REGULATORY AUTHORITIES ARE INCREASINGLY CONSIDERING PERFORMANCE-BASED REGULATION (PBR) to provide incentives for distribution utilities to gain economic efficiency. Typically, a specified service standard is established to discourage utilities from sacrificing service reliability while pursuing economic incentives. Performance standards are tied to a reward/penalty structure (RPS), which can be based on historic reliability records. By integrating historic data into the PBR plan, risk assessments and remedial work can be completed.

The Canadian Electricity Association has a long history of recording and disseminating information on utilities' reliability. Although individual utility data are confidential, overall Canadian utility performance is available in a comprehensive annual report that contains considerable detail on outage causes and relevant indices.

Table 1. Annual reliability data.

	Canada		System U		System I
Year	SAIFI	SAIDI	SAIFI	SAIDI	SAIFI	SAIDI
1994	2.55	3.39	3.12	1.76	1.90	3.09
1995	2.80	3.06	2.64	1.09	1.77	3.23
1996	2.39	2.86	0.90	1.11	1.77	3.14
1997	2.35	3.70	0.40	0.33	2.14	3.62
1998	2.40 3.58	3.32 30.31	1.69	1.36	1.85	3.56
1999	2.59	4.31	1.81	1.67	1.66	2.90
2000	2.26	3.23	0.61	0.41	1.68	3.03
2001	2.41	3.67	0.94	0.72	1.92	3.50
2002	2.33	4.06	1.35	1.05	1.50	2.89
2003	2.37 2.67	5.11 10.65	1.00	0.86	1.88	4.51
Ave	2.445 5.254	3.671 17.16	1.446	1.036	1.807	3.347
*Comment: The bolded data in all the tables exclude the effect of “Ice Storm 98” and the effect of the August 14 blackout and Hurricane Juan in 2003.

The Ontario Energy Board in Canada has stated that it will use historic utility performance to establish specified service reliability standards, and it requires that utilities have at least three years of reliability data. The board did not state in the first-generation PBR plan what would happen if utility reliability indices are out of the expected range. After sufficient data are collected, an RPS could be introduced into a second-generation PBR plan to encourage electric distribution utilities to maintain appropriate reliability levels. In this case, PBR provides the electricity distribution utilities with incentives to operate efficiently and to innovate. It also could introduce a potential financial risk due to the uncertainty associated with future system performance.

REWARD/PENALTY STRUCTURES AND PAYMENTS

An RPS integrated into a PBR plan works like a contract between a utility and its regulatory agency. The PBR protocol rewards the utility for providing good reliability and penalizes it for providing poor reliability. A common method of implementing an RPS is shown in Fig. 1. This structure has a “dead zone” where neither a penalty nor a reward is assessed. If the reliability is lower than the dead-zone boundary, a penalty is assessed. The penalty increases as the performance degrades and is frozen at a maximum penalty value. If the reliability is better than the dead-zone boundary, a reward is given. The rewards increase as the performance improves and is frozen when the maximum reward value is reached.

The primary service reliability indices used in North America and in many other parts of the world are the System Average Interruption Frequency Index (SAIFI) and the System Average Interruption Duration Index (SAIDI). The financial risk associated with an imposed RPS can be estimated by combining this structure with a related service reliability index expressed in the form of a probability distribution.

To initiate a PBR protocol, the historical average reliability index should reside in the dead zone of the proposed reward/penalty structure and preferably at the dead-zone center. The dead-zone width should be related to the standard deviation of the historic data and was set for illustrative purposes at twice the standard deviation in the studies described in this article. The remaining parameters in the RPS should be related to the incentive philosophy established by the regulatory authority.

HISTORIC RELIABILITY DATA

Table 1 shows the Canadian annual reliability indices (designated as Canada), and those of an urban utility (U) and an integrated utility (I) over the last 10 years.

The number of customers represented in the “Canada” system in 2003 is 11.58 million. The service area is 2.83 million sq km (1.09 million sq miles) and the peak load is 79,178 MW. The underground and overhead circuit lengths are 112,643 km and 680,275 km (69,993 miles and 422,703 miles), respectively. The circuit ratio of overhead to underground is 6.04, and the load density is approximately 28 kW/km². System U is a small urban utility, and System I is a large integrated utility. Their identities are unknown in accordance with the CEA confidentiality rules.

