DURING THE LATE 1960S AND EARLY 1970S, THE RESIDENTIAL DEVELOPMENT BOOM in suburban areas prompted electric utilities to begin investing heavily in underground cables. The quality of the cable was rather low by today's standards, and now a significant amount of the cable installed during that period is nearing the end of its expected life. This is of particular concern because increasing cable failures negatively impact system reliability and customer satisfaction.
Although the proactive replacement of underground cable systems helps reduce these negative impacts, it also requires a substantial capital investment for any utility. Therefore, it is extremely important to make timely, cost-effective decisions regarding the testing, repair and replacement of underground cable.
The direct-buried cable-replacement program at Baltimore Gas & Electric (BGE) is grouped into two major areas:
Proactive — Scheduled replacement of cable based on prioritization methodology involving various weighted factors.
Reactive — Replacement of a cable that faulted and resulted in customer service interruption.
A Six Sigma team, led by Project Manager Rick Knotts, was commissioned at BGE to improve its existing cable-replacement program. The team developed a new methodology by which it schedules proactive replacements. The methodology is based on the outputs of underground-cable asset-management software developed by the National Electric Energy Testing Research and Applications Center (NEETRAC) at the Georgia Institute of Technology, as well as some custom NEETRAC analysis performed for BGE and an optimization process developed by BGE's Six Sigma team.
Since 2002, NEETRAC has been developing a software application to assist with cable-failure analysis in cases where little historical data is available. Dr. Miroslav Begovic, a professor of electrical engineering at Georgia Tech, leads the NEETRAC software development with assistance from former graduate students Aleksandar Vukojevic and Gregory Henry, and current graduate student Joshua Perkel.
The software is designed to extract statistical information from data possessing significant uncertainty. The software uses a three-parameter Weibull distribution to describe cable life characteristics as well as a Monte Carlo method to obtain a range of possibilities for future behavior.
To avoid extensive data-mining of utility records, there are a few assumptions the NEETRAC software makes:
Reactive replacement is completed on such small segments of cable such that, statistically speaking, perfect repair occurs (and the failure rate is not reset).
Cable failure is due to cable age and is consistent with the Weibull statistics.
Proactive replacement is performed on the oldest cable in the system to garner the greatest benefit.
Because failures are not associated with a particular cable age group, the software was designed to require only three input variables for each year that the cable is in service: number of miles installed, number of miles replaced and number of failures. Each cable type (#2 aluminum jacketed, 350 aluminum nonjacketed, etc.) is processed independently.
The primary function of the NEETRAC software is to forecast, for a few years ahead, the amount of cable that must be replaced to achieve a desired failure rate. Weibull regression is applied to the inputs to construct the cable life characteristics, and then a Monte Carlo method is applied to make the predictions.
Because a Monte Carlo method is used, confidence intervals are naturally associated with the predictions. However, there is a caveat: the number of years described by the input data will affect the usefulness of the output. Short data sets will have wide confidence intervals, making accurate prediction nonviable. Longer data sets and data sets with extensive quantities of cable in service will have tighter confidence intervals, and subsequently allow more accurate prediction, further into the future.
Unless large cable populations are carefully tracked over a long period of time — 20 years or more — attempting to forecast more than a few years may provide unusable confidence ranges.
The “do-nothing” case is the starting point for analysis — the worst-case scenario. This involves making predictions of the future failure rate, assuming that no cable will be replaced. Figure 1 shows the annual failure rate for an example cable type. The estimated failure rate (median) is superimposed on the actual failure rate to demonstrate “goodness of fit” and predictions for the next five years are given along with confidence intervals. The predicted values are calculated with zero cable replacements. This is analogous to investing no money in a proactive cable-replacement program in any future year. In this particular example, it is evident that the number of failures will increase rapidly if no cable is replaced.
CONSTANT FAILURE RATE
To prevent system reliability from decreasing in the future, it is necessary to maintain at least a constant failure rate each year. Assuming that there is a desired failure rate (DFR) in the present year and that the failure rate is to be held constant at the DFR in the following year, the NEETRAC software can estimate the total number of miles that must be replaced (MR) with confidence intervals (Fig. 2). It is important to recognize that replacing MR miles does not guarantee the DFR will be achieved; the DFR could fall anywhere within the confidence interval.
Cable sensitivity — the failure rate reduction per cable mile replaced — can be extracted from Fig. 2 by calculating the slope of the tangent line through the appropriate point (MR, DFR) on the curve.
The median value is used in initial planning stages because there is an equal chance (50%) the actual failure rate will be less than or greater than the DFR. In order to evaluate results, it is beneficial to examine two relationships: the confidence of a DFR estimate for a specified MR and the confidence of an MR estimate for a chosen DFR. These relationships are shown in Figs. 3 and 4, respectively.
Figure 3 shows that 75% of the time, the actual failure rate will lie between 84% and 112% of the DFR, while Fig. 4 shows that 75% of the time, the actual number of MR miles to achieve the DFR will be 20% to 158% of MR. The upper bound is used for making worst-case estimates because greater confidence implies that greater amounts of cable must be replaced.
