (2000) Report to the Director General of Gas & Electricity Markets.
Transmission system performance. Financial Year 99-2000. National
Grid Company,

Nolasco, J. F. and Ferreira da Silva, J. B. G. (2002) 'Assessment and

improvement in availability of overhead lines and components' in
CIGRE Session 39. Set of papers presented to the CIGRE 2002
session 25-30 August 2002 Paris Paper 22-107

Nordin, R. S. B., Hoff, P. E. H. G. and Lord, W. S. G. (1995) 'Experiences

with diagnostic methods and condition monitoring for HV circuit
breakers and GIS in Sweden' in Stockholm Power Tech International
Symposium on Electric Power Engineering 18-22 June 1995
Stockholm, Sweden 130-5

326
North American Reliability Council (1994) Regional Reliability Criteria.
Summary of regional criteria and guides for planning reliable bulk
electric systems North American Reliability Council, November 1994

Office of the Tasmanian Regulator (2001) Electricity supply industry

performance report 2000 -2001 Office of the Tasmanian Energy
Regulator, December 2001

OFGEM (2001) Report on distribution and transmission performance

1999/2000 Office of Gas and Electricity Markets, January 2001

Ostendorp, M. (2000) 'Innovative airborne inventory and inspection

technology for electric power line condition assessments and defect
reporting' in Transmission and Distribution Construction, Operation
and Live-Line Maintenance Proceedings. 2000 IEEE ESMO - 2000
IEEE 9th International Conference on 123-128

Owens, D. (2001) United States of America Department of Energy National

Transmission Grid Study. Comments of the Edison Electric Institute 12
October 2001

Pelacchi, P. (1998) 'Automatic hot line insulator washing device positioned by

helicopter' in Transmission & Distribution Construction, Operation &
Live-Line Maintenance Proceedings, 1998. ESMO '98. 1998 IEEE 8th
International Conference on 133

Percebois, J. and Wright, P. (2001) 'Electricity consumers under the state

and the private sector: comparing the price performance of the French
and UK electricity industries 1990-2000' Utilities Policy, 10 3-4 167-
179.

Perez-Arriaga, I., Rubio Oderiz, F. J. and Peran Montero, F. (2002)

Benchmark of Electricity Transmission Tariffs. Final Report. Prepared
for DG TREN / European Commission Instituto de Investigacion
Tecnologica, Universidad Pontificia Comillas, February 2002

Pettersson, P., Wahlstrom, B., Eriksson, R. and Pettersson, K. (1986) 'Life

cycle cost evaluation for gas-insulated substations including a method
to determine outage costs' in International Conference on Large High
Voltage Electric Systems CIGRE: Proceedings of the 31st Session.
Paper 23.08/1-11

Pidd, M. (1996) Tools for thinking. Modelling in Management Science, John

Wiley & Sons, New York.

Pidd, M. (1998) Computer simulation in management science. 4th ed., John

Wiley & Sons, New York.

Plantagie, J. W. (2000) Infrastructure Finance. Analysis ESKOM Standard &

Poor's, 21 September 2000

327
Power Grid Corporation (2001) Annual Report 2000 - 2001 Power Grid
Corporation of India,

Powerlink Queensland (1999) Annual Report 1998-99 Powerlink

Queensland, 1999

Powerlink Queensland (2000) Forecasting capital expenditure for the

Queensland transmission grid - an approach to deal with a rapidly
changing power system October 2000

Powerlink Queensland (2001) Annual report. 00/2001 Powerlink Queensland,

2001

Productivity Commission (2002) Financial Performance of Government

trading enterprises 1996-97 to 2000-01, Performance Monitoring,
AusInfo, Canberra.

Pryor, B. M., Gibson, D., Blakeley, R. and Finn, J. (1998) 'Life cycle
management philosophies for transmission substations' in CIGRE
Session 1998. August 1998 Paris Paper 23-109

Queensland Government (2000) Queensland energy policy. A cleaner

greener state Queensland Government, May 2000

Read, E. G. (1997) 'Transmission pricing in New Zealand' Utilities Policy, 6 3

227-235.

