We Energies, like many utilities, has been actively tracking and benchmarking distribution reliability statistics for years, using the standard indices SAIFI, SAIDI and CAIDI for sustained outages. Its distribution reliability group is continually looking for insight into what drives these indices, how to improve them and how other utilities compare. Participation in several reliability benchmarking groups has shown consistent results for We Energies: 1^{st} quartile SAIFI, average SAIDI and 4^{th} quartile CAIDI.
Overall reliability performance in 2001 and 2002 earned We Energies (Milwaukee, Wisconsin, U.S.) the PA Consulting Group Reliability One Award for the Midwest. However, We Energies is not satisfied with its CAIDI performance and has established a goal to significantly improve CAIDI while maintaining SAIFI. Some natural questions have arisen: Do any utilities score in the 1^{st} quartile for all three indices, and what are their characteristics? Is high CAIDI partially due to very low SAIFI? And is it possible to drive down CAIDI without adversely affecting SAIFI?
CAIDI is perhaps the least straightforward of the indices. While SAIFI and SAIDI are driven primarily by frequency or time duration respectively, both variables drive CAIDI. Strategies for improving SAIFI and SAIDI can sometimes adversely affect CAIDI.
SAIFI is improved by reducing the frequency of outages (for example, by tree trimming and maintaining equipment). SAIFI is also improved by reducing the number of customers interrupted when outages do occur (for example, by adding reclosers and fuses).
Strategies that reduce SAIFI improve SAIDI because if an outage does not happen, it doesn't add to duration. SAIDI is also improved by improving CAIDI through faster customer restoration. However, system improvements can make CAIDI go up or down, depending on whether the improvements have a greater effect on outage frequency (customer interruptions) or outage duration (customer minutes of interruption.)
Usually, the three reliability indices are graphed independent of each other. Figure 1 (a) shows typical benchmark data, which can be obtained through participation in benchmarking groups, or from mining data in the public domain such as PSC/PUC Web sites. (Of course, when comparing reliability statistics among utilities one must be cautious to compare “apples to apples,” considering, for example, whether major storms are included or whether sustained outage definitions are the same.)
Figure 1 (b) shows trending data. Displaying the data separately makes it difficult to judge overall reliability. For example, is We Energies performing better or worse than Utility 11? Has We Energies reliability improved over the past five years?
While working to understand CAIDI better, We Energies found it useful to remember the relationship between the indices, as defined by IEEE 13662001:SAIDI = SAIFI * CAIDI
In this equation, SAIFI can be considered the independent variable — there can be no outage duration (SAIDI) without an outage frequency (SAIFI). This representation of the indices suggests graphing them together, with SAIFI on the xaxis and SAIDI on the yaxis. Lines of constant CAIDI pass through the origin. Thus, all three indices are shown at once.
Data in Fig. 1 is redisplayed in Figs. 2 and 3. We Energies refers to SAIFI/SAIDI/CAIDI graphed in this manner as the “reliability triangle” because the benchmark quartile lines form a triangle. The closer to the origin a point lays, the better the reliability.
This method also is useful for tracking the indices over the course of a year (Fig. 4).
Displaying the indices in a reliability triangle has several advantages:

Data reporting is condensed.

The fact that the indices are related is emphasized.

Reliability performance is clarified.

Cumulative tracking shows that SAIFI and SAIDI can only increase over the year, while CAIDI can increase or decrease.
The reliability triangle is a useful benchmarking and trending tool that provides insight into SAIFI, SAIDI and CAIDI. We Energies has used it to concisely communicate annual benchmarking results to managers and employees. It clarified that some utilities are able to achieve 1^{st} quartile performance in all three indices. Further research revealed that 1^{st} quartile utilities often have systems that are predominantly urban and underground.
Another benefit has been to promote the understanding of CAIDI and its relationship to SAIFI and SAIDI outside the distribution reliability group, and without relying on equations. The reliability triangle shows graphically that SAIFI and SAIDI may both improve (decrease) while CAIDI could still increase.
Ultimately, the reliability triangle helped We Energies with its approach to CAIDI improvement. Given the goal to improve CAIDI while maintaining SAIFI, strategies were implemented to reduce outage duration (CMI) while holding outage frequency (CI) constant.
Initiatives that improved CAIDI included:

Implementing automated callout of troubleshooters and crews for faster outage restoration.

Deploying faulted circuit indicators on the mainline of circuits that exceed annual CMI thresholds.

Increasing troubleshooter staffing and hours of coverage.

Promoting an attitude of “restore before repair.”
As of October 2004, We Energies was on target to achieve its yearend CAIDI goal, which represents a 20% improvement. CAIDI initiatives are working, with help from insights gained using the reliability triangle.
Jennifer Rothwell is a senior electrical engineer at We Energies with 15 years experience in distribution reliability and T&D protection. Rothwell is a licensed professional engineer in Wisconsin and a member of IEEE Power Engineering and Women in Engineering Societies. She holds BSBE and MSEE degrees from Marquette University.
jennifer.rothwell@weenergies.com
Interruption Indices
SAIFI, SAIDI and CAIDI are measures commonly used to report the average frequency and duration of sustained outages. They are defined over a fixed time period, usually a month or a year. The indices can be measured over the entire electric distribution system or over smaller portions of the system, such as an operating area or individual circuit.
For each outage that occurs, the utility tallies the number of customers interrupted (CI.) The product of CI and minutes of outage are calculated to determine customer minutes of interruption (CMI) for each outage.
The indices are defined by:
SAIFI (System Average Interruption Frequency Index): Sum of outage CIs/Total number of customers served.
SAIDI (System Average Interruption Duration Index): Sum of outage CMIs/Total number of customers served.
CAIDI (Customer Average Interruption Duration Index): Sum of outage CMIs/Sum of outage CIs.
SAIFI measures how often a customer can expect to experience an outage, SAIDI measures average outage duration per customer, and CAIDI measures average outage duration if an outage is experienced, or average restoration time.