BUDGET AND RELIABILITY PLANNING
Utility cable-replacement programs typically designate cables as main or tap cables. Main cable-replacement jobs typically cost two to two-and-a- half times as much as tap jobs, but, in turn, each avoided main cable failure can save approximately 25 to 35 times as many customer interruptions, and two to three times as many customer hours lost for a typical feeder design.
As mentioned previously, one excellent feature of the NEETRAC software is that it estimates the sensitivity of the cable-failure rate to miles replaced. By incorporating the replacement cost for main and tap cable, a weighted coefficient can be calculated that describes the greatest benefit for every dollar spent on replacement.
There are two ways to develop a cable-replacement program, with or without a financial constraint. If there is no financial constraint, then the number of miles to replace for each cable can be obtained from analysis using the NEETRAC software (for example, the constant-failure-rate scenario). If a financial constraint does exist, then the goal is to develop a cable-replacement program that will result in the greatest reduction in failures per dollar invested. This latter type of program is developed in Table 1, where the percentage of the fixed budget to spend on each cable type is indicated by “PFB” and the percentage of the total miles replaced for each cable type is indicated by “PTMR.”
Taking into account the confidence intervals on the number of miles to replace and multiplying that by the corresponding installation cost will yield the “investment confidence” plot shown in Fig. 5. Investment of 1.0 per unit corresponds to the targeted program budget.
This figure demonstrates that, in order to keep the number of failures at the minimum level achievable with the financial constraint, it might cost 40% more than originally planned to be 95% confident in the estimate. This may seem extreme, but it is important to remember that the actual number of failures may fall anywhere in the confidence range; thus, risk must be carefully leveraged against cost.
To determine the potential reliability benefit to the system, it is necessary to calculate the system average interruption frequency index (SAIFI) and the system average interruption duration index (SAIDI) impacts of cable replacement. For example, assume that every failure avoided on a tap cable saves 20 customer interruptions with an average duration of 1 hour, and 600 customer interruptions on a main cable with an average duration of 3 hours. Also assume a total of 1 million customers on the system. Table 2 gives example values for expected SAIFI and SAIDI improvements in the following year if the PFB values of Table 1 are used. The “avoided failures” column in Table 2 lists example values, but would normally be calculated using values in Table 1 combined with the actual budget.
Figure 6 displays the theoretical benefit to SAIFI and SAIDI for increased investment in the cable-replacement program.
|Cable Type||One Year Estim. Failures With No Replace- |
tivity (Failures Saved Per Mile Replaced)
ation Cost Units
|Failures Saved Per Unit Cost||PFB||Miles Replaced Per Unit Budget||PTMR|
|Cable Type||Avoided Failures||Avoided Customer Inter- |
|Avoided Customer Hours Lost||SAIFI Improve- |
|SAIDI Improve- |
Based on its Six Sigma project in 2005, BGE has begun a cable-replacement program with a new prioritization methodology that uses NEETRAC software to forecast cable failures. The improved cable-replacement program will help BGE better meet its strategic objectives in reliability, customer satisfaction and cost management in upcoming years.
Because of the statistical nature of the predictions made by the NEETRAC software, the precision of the results is a function of both quantity and quality of the available input data. For example, more accurate estimates can be made if failures are associated with the year of installation, rather than assuming, in the absence of better data, that the failure occurs on the oldest cables on the system.
The authors would like to acknowledge contributions by several BGE employees: John Borkoski, director of system and reliability planning; Gerry Schmidt, general supervisor in system protection and control; Andy Dodge, director of system of operations; Frank Tiburzi, principle engineer in restoration services and operations support; and everyone who participated in the Six Sigma project. Many thanks also to Dave Gaugh, John Birrane and Joe Mullen for their effort in procuring photos.
The authors also would like to credit and give thanks to Dr. Miroslav Begovic, professor of ECE at Georgia Tech; Joshua Perkel, graduate student at Georgia Tech; and Rick Hartlein, NEETRAC program manager, underground systems.
Greg Henry is an EIT-certified engineer in system operations at BGE. He earned his BSEE degree from the University of Texas at Austin in 2003, and his MS degree in electrical and computer engineeering from the Georgia Institute of Technology in 2004. firstname.lastname@example.org
Rick Knotts is a Six Sigma Black Belt at BGE and has been leading Six Sigma projects for the last two years in BGE's Business Performance Improvement area. He has worked for BGE for more than 15 years in various engineering capacities in the following areas: system operations, system planning and substation design. He earned his BSEE degree from the University of Delaware in 1986. He has been a professional engineer in Maryland since 1993. email@example.com
Aleksandar Vukojevic joined BGE as a system protection and controls engineer in 2004. He works on the design of protection and control schemes, switchgear design, relay settings and coordination, and relay commissioning tests. He earned his BS degree in applied mathematics from Kennesaw State University in 2000, and his BSEE and MSEE degrees from the Georgia Institute of Technology in 2003 and 2004, respectively. He is EIT certified. firstname.lastname@example.org