Table 2 shows the average values of SAIFI and SAIDI and their standard deviations for each utility system based on the 10-year historical data. The annual Canada historic performance is shown in the histograms in Fig. 2. These histograms clearly illustrate the annual variability in the indices.

Only 10 observations of each index are in this analysis. It was assumed that each system remains relatively constant over the period in regard to design and operational changes. This is a broad assumption. These histograms, therefore, provide approximate probability distributions of the indices. The histograms contain influences such as topological changes, maintenance practices and operational policies. The historical data, however, contain valuable information on the variation in the annual SAIFI and SAIDI service continuity indices, and provide an appreciation of the variation that can be expected in the future.

SYSTEM RISKS USING HISTORIC RELIABILITY DATA

As noted earlier, the average historic values of the reliability indices are preferably located at the dead-zone center. The dead-zone width was taken to be twice the index distribution standard deviation in the following studies. This process was used to create the dead zones shown in Table 3, using the 10-year historical data for the three representative utility systems.

Figure 2 shows the combination of the historical data for the Canada system and two hypothetical RPSs. The attention in this case is focused on the location of the dead zone rather than on the calculation of the expected reward/penalty (ERP) payments.

Figure 2 shows there is a 10% probability that the system SAIFI will lie in the penalty zone and also a 10% probability that it will lie in the reward zone. The system SAIDI has a 20% probability of residing in either the reward or penalty zone. The utility could realistically expect some future penalty payments according to the historic performance. It can be seen from Fig. 2 that the most probable outcomes lie in the designated dead zone. The utility could possibly make some improvements, which would move its performance toward the reward zone and therefore receive financial remuneration from the regulator.

Ontario's first-generation PBR plan only requires that a utility's performance be the same or better than previously experienced. The SAIFI and SAIDI ranges in the last 10 years (1994-2003) for the Canada data are [2.26, 3.58] and [2.86, 30.31], respectively. This includes the effects of the 1998 ice storm as well as the effects of the 2003 August blackout and Hurricane Juan. These were obviously major system disturbances. If these data are excluded, the SAIFI and SAIDI ranges are [2.26, 2.80] and [2.86, 5.11], respectively. The regulator will have to decide if events of this magnitude are to be included in the assessment of system performance.

Table 2. The average SAIFI and SAIDI values and their standard deviations.

System	SAIFI		SAIDI
	Ave.	S.D.	Ave.	S.D.
Canada	2.445	0.159	3.671	0.672
System U	1.446	0.880	1.036	0.479
System I	1.807	0.175	3.347	0.486

Table 3. The dead zones for the three representative utilities.

System	Dead Zones
	SAIFI		SAIDI
Canada	2.286	2.604	2.999	4.343
System U	0.566	2.326	0.557	1.515
System I	1.632	1.982	2.861	3.833

The methodology used to establish the dead-zone values shown in Table 3 provides a consistent approach to create the maximum and minimum bounds based on the utility's past performance. The decision to use ±1 standard deviation is arbitrary and should be studied by the regulator. The same approach also can be used to examine the effect of dead-zone location on the remaining two utility systems.

ANALYSIS OF INTERRUPTION CAUSES

The system reliability characteristics of individual utilities differ due to the diversities in service areas, load densities, system topologies, weather environments and service standards. Urban systems usually have short supply feeders, underground circuits and alternate power supplies. Therefore, their reliability indices are generally better than those in rural or integrated urban/rural systems.

An examination of the contributions to the service continuity indices from various system factors provides considerable insight into how the performance can be improved. The CEA reporting system divides the customer outages into the following cause codes:

• Unknown (Unkn)
• Scheduled Outage (Sch.O)
• Loss of Supply (Los.S)
• Tree Contact (Tr.C)
• Lightning (Lightn)
• Defective Equipment (De.E)
• Adverse Weather (Ad.W)
• Adverse Environment (Ad.En)
• Human Element (Hu.E)
• Foreign Interference (For.I).

