STRATEGIES AND PROGRAMS FOR IMPROVING THE RELIABILITY of our transmission and distribution networks should be targeted for maximum benefit. Outage databases can help identify which problems are leading to disturbances. Specifically, they can help identify which circuits to target, which areas have the most problems with trees, which areas have the most problems with animals and so on. In addition, outage databases also can help judge the results of improvement programs.
Without accurate and thorough outage databases, there is no evidence to locate legitimate targets for reliability improvement programs. At the same time, we need relatively simple database entry codes and remarks fields, along with guidelines and training for personnel who enter outage data. Audits and assessments are performed to assure the codes are entered accurately. In many utilities, budgets for reliability improvement programs are based on reliability information extracted from the outage database. Reliability-based budget strategies typically address underground cable replacement, vegetation maintenance and “worst-circuit” programs. If incorrect failure codes are entered, a region may lose money for certain reliability and asset management programs.
ALL OUTAGES CONTAIN AN ELEMENT OF MYSTERY
The initial coding of outages is based on the judgment and training of both the first responder and the employee entering the data. However, the initial coding often does not identify the primary root cause. Most outages have more than one root cause. The primary root cause is something for which a utility can implement economical corrections within a reasonable period of time. The purpose of codes and coding conventions is to identify the primary root cause of each outage.
Having accurate and precise outage-code systems increases the usefulness of outage databases. Weather is a common outage-cause code, but what does that mean? If a tree knocks down a distribution line during a storm, is that weather? How do you differentiate between lightning and tree-caused outages during a thunderstorm? Another common blunder is tagging a cause as “cutout” when the cause was really something downstream that caused the operated fuse to clear the fault. As much as possible, good outage-code systems should separate the root cause of the outage from the weather, the protective device in operation and the equipment affected. One suggestion is to use a separate category for weather and indicate a major storm separately. Also, consider codes that reveal deficiencies, including inadequate clearances, deteriorated equipment, missing animal protection or broken insulation.
Duke Power (Charlotte, North Carolina, U.S.) characterizes an outage by specifying four code fields: interrupting device, cause or failure mode, equipment code and weather. These codes, used in combination with each other, are quite accurate in describing what is known about the cause of an outage, or in some cases, what is not known about an outage.
DOES WEATHER CAUSE OUTAGES?
There is often discussion regarding if weather is an outage cause or a contributing factor. From a reliability perspective, the electric distribution system is designed to withstand moderate to bad weather conditions. Weather that exceeds the design parameters of the system can be considered a primary root cause. Weather situations that may be considered primary root causes include: lightning, extreme wind, extreme ice or flooding. Given that, utilities should avoid declaring stormy weather such as wind or lightning as a cause. There may be evidence (burn marks for example) to show that lightning caused an arrester failure, but if there is not, it is better to code the cause as an equipment failure during a lightning storm. Utilities should make sure that inspectors don't get into the habit of using weather as an excuse to explain outages.
Reliability programs can address system weaknesses and thus prevent many outage instances during stormy weather. Properly funded and executed vegetation management programs cut down on outages during high winds. Properly designed overhead lines and equipment protection substantially reduce outages during lightning. If a utility is not attempting to make the system more outage resistant during stormy weather, it may be thinking of weather as a cause rather than a contributing factor.
Engineers often want to specify additional codes including vintage year (for equipment failures), overhead/underground indicators and how the outage was restored. Most of these codes can be incorporated in Duke Power's four main codes given previously. If a utility offers too many codes, first responders will likely ignore some of them. Experience on the Duke Power system has shown that the four codes mentioned previously are about the limit that a first responder can accurately handle, even if other codes are requested.
DO NOT OVERLOOK THE COMMENTS FIELD
A generic comments or remarks section provides specifics that might not be clear from the outage codes. This helps with later analysis and allows for keyword searching, thus revealing patterns. The remarks field contains vast amounts of information if used liberally. First responders and dispatchers can use the remarks field to describe what happened during the outage. Modern database applications can quickly search and count outages that contain certain keywords or phrases within the remarks field. These database applications can also filter out records that contain certain keywords.
Here are some examples of the use of the remarks section at Duke Power:
There was no code to differentiate between live or dead tree outages in the code system. However, work crews almost always noted in the remarks field if the tree or limbs involved were “dead” or “rotten.” This method of counting dead tree outages versus live tree outages was found to be a highly accurate for determining if the annual danger-tree survey and removal was effective.
