As part of the circuit breaker asset management analytics research EPRI collects and analyzes industrywide data to develop maintenance, asset management and model specific insights. These insights help in understanding where utilities are spending maintenance dollars, which models are performance outliers and whether changes are necessary to their maintenance programs.
Presented below are insights from analysis performed to date:
Insight 1 – Understanding breakdown of SF6 circuit breaker corrective maintenance (CM) by sub-system
Analyzing SF6 Circuit Breaker Corrective Maintenance Causes Based on CM Records from 10794 Breakers and more utilities (> 69kV) Reported over 10 Years
Gas Calls constitute 40-55% of the CM work orders for most of the utilities represented, that is, almost half of the corrective maintenance came from adding gas or fixing gas leaks.
Another observation is that for Utility B, Not Specified category accounts for about 25% of maintenance records. This is because the work order descriptions were blank or lacked sufficient information to put into a category for this utility.
In addition, Utility J’s Gas Call percentage is significantly lower than the other utilities. It’s possible that Utility J is very effective in containing gas leaks but may also be experiencing a high percentage of other maintenance issues in the Other CM category. This indicates that further exploration is needed to understand what is driving the Utility J results.
All these observations are potentially useful insights and point to avenues for further investigation.
Insight 2 – Percentage of breakers with Gas Calls To focus further on Gas Calls and breaker ages for the SF6 breakers, EPRI analyzed work orders for breakers at various ages to calculate the percentage of breakers that had at least one gas call in last 10 years and plotted with breaker age.
Percentage of breakers with at least one gas call in last 10 years vs breaker age. Analyzed 18362 SF6 breakers from 12 utilities (≥ 69kV).
Using empirical data, we can try and establish the relationship between gas call requirements and breaker age. For example, the point indicated in the chart shows that 28% of breakers at age 20 had at least one gas call in the last 10 years.
As expected, pattern shows that as breakers age, more percentage of breakers require gas call corrective maintenance.
For younger ages there are more breakers in the population analyzed, and the confidence bounds are smaller, clearly showcasing the confidence in the results. As we get into older ages, the number of breakers Inservice are less at these utilities and observe more scatter with larger confidence bounds.
Insight 3 - Comparing Gas Call Rates of Various SF6 Circuit Breaker Models Furthermore on Gas Calls, EPRI calculated the Annual Average Gas Call values of various SF6 breaker families as shown in the figure below.
Annual Average Gas Call is calculated for each breaker using up to the last 10 years of maintenance history. Annual Average Gas Call is calculated by summing the gas call counts for each breaker and averaging it annually by dividing by the number of years of recorded history.
Annual Average Gas Call for breaker = (Number of Gas Calls incurred by breaker)/ (Number of Years over which work order histories were observed)
Comparing Gas Call rates of various SF6 circuit breaker models for 8021 breakers (rated voltages = 123kV, 145kV, 170kV) from 12 utilities over 10 years.
The table includes the breaker count and the average age of the breakers by circuit breaker models.
As the plot reveals, there are significant differences in average annual gas calls among the different breaker families.
The Siemens-SPS, Siemens-TCP, GE-HVB and ABB-PA have higher annual average gas call than the other breaker models and, even though the average ages of the Siemens-SPS, Siemens-TCP, GE-HVB are similar (24, 26 and 28 years), there is a significant difference in Annual Average Gas Call values. ABB-PA has the highest value but is also the oldest family represented. HITACHI-HS and ABB-PM/PMI are younger families but incur higher Annual Average Gas Call.
This type of representation allows “apple-to-apple” comparisons of gas breaker families with a single metric, the Annual Average Gas Call.
The data can also be represented with a utility view to show which utilities have higher values in order to draw additional comparisons and inferences.
We have developed a dynamic report along similar lines that is shown below
Comparison of Annual Average Gas Calls for SF6 Breakers
The intention of the following interactive plot is to display a comparison of various SF6 circuit breaker models for a common calculated metric: Annual Average Gas Call, using the 10 year observation period. Based on the user selection for Voltage and Mechanism Stored Energy, the information for select breakers will display. There are two box plot distributions. The one on the top is the box plot distribution of Annual Average Gas Calls by breaker families and the one on the bottom is age distribution corresponding to the specific families. The table under the box plots shows the count of breakers and percentage of breakers with gas calls, for which these plots were generated. The takeaway is that this kind of representation allows comparisons among circuit breaker families, and one family across different utilities, and different families across one utility. This is an example demonstration and efforts are underway to curate and include more data from more utilities.
To interact with the plot:
Select the “1. Voltage” and “2. Mechanism Stored Energy” (By default 123, 145, 170kV and all mechanism types are selected).