The contributions to the service continuity indices can come from different causes in urban and rural systems. This section presents the interruption contributions for the selected utility systems over the last 10 years. The designated contributions to the Canada SAIFI are shown in Fig. 3. The Canada aggregation represents a huge integrated system. It can be seen from Fig. 3 that loss of supply makes a significant contribution to the overall SAIFI and accounts for approximately 30% of the index. It also exhibits considerable variability and therefore makes a significant contribution to the annual index variability. Defective equipment and tree contact provide significant contributions to the annual index but are relatively constant contributors over the 10-year period.

Figure 4 shows the major cause contributions to the SAIDI index for the Canada system. Loss of supply makes a smaller contribution to SAIDI compared to its contribution to SAIFI. Adverse weather is a major contributor to both the annual SAIDI and the variability of the index.

Figures 5 and 6 show the SAIFI and SAIDI values for System U, which is a small urban utility with a low circuit ratio and high load density. It can be seen that loss of supply makes a significant contribution to the annual indices in 1995, 1998, 1999 and 2002. It accounts for 40% of the total index in some years and greatly affects the overall SAIFI and SAIDI profiles. Defective equipment creates major fluctuations in SAIDI in many years. Its influence on SAIFI is notable but less than that on SAIDI. Lightning and foreign interference are the other major contributors to the magnitude of the annual indices and their variability. The contribution of adverse weather is generally low for System U, other than in 1996.

System I is an integrated utility with a high circuit ratio and low load density. The major contributions are shown in Figs. 7 and 8. It can be seen that loss of supply has minimal effect on both the SAIFI and SAIDI indices in the first six years but makes relatively large contributions in subsequent years. Lightning makes a noticeable contribution to the variability of the total SAIFI value. Scheduled outages make significant contributions to both SAIFI and SAIDI for this utility. The contributions due to defective equipment and the remaining causes are constant. Adverse weather has more influence on SAIDI than on SAIFI and contributes to the variability of the total index.

In general, it can be seen that loss of supply has a significant effect on the annual indices for Canada and the high load density System U, and provides much smaller contributions for the low load density System I. Loss of supply also has a larger impact on SAIFI than on SAIDI, due to the relatively short restoration times associated with bulk system outages. Defective equipment makes a significant contribution in each system, which remains constant. Tree contact has considerable influence on the Canada indices. Lightning is a significant cause contributor to SAIFI, and interruptions from foreign interference occur more frequently in System U.

The magnitude and variability associated with the cause code contributions provide a new dimension in the examination of the overall SAIFI and SAIDI indices. The variability in the system indices shown in Fig. 2 is due to the variability in the cause code components. The loss of supply component could be considered to be outside the control of a distribution company, and therefore excluded from consideration in a PBR analysis. This exclusion not only will affect the average annual index, but will also affect the index distribution, and therefore the financial risk faced by the distribution utility in a PBR framework. Remedial action to improve the SAIFI and SAIDI value in an attempt to increase the probability of reward payments should be related to both the magnitude and the variability associated with the cause code contributions.

The reliability standards used in Ontario's first-generation PBR plan are intended to minimize bad outcomes in the future. The Ontario Energy Board has indicated that after enough data are collected and experience is accumulated, a new scheme involving reward/penalty policies may be introduced in the second-generation PBR plan. Historic utility service continuity performance can be used to establish an appropriate RPS, which reflects the existing situation and also includes incentives determined by the regulator in regard to future desired performance.

The index distributions created from the historical performance data are based on a relatively small number of actual observations. These are the data utilities and regulators will be using to make important social and financial decisions. This includes both the overall indices and the cause code contributions. The concept of using index distributions in addition to the average annual values is an important tool in assessing financial risk and creating appropriate reward/penalty structures in a PBR regime.

---

Roy Billinton obtained bachelor's and master's degrees from the University of Manitoba, and Ph.D. and DS degrees from the University of Saskatchewan. He joined the department of electrical engineering at the University of Saskatchewan in 1964 and is now an emeritus professor. roy.billinton@usask.ca

Zhe Feng obtained a BSEE degree in 1999 from Anshan University of Science and Technology, Anshan, China, and worked for Anshan Design Institute of Mining Company, China, in the electrical department as an assistant engineer. She is currently working on a master's degree at the University of Saskatchewan, Canada.