There was a hypothesis that 3-D or smaller transformer fuses were more susceptible to nuisance fuse outages during lightning than 5-D or larger transformer fuses. There were a large number of outage records of fuse outages on transformers during periods of lightning, but the fuse size was not recorded. However, it was discovered that the crew almost always called in the fuse size when they replaced a fuse. The dispatchers put this information in the remarks field. A contextual database search on the remarks field of transformer outages found more than 7000 cases per year where the transformer fuse size was supplied. This sample size was sufficient to allow normalization of 3-D and 5-D transformer outages against the total population of transformers that would have such fuses. The result of this analysis showed that there was no significant difference in the performance of these two fuse sizes.
There was a concern that squirrels were causing significant damage to overhead conductors, connectors and equipment by literally biting and gnawing on various items. The remarks field was searched to find how widespread this behavior was. Duke Power determined that squirrels will chew and gnaw almost anything, but bare aluminum conductor is their favorite. Also, while widespread, this phenomenon was not found to be a significant reliability problem.
SELECT THE RIGHT OUTAGE CODES
When designing or modifying outage-code systems, balance the desire for as much information as possible with the reality that overloading crews with too many options can be counterproductive. Use codes that are specific, not too generic to provide useful information. Use codes that are simple. A balance must be maintained between being too broad and too detailed.
Here are additional guidelines for outage coding:
Where appropriate, subcodes provide extra information. As an example, it may make sense to have a code or subcode to denote whether a structure has an animal guard. If that adds too much complexity, have crews enter a generic comment to indicate the structural deficiency (the missing animal guard). Also, for animal outages, crews should only mark it “animal” if they find the remains of an animal or similar proof.
Using unknown as a cause is better than guessing. A wrong cause code can result in resources being directed to the wrong problem.
- Equipment failures
Equipment failures are prime candidates for fine-tuning. For example, a subcode could indicate the equipment construction, vintage or even the manufacturer. For cables and especially for splices, knowing the manufacturer and vintage can help determine targeted replacement programs. The mode of failure is also an important consideration. It is valuable to know whether the equipment was broken or the failure was from decay?
Consider an example outage code for a tree contact during a storm that caused a downed wire that might be tagged as follows:
Interrupting device: Breaker
Cause: Vegetation (dead tree)
Equipment: Bare conductor (downed wire)
In this coding example, the inspector clearly separates the cause (the tree) from the impact on the system (downed wires) and notes the weather conditions during the event. The items in brackets denote subcodes that add useful information; in this case, the subcodes reveal that the tree was dead (points to a better hazard-tree program) and that it caused a downed wire.
An arrester failure found during a lightning storm might be classified as:
Interrupting device: Transformer fuse
Cause: Equipment failure
Equipment: Arrester (catastrophic)
Weather: Lightning storm.
In this case, the inspector doesn't really know the cause of the arrester failure, although it was probably caused by lightning. The inspector doesn't even know whether the arrester failed during the storm (the arrester may have failed earlier and been found only when crews arrived at the scene). But because the arrester was obviously failed just downstream of the fuse and the transformer was still operational, the arrester likely caused the fault.
Consider a squirrel outage across a bushing:
Interrupting device: Tap fuse
Equipment: Bushing (no animal guard)
The inspector's use of a subcode for bushing that highlights a common deficiency (no animal guard) helps direct resources to repair the deficiency.
Crew tendencies should be taken into consideration. The use of mobile computers allows direct data entry, but crews are prone to incorrectly using the software. Training and straightforward user interfaces can help. The advantage of computer data entry is that the form can adapt to the scenario at hand, while providing data from the outage-management system (including outage start time). Having a crew call back outage information to dispatchers allows the dispatchers to query the crew to make sure the codes are entered consistently. Paper data entry sheets can be relatively easy for crews to interpret but limit flexibility for subcodes and options.
A utility culture should be fostered to encourage crews to enter accurate outage codes. It helps to provide training and documents to demonstrate how all outage codes are to be used. Guidelines and code sheets should be provided to all who work with entering outage codes. At Duke Power, dispatchers enter outage codes based on information entered in by first responders. At the 24-hour center, dispatchers are graded on how accurately the code guidelines are followed.
Local reliability engineers or technicians are the “code police.” Each day, these employees review all the outage records in their location for the past 24 hours or over the weekend to make sure code guidelines are followed. If information in the outage record is missing, unclear or contradictory, the reliability engineer tracks down the information and completes the entry correctly. The code police are also assessed quarterly on the accuracy of outage records in their location. Those who police outage data should be rewarded for the accurate and complete reporting of outages. To avoid conflicts of interest, these individuals should not be evaluated based on the annual outage indices such as SAIFI or SAIDI.
A number of tools and options are available for analyzing outage databases. Spreadsheet and database sorting and queries can reveal a number of patterns. More sophisticated data mining tools also are available. EPRI Solutions (Knoxville, Tennessee, U.S.) has developed and used the open-source tool Rpad (www.rpad.org) for a number of outage data mining projects. This tool has advanced data manipulation with statistical capabilities that allow for sophisticated trending and correlations, which can be useful for evaluating reliability programs.
