A thousand customers just lost power because a bushing failed out on a feeder. Did it just happen or were there early warning signs? Could the failure have been predicted in advance or, better yet, prevented? Pickwick Electric Cooperative (PEC) and Arizona Public Service (APS) have been working with a new technology that enables them, for the first time, to avoid faults by detecting incipient problems and responding proactively.
The new technology, known as distribution fault anticipation (DFA) technology, works by measuring high-fidelity current transformer (CT) and potential transformer (PT) waveforms, typically at the substation, and applying sophisticated analytics to those waveforms. It detects failures, incipient failures and other misoperations out on the feeder, thus providing situational intelligence and enabling feeder-level condition-based maintenance. It does so without complicated setup and without requiring communication with downstream line devices.
Waveform-based analytics represent a new paradigm in distribution system operations and health monitoring. Utilities historically have had little situational intelligence regarding the health of their distribution systems. Modern smart components such as advanced metering infrastructure and distribution automation systems may provide feeder loading levels or let the utility determine whether particular customers have service, but they do little, if anything, to detect feeder anomalies or assess line health.
A breaker locked out an APS feeder — for a fault past a recloser that should have sectionalized the faulted segment without breaker involvement. APS notes such improper operations and performs root-cause investigations. Investigations require multiple sources of information and labor-intensive analysis. This includes downloading records from field and substation devices, manual analysis and correlation of those records, review of coordination settings, operational testing of the recloser and the breaker/relay, and possibly other steps. Some investigations identify the root cause but others conclude with no cause identified.
In the subject case, online DFA waveform analytics saved substantial manpower by automatically identifying the root cause within minutes of the event. The cause was diagnosed as conductor slap, a phenomenon that occurs when fault current induces magnetic forces in upstream conductors, causing them to slap together. This creates a second fault upstream of the first and necessitates operation of upstream protection, in this case the breaker.
After learning the root cause, APS used analytics-derived parameters to locate the offending span, where it found conductors with bright spots and pitting consistent with recent arcing. A traditional investigation would have focused on identifying a defect in the protection system when, in fact, the root cause had nothing to do with the protection system, but rather a problem with the physical span characteristics. APS does not believe that, in the subject case, a conventional investigation would have identified the true root cause.
Field research has documented that fault-induced conductor slap does not occur at random locations, but rather it recurs over time in spans whose construction is susceptible to the phenomenon. This makes it important to diagnose conductor-slap incidents correctly. Each incident causes one or more unnecessary interruptions and a possible outage, often on a feeder-wide basis. Slap-induced arcing causes progressive conductor damage, which, in extreme cases, results in broken conductors. Each slap episode also can throw off particles that might start a fire. Knowing a span's susceptibility to slap enables the utility to take corrective action to avoid future events.
Many utilities, including PEC and APS, have supervisory control and data acquisition (SCADA) systems that tell them when their substation-based feeder breakers operate. Downstream of the breaker, however, sectionalizing reclosers operate autonomously, often without the utility being aware of individual operations. PEC and APS both have long rural feeders with 10, 20 or even more reclosers. Many are hydraulic or, even if electronic, do not have communications installed.
From the substation, DFA analytics detect and report recloser operations in great detail. Knowing details of a measured recloser operating sequence and the estimated load beyond the recloser often enables a utility to determine which recloser has operated, even on a feeder with many reclosers.
It is obvious this provides the ability to know when unsupervised line reclosers operate. What may be less obvious is it also provides an opportunity to assess whether a particular recloser has operated correctly. Using DFA recloser reports, PEC has detected improper operations, such as a recloser that was supposed to lock out after four trips but instead tripped six times before locking out. Conversely, in another case, a field crew observed what appeared to be failure of a recloser to lock out, but a 5-minute analysis of DFA recloser reports showed the recloser was operating correctly. The apparent discrepancy was because 2 to 3 minutes elapsed between operations and the recloser restarted its timing sequence.
Using substation measurements, DFA analytics provide ongoing, real-time information on line recloser operations, enabling the utility to validate proper operations and detect improper operations. This complements and enhances periodic inspection and testing.
Finding Failures Without Outages
PEC and APS have detected and proactively repaired failing apparatus and other conditions that cause intermittent faults:
Cracked transformer bushing
Wind-blown conductors clashing in a long (1,000-ft [305-m]) span
Failing lightning arrester
Service transformer with a hole in its lid
Tree branches bridging conductors or pushing them together.
Prior to failure, these conditions often cause intermittent faults. PEC documented such an example, in which a bushing failed and put 903 customers in the dark. In PEC's case, the failing bushing gave six weeks of early warning, during which time it flashed over on five separate occasions, each time causing an unmonitored recloser to momentarily interrupt 903 customers. Despite 4,515 customer interruptions, no customers complained until a sixth flashover put them in the dark. This failure occurred in the early days of DFA research, before the system worked autonomously. More recently, PEC has preempted similar events and is confident it would have preempted this one if it had today's DFA analytics.
Proactive notification of these failures is providing multiple advantages:
Sustained outages can be avoided, resulting in improved reliability.
Momentary sags and interruptions can be avoided, resulting in improved power quality.
The first two advantages improve customer satisfaction and reliability indices, such as SAIFI and SAIDI.
Reducing the number of faults reduces fault-current stress on transformers, lines, switches and other line components.
Searches and repairs often can be carried out during normal working hours and in fair weather.
Searches can be made while customers' lights are on, instead of during an outage.
