Presentation to the IEEE Power and Energy Society, Phoenix Chapter Radisson Hotel Phoenix Airport Phoenix, Arizona, March 22, 2012 Carl L. Benner, PE Sr.

Presentation to the

IEEE Power and Energy Society, Phoenix ChapterRadisson Hotel Phoenix Airport

Phoenix, Arizona, March 22, 2012

Carl L. Benner, PESr. Research Engineer

Assistant Director, Power System Automation LaboratoryDepartment of Electrical and Computer Engineering

Texas A&M UniversityCollege Station, TX 77843-3128

(979) 845-6224, [email protected]

Bringing Visibility to Power Distribution System Operations and Engineering

2© 2012 The Texas A&M University System. All rights reserved.

In The Dark

• A typical distribution feeder consists of thousands of components: wires, insulators, capacitors, transformers, …. We generally have no idea of their condition, until the fail.

• This puts utilities in a mostly reactionary posture: wait until something breaks and then fix it (with exceptions, such as inefficient periodic capacitor maintenance).

• In addition, some problems (e.g., “flickering lights”) can be difficult to diagnose, resulting in incorrect diagnoses or “no cause found.”

• DFA technology provides greater visibility, or awareness, of system events and conditions, enabling multiple operational improvements, including proactive condition-based maintenance, accurate diagnosis of problems, etc.


DFA Overview

• Distribution Fault Anticipation (DFA) technology is an advanced, multi-purpose monitoring and diagnostic system that alerts users to faults, incipient failures, and other events and conditions on distribution systems.

• DFA technology has several self-imposed constraints:– Standard CTs and PTs as inputs– Minimal setup – no requirement for system model– No requirement for distributed electronics– No requirement for distributed communications (sub only)– Provide “intelligence,” not just data

• Two DFA-related patents have been granted, and several other applications are in process.


DFA Hardware Installation Photos


DFA Overview (cont’d)

• DFA devices in substations use sensitive triggering to record waveforms, even for events causing only slight changes to electrical waveforms. The devices then process those waveforms using sophisticated algorithms, to create “actionable intelligence.”

• A centralized server retrieves this “actionable intelligence” from the DFA devices and makes it available via web interface or email.

• DFA complements other smart grid technologies, by enabling proactive condition-based maintenance, improved diagnosis and response to events, forensic investigations, etc, by providing a level of operational awareness and visibility not available with other technologies.


Historical Background

• EPRI supported early DFA research at Texas A&M University, starting in 1997.

• The initial goal was to predict, or “anticipate,” future failures of line apparatus.

• Foundational Hypothesis: Line currents and voltages reflect feeder activity, and early-stage apparatus failures affect electrical waveforms in specific, predictable ways. Therefore, continuously monitoring these signals provides the opportunity to detect and diagnose impending failures and other feeder activity.


Historical Background (cont’d)

• Early project efforts collected massive amounts of substation waveform data, by connecting instrumentation, configured with sensitive triggering, to conventional CTs and PTs on 70 utility feeders for multiple years. This produced the largest database ever created of high-fidelity electrical waveforms corresponding to normal and abnormal events on distribution systems.

• Correlating utility-documented apparatus failures with recorded waveforms enabled researchers to identify waveform characteristics and subsequently develop detection algorithms.

• The research process serendipitously demonstrated that applying intelligent algorithms to high-fidelity waveforms can provide considerable value beyond the “anticipation” function originally envisioned.


Current Status of DFA Technology

• Today DFA hardware and analysis algorithms detect voltage and current signatures indicative of multiple types of incipient failures.

• The research process revealed that electrical waveforms reflect a broad range of feeder activity, not just incipient failures.

Intelligent Algorithms (Analytics)

Reliability

Condition-BasedMaintenance

Asset Management

ImprovedSafety

Outage Management

Forensics

Improved Power Quality

O&M Cost Reduction

Fault Anticipation Diagnosis of ProtectionProblems

• Applying on-line signal processing and pattern-matching algorithms to electrical waveforms can provide situational awareness and actionable intelligence for dispatchers, analysts, and repair crews.