Tying outage data to GIS or other mapping systems can reveal location-specific trends. Many types of faults are highly repeatable because of location-specific structural deficiencies. Totarget these deficiencies, it helps to be able to tie outage location to a circuit map, enabling utility personnel to graphically review outage locations and outage type.
The circuit on page 58 historically had high numbers of tree outages, many of which were clustered along certain parts of the circuit. This provides a first indication of where to look for problems in the field. Just as it helps to tie outage locations to map locations, it also helps to tie outage locations to construction type. With this tie-in, one can evaluate fault rates for different construction types and target maintenance, upgrades and design changes as appropriate.
With clear outage codes and accurate outage databases, utilities can more effectively target reliability programs. Vegetation management is a prime example. Utilities can target their danger-tree programs to circuits where the outage databases reveal high numbers of mainline faults, so that circuits with poor tree-outage histories receive more frequent pruning. Outage databases can then be used to judge the effectiveness of program changes. By implementing a danger-tree program and comparing the results with the remaining circuits, the effectiveness of the program can be evaluated.
GUESS AND TEST
An accurate outage database provides a gold mine of data that can be used to test various hypotheses. Duke Power has learned that at least 100 incidents are needed to avoid undue bias. The outage database may have sufficient data to test a given hypothesis. As an illustration, Duke Power observed that 30% of its line transformer failures occurred on dead-end poles. A number of hypotheses were considered to explain this statistic (possibly from increased lightning stress). After further investigation, Duke Power found that 29% of its transformers were installed on dead-end poles, so the conclusion was reached that dead ends do not contribute significantly to transformer failures.
Outage databases can also be used to develop any number of reliability improvement programs including: recloser applications, grounding improvements, animal-guard installations, tree pruning, and application of automation. Each program requires different database information.
Tom Short is a senior engineer with EPRI Solutions Inc. in Schenectady, New York. Before that, he worked for Power Technologies Inc. for 10 years. He has performed lightning protection, voltage sag, flicker, capacitor application and load-flow analysis studies and has been involved in several distribution system monitoring projects. He received an MSEE degree from Montana State University in 1990. As chair of the IEEE Working Group on the Lightning Performance of Distribution Lines, he led the development of IEEE Std. 1410-1997. Short authored the Electric Power Distribution Handbook (published by CRC Press, 2004) and co-authored with Lee Taylor chapter five of the Distribution Reliability and Power Quality (published by CRC Press, 2006). firstname.lastname@example.org
Lee Taylor has worked as a distribution engineer at Duke Power for 34 years, serving as the lead distribution reliability engineer for the past 15 years. Taylor holds the title of consulting engineer. His areas of responsibility include distribution reliability data analysis, field investigations of fault sources, tactical and strategic fault prevention. Taylor received his bachelor's degree in physics from the University of North Carolina in 1971. He is a licensed professional engineer, IEEE member, and currently serves as an officer on the SEE Power Quality and Reliability Sub-committee. With Tom Short, he co-authored chapter five of the Distribution Reliability and Power Quality (published by CRC Press, 2006). email@example.com
DUKE POWER'S CODE FIELD DESCRIPTIONS
Duke Power has specified four code fields to characterize its outages:
Interrupting device — Fuse, line breaker (recloser), station breaker and transformer fuse.
Cause or failure mode — Animal, tree and unknown. If the cause is not known, or does not fall into a predetermined cause category, then the cause code becomes the failure mode. Often, the failure mode is the only information available. Examples are burned, broken, malfunctioning and decayed. Failure modes imply equipment failures, so an equipment code is usually a required entry in a failure mode. Failure modes represent partial information about the cause, in that the root cause may remain unknown, but certain things about the failure are known. For example, if a station breaker fails to trip, then you may have a specific failure mode called “fail to interrupt.” For the equipment code, enter the code for “station breaker.” This system is flexible. The same failure mode can be used with other equipment codes to describe other devices that fail to interrupt. If there is no specific failure mode, then a generic failure mode such as malfunctioning or broken can be used.
Equipment code — To specify the equipment that failed.
Weather — It is useful to know if an equipment failure or unknown interrupting device outage occurred during lightning or other stormy weather. Animal outages normally occur in fair weather. By studying combinations of time-of-day, weather and interruption device activity on a utility, it is often possible to classify unknown outages into likely cause categories. For example, on the Duke Power system, an unknown outage on a tap fuse on a spring morning in fair weather is highly likely to be an animal outage.
Further Reading: “Reliability and Power Quality Improvement Programs,” chapter five of Distribution Reliability and Power Quality, by T.A. Short and L. Taylor (published by CRC Press, 2006).