Reliability Criteria Subcommittee of the Engineering Committee, N. (1985)

Reliability concepts in bulk power electric systems North American
Electric Reliability Council, February 1985

Render, B. and Stair, R. M. J. (1994) Quantitative Analysis for Management.

Fifth Ed., Allyn and Bacon, Boston.

Richer, C. E. (2000a) EU Directive update: the pace quickens Standard &

Poor's Accessed on 20/9/02
www.standardandpoors.com/Forum/RatingsCommentaries/Insurance/
Articles

Richer, C. E. (2000b) International utility comparative financial statistics for

1999 Standard & Poor's Accessed on 20/9/02
www.standardandpoors.com/Forum/RatingsCommentaries/Corporate
Finance

Richmond, B. (2001) An introduction to systems thinking, High Performance

Systems, Hanover,New Hampshire.

Rudnick, H. and Zolezzi, J. (2001) 'Electric Sector deregulation and

restructuring in Latin America: Lessons to be learnt' IEE Proceedings
Generation, Transmission and Distribution, 148 2 180-184.

328
Saeed, K. (1998) 'Sustainable trade relations in a global economy' System
Dynamics Review, 14 2-3 107-128.

SAIIR (2001) 2nd Annual performance report. Performance of regulated

electricity businesses in South Australia 2000 - 2001. Office of the
South Australian Independent Industry Regulator, November 2001

Scarf, P. A. (1997) 'On the application of mathematical models in

maintenance' European Journal of Operational Research, 99 493-506.

Senge, P. M., Kleiner, A., Roberts, C., Ross, R. B. and Smith, B. J. (1994)
The fifth discipline fieldbook. Strategies and tools for building a
learning organisation, Nicholas Brealey Publishing, London.

Shameen, A. (2001) 'Asia's big power play' Asia Week 27 2 19 January 2001

Sherard, T. (2001) Asset management at Duke Energy - Transmission, Semi-

structured interview (Expert interview), Recipient Crisp, J., 31 October
2001

Simpson, R. (2001) Powerlink Operating and Maintenance Expenditure

Review. Prepared for Australian Competition and Consumer
Commission. PB Associates, 10 April 2001

Singh, A. and Poh, S. K. C. (1996) 'A tussle over power profits' Asia Week 7
December 1996

Singh, A. and Poh, S. K. C. (1997) 'Hot wires and cold cash. Money
problems hobble Tenaga's expansion' Asia Week 17 January 1997

Sioshansi, F. and Morgan, C. (1999) 'Where function follows form -

international comparison of restructured electricity markets' The
Eectricity Journal, 20-30.

Sioshansi, F. P. (2001) 'Opportunities and perils of the newly liberalized

European electricity markets' Energy Policy, 29 419-427.

Soler, R. G., J. (1993) 'Robotic maintenance of the EDF transmission (63 to

400 kV) network: feasibility study and effects on tower design VO -' in
Transmission and Distribution Construction and Live Line
Maintenance, 1993. Proceedings from ESMO-93., Sixth International
Conference on 459-468

Standard & Poor's (1999) Electric Utilities Standard & Poor's, September
1999

Standard & Poor's (2001) Project and Infrastructure Finance Review 2001.
Criteria and commentary Standard & Poor's, October 2001

Standard & Poor's (2002) Asia Ratings List Standard & Poor's Accessed on
30/9/2002

329
 www.standardandpoors.com/ratingsActions/RatingsLists/asianratings/
printmalaysia.htm

Standards Australia (1999a) AS/NZS 4360 Risk Management, Standards

Association of Australia, Strathfield, NSW.

Standards Australia (1999b) AS/NZS 4536:1999 Life cycle costing - An

application guide, Standards Australia and Standards New Zealand,
Homebush NSW, Australia.

Stern, J. and Holder, S. (1998) 'Regulatory governance: criteria for assessing

the performance of regulatory systems - An application to
infrastructure industries in the developing countries of Asia' Utilities
Policy, 8 1 33-50.

Stewart, S. A. (2001) Asset management in Scottish Power, Semi-structured

interview (Expert interview), Recipient Crisp, J., 19 October 2001

Sudhir, V., Srinivasan, G. and Muraleeharan, V. R. (1997) 'Planning for

sustainable solid waste management in urban India' System Dynamics
Review, 13 3 223-246.