Upon selection, the data appears in the plots and the table below. Moving the mouse on the plots reveal mean, median values for the breakers data based on the selection criteria. For example, the mean Annual Average Gas Call value of 1 indicate that the set of breakers included in this study on a average have 1 gas call per year per breaker. Relative comparison is possible among different breaker families. Another metric calculated is the percentage of breakers with gas calls. For example, a 50% value indicate that half of the breakers in the family have seen at least one gas call in the 10 year observation period. Thus gives an idea on the number of leaking breakers by family.
The “Display all utilities together” is a toggle button, can be used to display data from many utilities together or separately.
Additional selection is possible using “3. Breaker Family” and “4. Utility Designation” to display data from one or more utilities and families (press Ctrl to select multiple options).
1. Voltage
2. Mechanism Stored Energy
3. Breaker Family
4. Utility Designation
1 - Corrective Maintenance Requirements Based on Work Order Counts - Annual Average CM
To compare corrective maintenance activities across breaker families and across utilities, a new metric was developed, Annual Average CM.
It is calculated for each breaker and aggregated for families, as follows:
Annual Average CM for breaker = (Number of CMs incurred by breaker) / (Number of Years over which the work order histories were observed)
Annual Average CM for family = Sum [(Annual Average CM for breakers in the family)] / (Number of Breakers in Family)
Annual Average CM is calculated for each breaker using the maintenance history provided by the utility. For older breakers, EPRI has chosen to look back no more than ten years. As the name suggests, Annual Average CM for a breaker is calculated by summing the CM counts for that breaker and averaging it annually by dividing by the number of years over which work order histories were observed. For example, a breaker that had 4 corrective maintenance work orders over the last 5 years of available history has a calculated Annual Average CM of 0.8.
This metric can be used to assess corrective maintenance requirements for different circuit breaker families and draw comparisons across breaker families and different utilities.
In Figure 2-2, using standard box plots, the top row shows annual average CM values for four oil breaker families—two models from Allis/Siemens BZO, GE FK and Westinghouse GM breakers. Within the boxes the horizontal line shows the median and the X shows the mean. The box colors represent different utilities. The middle plot shows the age distribution and the table at the bottom gives the breaker count used to generate the plots.
This visualization is helpful in making comparisons. For example, the Westinghouse breakers have significantly more varied performance at the different utilities. Furthermore, Utility B shows better overall performance across breaker families. These observations raise questions such as:
Why is there a performance difference?
What is Utility B doing differently in maintaining its oil circuit breaker fleet?
These questions cannot be answered directly from this analysis, but it is possible to see which breaker family is incurring a higher level of maintenance than other families, which may influence CM requirements in the coming year. It is also possible to identify how a given breaker family is performing at different utilities, and how breaker families differ in their maintenance requirements.
The metric can also be used to look at different families within a single utility, or to make comparisons of gas and oil breakers.
Figure 2-3 is an example of the same metric used in an analysis demonstrating increasing maintenance with age for one family of gas breakers. The plot on the left shows the annual average CM count versus average family age for each utility. There is a clear trend of increasing CM with increasing age. The plot on the right shows the percentage of breakers with one or more CM actions. For example, at utility G (family average age around 12) only a little over 30 % of the breakers have needed corrective maintenance while at utility F (average age around 23), 90 % have. This analysis is only possible because the IDB has a collection of breakers of the same family from different utilities with different average ages for this family.
Such a metric may be useful in providing a quantitative, rather than anecdotal, view of the increasing need for corrective maintenance, and to help asset managers to forecast maintenance requirements as breakers age.
2 - Comparing Gas Call Rates of Various SF6 Circuit Breaker Models
To focus further on gas calls, EPRI analyzed work orders to calculate the annual average gas call values of various SF6 breaker families as shown in Figure 2-6.
Annual Average Gas Call is calculated for each breaker using up to the last 10 years of maintenance history. Similar to the Annual Average CM discussed before, Annual Average Gas Call is calculated by summing the gas call counts for each breaker and averaging it annually by dividing by the number of years of recorded history.
Annual Average Gas Call for breaker = (Number of Gas Calls incurred by breaker)/ (Number of Years over which work order histories were observed)
The table includes the percentage of breakers with gas calls. This is calculated for each model group by dividing the number of breakers in the group that have incurred at least one gas call corrective maintenance in the recorder history period by the total number of breakers in the group. As Figure 2-6 reveals, there are significant differences in average annual gas calls among the different breaker families. The ABB-PM, Siemens-TCP and GE-HVB have higher gas call work orders than the other three types and, even though the average ages of the Siemens and GE breakers are similar (24 and 28 years), there is a significant difference in the percentage of breakers with gas calls—62% of the GE breakers versus 27% of the Siemens breakers.
This type of representation allows “apple-to-apple” comparisons of gas breaker families with a single metric, the Annual Average Gas Call. The data can also be represented with a utility view to show which utilities have higher values in order to draw additional comparisons and inferences.