Working in daytime, fair-weather conditions results in greater efficiency and improved worker safety as compared to working in the dark or during inclement weather.
DFA analytics have enabled PEC and APS to detect and locate multiple such conditions and make preemptive repairs. One example is a long PEC feeder on which DFA analytics reported an impending failure. PEC used fault parameters, provided by DFA analytics at the substation, to direct the search to a small area of the feeder.
Searching that small area, a crew found a service transformer with a hole punched through its lid. The crew then replaced the transformer, during daytime hours on a fair-weather day, thus avoiding further interruptions, sags, outages, system stresses or other trouble (such as the remote possibility of an exploding transformer).
This is not an isolated example. PEC and APS have used DFA waveform analytics to detect multiple such conditions and make preemptive repairs.
Crews responding to lights-out or flickering-lights calls often receive only vague descriptions of symptoms, supplied by customers. Moreover, some problems are intermittent and may not be manifesting themselves when the crew arrives on the scene. The figure above illustrates a sequence of events that required four crew trips and equipment replacements, all ultimately determined to have had a single hard-to-diagnose clamp failure as their root cause.
DFA waveform analytics had been alarming this clamp failure, intermittently, for three weeks. But, because the DFA project had experimental status, responding crews were unaware of these alarms. As a result, this single clamp failure cost PEC four customer calls, four crew trips (all on overtime) and the change-out of two customer transformers that later tested good. Giving responders analytics-generated diagnoses will reduce incorrect diagnoses, no-cause-found events, customer complaints, return trips and change-outs of healthy apparatus, such as the two transformers in this case.
Light in the Dark
Figuratively speaking, distribution systems largely operate in the dark, with utilities having little visibility into failures, incipient failures and other feeder misoperations. PEC and APS have been working with the new DFA technology to provide newfound situational intelligence, enabling better operational efficiency, and improved reliability and quality of service. It does so with substation-based monitoring, without complicated setup and without requiring sensing, electronics and communications along the feeder. PEC and APS have used this technology to avoid multiple faults on their systems and to efficiently diagnose problems that otherwise would have taken substantially more effort and likely would not have been resolved at all.
Ken Sanford (firstname.lastname@example.org) is senior engineer at Arizona Public Service Co. (APS). He works in construction and operations for the southeast division covering three counties. He also is working on the smart grid technologies team with the innovation/technology solutions department at APS. Sanford graduated from Arizona State University in 1986 with a bachelor's degree in construction engineering.
John S. Bowers (email@example.com) is the vice president of operations at Pickwick Electric Cooperative in Selmer, Tennessee, U.S. He is a 1991 graduate of Tennessee Technological University and holds a BSEE degree. Bowers also is a registered professional engineer in Tennessee. He has been actively involved with the distribution fault anticipation technology since 2002.
How Distribution Fault Anticipation Works
The distribution fault anticipation (DFA) system's building blocks are 19-inch rack-mount DFA devices, mounted in substations on a per-feeder basis. A feeder's DFA device performs high-fidelity digitization of electrical waveforms from that feeder's current transformers and potential transformers. It records even minor anomalies and uses sophisticated analytics to determine the underlying failures or other feeder events.
Configuring a DFA device does not require special programming or entering of feeder maps, protection settings, feeder connectivity or device placements. Neither is it necessary to communicate with line devices. Analytics detect the presence of line devices, such as switched capacitor banks and reclosers, including hydraulics, by analyzing the waveform signatures those devices produce as they operate.
Within each feeder's DFA device, analytics generate reports that are communicated, through the Internet, back to a central server computer. The central server provides web-based access to reports from DFA devices across the system. Device-to-server communication can use digital subscriber line, cell modem, cable modem or radio.
Distribution fault anticipation waveform analytics examine waveform data to detect and characterize recloser operations. For example, the graph plots a 20-second period of current, with an inset showing the analytics-calculated operating sequence. The operating sequence is interpreted as three trips (one fast and two delayed) and a single-phase recloser operation for a phase B fault of 585 A to 591 A. Each trip momentarily interrupted 20% of phase B (only) load. Each open interval was 1.9 seconds to 2.0 seconds.
The report often provides the utility's only notice that a recloser has operated. In addition, the utility often can determine exactly which recloser has operated, even on a feeder with numerous reclosers, by comparing reported operating sequences to system-model information.
Distribution fault anticipation technology was founded on research led by Carl L. Benner (firstname.lastname@example.org) and Dr. B. Don Russell (email@example.com) at Texas A&M University, and largely supported by the Electric Power Research Institute (EPRI). More than 10 EPRI-member utilities participated in early research to identify and correlate waveform fingerprints with specific feeder phenomena. The technology has evolved and now uses online 24/7 waveform analytics to recognize faults, incipient failures and other feeder events. Tennessee Valley Authority (TVA), through EPRI, has been a supporter of these efforts since 2001 and works with Pickwick Electric Cooperative, one of more than 150 TVA distributors, as a host site. Arizona Public Service became involved with DFA on its system in 2011.
Texas A&M maintains a website (https://dfaweb.tamu.edu/DfaReports/DfaSuccess.aspx) that details other examples, illustrating how analytics can improve knowledge of and response to multiple feeder problems, including vegetation faults, capacitor failures and secondary cable failures among others.
Arizona Public Service | www.aps.com
Electric Power Research Institute | www.epri.com
Pickwick Electric Co. | www.pickwick-electric.com
Texas A&M | www.tamu.edu
Tennessee Valley Authority | www.tva.gov