Failure Types Documented in DFA Database

• Voltage regulator failure• LTC controller misoperation• Repetitive overcurrent faults• Lightning arrestor failures• Switch and clamp failures• Cable failures

– Main substation cable– URD primary cables– URD secondary cables– Overhead secondary cables

• Tree/vegetation contacts– Contacts with primary– Contacts with secondary

services

• Pole-top xfmr bushing failure• Pole-top xfmr winding failure• URD padmount xfmr failure• Bus capacitor bushing failure• Capacitor problems

– Controller misoperations– Failed capacitor cans– Blown fuses– Switch restrike– Switch sticking– Switch burn-ups– Switch bounce– Pack failure

Certain types of failures occur frequently and are well understood. Other types of failures occur infrequently and few incidents have been documented. The DFA system was designed to accommodate improved algorithms, as ongoing field experience provides additional data.

EPRI

As a follow-on to the letter reports, I would like to get at least one or more of each of these listed items documented with Description of the issue - some waveshapes - sequence of event discussion - and field finding. I note that about five or six are already inlcuded in this ppt and that is an acceptable method of doing the documentation - but perhaps with a few extra waveforms and some animations similar to what you did with a few of APS's events.


Evolution of DFA Algorithms (Example)

Early Algorithm: Generic Arcing Detection

Arcing(capacitor) Arcing

(series)

Arcing(shunt)

Arcing(clamp)

Arcing(line switch)

Arcing(clamp w/low load)

Arcing(clamp w/high load)

Arcing(switch w/low load)

Arcing(switch w/high load)

TIME


System-Level DFA Data Processing Paradigm

• DFA technology currently employs a distributed-processing paradigm, in which individual DFA devices perform the number crunching and deliver pre-analyzed events to a central server.

• This model provides multiple advantages as compared to models in which all devices send waveform data to a central server for processing.– The central server does not have to retrieve waveforms of routine events

(e.g., motors starting, normal capacitor operations). The user can decide which types of events are important and configure the system to retrieve waveforms only for specific types of events.

– Event reports can be delivered without the time delay of waiting for waveform retrieval. This provides a particular advantage where communications are limited.

– Distributed processing promotes scalability for system-wide deployment.

• Current system implementation makes it straightforward to update algorithms on DFA devices as algorithms continue to evolve.


On-Line Signal Processing and

Pattern RecognitionAlgorithms

(Performed byDFA device in

Substation)

Line recloser*tripped 8% ofphase-A load twice,but reclosed and didnot cause outage

Breaker lockout caused by fault-induced conductor slap

Inputs: Substation CT and PT Waveforms

*DFA algorithms analyze high-fidelity substation waveforms, to report hydraulic reclosers, switched line capacitors, apparatus failures, etc, without requiring communications to those devices.

Algorithm OutputsDFA Algorithms

Data Processing by Individual DFA Devices

Failing hot-lineclamp on phase B*

Failed 1200 kVARline capacitor*(phase B inoperable)


Division of Responsibilities between DFA Devices and Central Master Station

Performed by DFA Devices at SubstationsDevices to Master•Algorithm Outputs•Selected Waveforms•Interval Data•Real-time Waveforms(on demand)Master to Devices•Configurable Settings

Central DFA Master Station

- Format Web Reports- Provide Web Interface- Generate Waveform Plots- Format Email Reports- Send Email Reports

Comm Link


System-Level DFA Data Processing Paradigm:An Illustrative Example

• Assume a particular feeder has a failing hotline clamp. A typical failure of this type may have the following characteristics:– It causes waveform variations the DFA system can detect and identify.

– It produces these variations intermittently, typically for periods of a few minutes at a time, interrupted by quiescent periods often lasting for hours or days.

– As measured from the substation, it causes variations on the order of a few primary RMS amperes and almost no primary RMS volts. (This illustrates the need for sensitive triggering.)

– Over a period of days to weeks, it produce dozens, hundreds, or even thousands of individual “events.”

• The DFA device records each waveform anomaly, classifies it as a failing clamp, and labels the “cluster” of events as a failing clamp.

• The master station retrieves from the device the single report of “failing clamp,” and reports this via website and/or email.