3 - Circuit Breaker Replacement Ranking
Because of the older demographic distributions common in many utilities, improvements are needed to provide more effective management of aging high-voltage circuit breaker fleets. Operating such equipment reliably and with a low risk of failure at or beyond typically assumed design lives is a subject of interest for many utilities. Consequently, developing and justifying a repair/replace (R/R) management strategy for these populations, and the rational basis for it, are increasingly important.
A rigorous repair versus replacement decision analysis would be mathematically quite complex and require many assumptions. However, an understanding of common utility asset management decision processes has allowed EPRI to develop a methodology to help identify candidate breakers for replacement and to rank a group of candidates in a rational and defensible fashion.
The circuit breaker replacement ranking methodology uses a risk-based approach designed to help utilities identify replacement candidates. It provides an analytical basis that is consistent and repeatable, objective and documented, risk based, and addresses regulatory concerns.
Using existing data, the methodology infers the relative costs and risks of keeping a breaker in service. Input parameters for applying this method are selected based on understanding how circuit breakers are built, operate, and wear and the selection and utilization of these parameter has been informed by review and analysis of circuit breaker IDB data.
The circuit breaker replacement rank (CBRR) is a function combining three major indices:
Functionality Index: How likely will the breaker be fit to continue operation?
O&M Index: What is the anticipated O&M cost that the breaker might incur?
Consequence Index: What are the consequences if the breaker fails to perform its function?
The method employs a relative ranking concept. The worst performer is assigned a maximum value of 100 and all remaining breakers are scaled proportionally. Figure 2-21 highlights the raw data curation, loading, consolidation and replacement ranking application process.
The CBRR algorithms utilize various weightings and thresholds. Analyses of circuit breaker IDB data using some of the metrics described earlier has informed and improved the selection of these values. Member utilities can work with EPRI to adapt the replacement ranking methodology and implement it to suit their unique situation and requirements.
EPRI releases a new version of the circuit breaker replacement software every year, which is available to funders of Program 34 or 34.002.
4 - Performance of Newer Breakers Based on Time to First Corrective Maintenance
Figure 2-1 shows Kaplan-Meier plots of the time to first CM for two specific breaker families; The colors mark the performance of these families at five different utilities. The populations of the different families vary among the utilities. The plots illustrate the different maintenance requirements of the families and reveal another insight: the performance of these families over this observation period is independent of the utility in which they are installed. It is quite reasonable to say that the circuit breaker performance in regard to need for maintenance in the first ten years is independent of how the breaker is used or maintained at a utility. In other words, it is an intrinsic characteristic of the breaker design. As shown in Figure 2-1, breaker families tend to behave similarly for first corrective maintenance (CM) requirements for younger breakers (< 10 years) at multiple utilities. The figure also shows that the spring breakers have a shallower slope than the hydraulic breakers. This indicates the hydraulic mechanisms require earlier CM compared to the spring breakers. Similar plots can be created for all the gas breakers, or families can be compared across utilities for benchmarking purposes.
5 - Maintenance Work Order Classification
The preceding examples used only the count of corrective maintenance work orders. Additional insights can be gained from understanding the specific maintenance done under each work order. Classification of corrective maintenance work orders into meaningful categories serves to more precisely associate breakers to the maintenance they have required. The EPRI maintenance template includes breaker work order classifications in four high-level categories as shown in Figure 2-4.
“Gas Calls” include work addressing gas containment subsystems
“Mechanism” covers work addressing mechanism components
“Others” encompasses all other work orders addressing circuit breaker functional failures
“CM Not Specified” denotes work orders with not enough information to classify
Extracting useful information from the work order descriptions and assigning them into meaningful bins is time and labor-intensive. EPRI is evaluating the use of Natural Language Processing as a means to make the process more efficient. See Applying Advanced Analytics to Improve Substation Asset Management and Maintenance: Assessing and Applying Natural Language Processing To Analyze Unstructured Text Records. EPRI, Palo Alto, CA: 2021.3002021202.
The effort put into classifying work orders yields potentially valuable insights into breaker maintenance.
For example, Figure 2-5 provides a breakdown of all corrective maintenance work orders over a ten-year period for eight utilities’ fleets of SF6 breakers. One observes that Gas Calls constitute 40-55% of the CM work orders for most of the utilities represented, that is, almost half of the corrective maintenance was for adding gas or fixing gas leaks.
Another observation is that for Utility B, Not Specified accounts for about 25% of maintenance records. This is because a number of work order descriptions were blank or lacked sufficient information to put into a classification for this utility.
In addition, Utility J’s Gas Call percentage is significantly lower than for the other utilities. It’s possible that Utility J is very effective in containing gas leaks, but may also be experiencing a high percentage of other maintenance issues in the Other CM category. This indicates that further exploration is needed to understand what is driving the Utility J results.
All these observations are potentially useful insights and also point to avenues for further investigation.