Sullivan, M., Noland-Suddeth, B., Vardell, T. and Vojdani, A. (1996)

'Interruption costs, customer satisfaction and expectations for service
reliability' IEEE Trans. on Power Systems, 11 2 989-95.

Tanaka, T., Okamoto, T., Nakanishi, K. and Miyamoto, T. (1993) 'Aging and
related phenomena in modern electric power systems' IEEE Trans. on
Electrical Insulation, 28 5 826-44.

Teixiera, J. (2001) Interview - National Grid USA, Semi-structured interview

(Expert Interview), Recipient Crisp, J., October 31 2001

Theunissen, J., Klebanowski, A., Ooi, K., Bennett, K. and Rednell, D. (1998)
'Management of protection & local control practices within a privatised
electricity supply industry - Transmission utility perspective' in
International conference on Large High voltage Electric Systems 30
August - 5 September 1998 Paris, France Paper 34-201

Trade Partners UK (2002) Power market in Malaysia Trade Partners UK

Accessed on 30/9/2002 www.tradepartners.gov.uk

UNESCO (1999) Part 1. Developments in Electric Power in the ESCAP

region United Nations, 1999

Urwin, R. (1999) 'Engineering Challenges in a competitive electricity market'

in High voltage engineering Symposium 22-27 August 1999 366-378

Valdez Flores, C. and Feldman, R. M. (1989) 'A survey of preventive

maintenance models for stochastically deteriorating single unit
systems' Naval Research Logistics Quarterly, 36 419-446.

330
Vermeulen, S. T. J. A., Rijanto, H. and van der Duyn Schouten, F. A. (1998)
'Modelling the influence of preventive maintenance on protection
system reliability performance' Power Delivery, IEEE Transactions on,
13 0885-8977 1027-1032.

von Hirschhausen, C. and Opitz, P. (2001) DIW Discussion papers.

Discussion paper no. 246. Power Utility re-regulation in East European
and CIS transformation countries (1990-1999). An institutional
interpretation. Deutsches Institut fur Wirtschaftsforschung Accessed
on 19/9/02 www.diw.de

Wang, H. (2002) 'A survey of maintenance policies of deteriorating systems'

European Journal of Operational Research, 139 3 469-489.

Wilson, M. (1998) Reforming the Russian electricity sector Public Policy for
the Private Sector. Note No.139. World Bank Group Accessed on 20
September 2000 www.worldbank.org/viewpoint/index.asp

Woolweber, B. N. (1998) 'Graded approach to substation preventive

maintenance activities' in Proceedings of the American Power
Conference. Reliability and Economy - Technology Focus for
Competition and Globalization. 14-16 April 1998 Chicago, Il,USA
p.470-4

World Bank (1999a) Kazakhstan - Electricity Transmission .. Rehabilitation

World Bank, Public Information Document, Project KZPE65414
Accessed on 20/09/00 www.worldbank.org/pics/pid/kz65414.txt

World Bank (1999b) World Development Indicators World Bank Accessed on

20/9/2000 www.worldbank.org/data/databytopic

World Bank (2000a) World Development Indicators World Bank Accessed on

20/9/2000 www.worldbank.org/data/databytopic

World Bank (2000b) World Development Indicators. Table 2.6 Wages and
Productivity World Bank Accessed on 20/9/2000
www.worldbank.org/data/databytopic

World Bank (2001a) Transition. The first ten years. Analysis and lessons for
Eastern Europe and the former Soviet Union, The World Bank,
Washington DC.

World Bank (2001b) World Development Indicators. Table 3.9 Sources of

Electricity World Bank Accessed on 20 March 2002
www.worldbank.org/data/databytopic

World Bank (2001c) World Development Indicators. Table 5.9 Power and
Communications World Bank Accessed on 20 March 2002
www.worldbank.org/data/databytopic

Zadeh, L. A. (1965) 'Fuzzy Sets' Information and Control, 8 3 338-353.

331
Zadeh, L. A. (1988) 'Fuzzy Logic' Computer, 83-93.

Zsiga, A. (2002) Nordic power sector: return to credit stability? Standard &
Poor's Accessed on 20/09/02
www.standardandpoors.com/Forum/RatingsCommentaries/Corporate
Finance

332