Sample DFA Report (via Email)

Incident: Remote fault occurred. Line recloser should have sectionalized feeder, but breaker tripped and locked out feeder. Utility noted apparent miscoordination, prompting need for investigation.DFA response: Minutes later, DFA emailed a report indicating the feeder lockout was caused by fault-induced conductor-slap, not miscoordination.Location: The DFA report also provided fault-current estimate, enabling location of span that slapped and caused lockout.Special note: Police office had reported “pole fire” at site of conductor slap, but utility responder found no fire. Officer apparently had seen shower of sparks from slapping conductors.


Example: Fault-Induced Conductor Slap (FICS) and Predictable Feeder Lockouts

• 11/12/2007 Close Call• As a result of a downstream fault, upstream wires swung

together, resulting in fault-induced conductor slap (FICS). This caused a second fault and necessitated a trip by the substation breaker. Breaker reclosed twice and held, enabling feeder to sectionalize properly.

• DFA uses such episodes to warn of FICS susceptibility and provide information to locate slap.

• 12/02/2007 Breaker Lockout (43 minutes)• FICS occurred in the same span as the 11/12/2007 episode. This

episode caused the substation breaker to lock out.

• DFA uses such episodes to provide renewed warning of FICS susceptibility, including location.

• 11/13/2009 Breaker Lockout (36 minutes)

• 11/18/2009 Close Call

• 12/25/2011 Breaker Lockout (13 minutes)

EPRI

There are very few of us who can recognize what is going on with the current traces so I fear we may loose the audience without adding some circles and balloon style comments to the waveshapes. May requre a few extra slides to make it more clear that the amount of current associated with each event is the distinguishing identifier that - onced combined with some sequence analytics - yields the likely cause(s)Perhaps we should strart with the next slide and then add a tutorial slide called conductor slap - What are we looking for? and use that as the predecessor to this slide - and follow those two slides with this Example one...


Discussion of FICS Detection

• When FICS occurs, faults that should be cleared by sectionalizing devices instead create more widespread disruptions, including feeder lockouts.

• FICS typically occurs in an overhead span with unusual construction.

• DFA research has documented that utilities experiencing FICS mistakenly may believe they have a problem with protection coordination. Having limited information to guide investigations of these presumed protection problems, those investigations often result in “no cause found.”

• DFA reports FICS and provides guidance to find the offending span.

• DFA research demonstrates that spans susceptible to FICS experience it repeatedly. Intervals between episodes may be months or years, however, making it unlikely that personnel will recognize the connection between episodes.

• FICS impacts SAIDI and SAIFI, because each FICS episode tends to interrupt numerous customers for a significant period of time.

• DFA alerts the utility to FICS the first time it detects it electrically. Subsequent FICS interruptions that would have occurred without the DFA’s alert constitute outages anticipated and averted.


Discussion of Recurrent Fault Detection

• Recurrent faults occur when incipient failures cause flashovers that momentary interruptions by sectionalizing reclosers temporarily “heal.” The fault recurs each time environmental or other conditions become conducive to flashover. A single incipient failure may cause many such episodes.

• DFA research has documented multiple causes for recurrent faults, including vegetation, cracked bushings, a “death-grip” bird, and an overlong jumper.

• Left uncorrected, recurrent faults typically evolve, eventually causing sustained outages and possibly other damage.

• DFA research indicates that customers often do not report momentary interruptions. For example, over a six-week period, a failing bushing flashed over five times, each time causing a momentary interruption for 900 customers. No customers complained until a sixth flashover caused a sustained outage.

• DFA analyzes each fault as it occurs. It also examines its historical records and alerts the utility when it detects multiple episodes of what appear to be the same fault, even if multi-day or multi-week intervals separate the episodes.

• DFA enables location of these incipient failures before they cause outages. One utility routinely uses a procedure it developed for finding recurrent faults, and has averted multiple sustained outages, thereby improving SAIDI and SAIFI.


Step 1: Learn of recurrent fault from DFA DFA reported two identical faults, 18 days apart. This circuit has 139 miles of exposure, making

location of this incipient failure challenging.Step 2: Compare DFA info to system model at each line recloser (illustrated here for recloser R)Protection– DFA: Reported operation of 1ø recloser– Model: R is bank of 1ø hydraulic reclosersMomentary Load Interruption– DFA: Estimated 19-21% load interruption– Model: 23% of load is beyond RReclosing Interval– DFA: Reported 2-second open interval– Model: Reclosers at R have 2-second openConclusion: Failure is downstream of recloser R (reduces search area by 74%).

Sub

R

Circuit NS 344(139 circuit miles)

Detailed ExampleRecurrent Faults and Proactive Maintenance


• Comparing DFA information with the feeder model identified which recloser was operating, indicating fault was downstream of that recloser.

• Once we focus downstream of the particular recloser, comparing the DFA fault-current estimate of 510A with the feeder model targeted area outlined in oval (~1% of total circuit miles).

REnlarged View of LineDownstream of R

Sub

NS 344(139 circuit miles)

R

Detailed Example (cont’d)Recurrent Faults and Proactive Maintenance


Sub

NS 344(139 circuit miles)

R

Enlarged View of Oval

Integrating DFA information with system model put failure inside oval area, encompassing 4 spans on either side of a tee (1% of circuit). Searching that area identified a arrestor that was flashing over. Replacing arrestor averted further interruptions and future outage to 53 customers.

Detailed Example (cont’d)Recurrent Faults and Proactive Maintenance


ExampleMisdiagnosed Trouble

• T=0 hours: 16 customers with lights out Blown tap fuse, but no obvious system failure. New fuse holds. Close ticket.

• T=36 hours: Flickering lights in same area Crew hears buzzing transformer. Replace transformer. No buzzing. Close ticket.

• T=38 hours: Lights flickering again

Now what do I do?


• For three weeks, the DFA system had been reporting a failing clamp on this phase of this circuit.

• Without DFA, this single failure“cost” the utility four customercomplaints, four truck rolls, andtwo unnecessary transformerchange-outs.

• Equipping a dispatcher with this information would enable one tripand one clamp replacement.

ExampleMisdiagnosed Trouble


Discussion of Switch/Clamp Failures

• As the jaws of hotline clamps and switches begin to fail, they produce unique arcing signatures, often detectable from substation CT/PT waveforms.

• The incipient-failure signature tends to occur for minutes at a time, interspersed with quiescent intervals lasting multiple days.

• A failing device may exhibit hundreds or even thousands of distinct arcing episodes. The DFA system uses intelligent reporting to provide current status of the incipient failure, without overwhelming personnel with alarms.

• DFA reports incipient failures of both clamps and switches, and it distinguishes between the two types with a high degree of confidence (80%/20%).

• DFA estimates the load past the failing device, enabling the utility to assess whether the failure likely involves a mainline device serving hundreds of customers, versus a device serving a small number of customers.

• Locating failing mainline devices would have high importance and should be relatively straightforward, because few such devices exist on a given feeder.

• DFA enables the utility to know the nature of the trouble, when a customer calls to report outage or flicker. This provides for better diagnosis, fewer unnecessary equipment change-outs, increased crew efficiency, and decreased “no cause found” incidents requiring call-backs.


• A secondary (120/240V) underground cable arced for an extended period and then caused an outage.

• The DFA device at the substation detected the secondary arcing, as reflected to the primary, dozens of times over a 44-minute period.

• Reflecting 60A of measured primary current to the transformer secondary yields 60A x 60:1 turns ratio = 3600A of secondary current.

• Typical bursts of current lasted several hundred milliseconds. Repeated 3600A bursts of secondary current failed to blow the secondary fuse.

• The feeder’s 300+ amps of normal load current renders the fault current “invisible” to conventional monitoring. Detection requires sensitive triggering, such as used by the DFA system.

ExampleSecondary (120/240V) Buried-Cable Arcing


Discussion of Secondary Buried-Cable Arcing

• This recent case is the second example of secondary buried-cable arcing documented by the DFA research program.

• The past case involved recurring secondary-arcing episodes. Some episodes arced for a few hundred milliseconds, but then self-extinguished for days. Others episodes tripped a CSP transformer’s breaker. Each time the breaker tripped, customers called and the utility had to dispatch a crew. When the crew reset the CSP breaker, it held, restoring normal service, so trouble tickets were closed with “no cause found.” This happened multiple times over a three-month period.

• Waveforms from the previous and current cases are being analyzed, to determine the degree to which they have similar, unique characteristics that could serve as the basis for implementing cause-specific reporting.

• Cause-specific reporting makes plausible the idea that, in the future, the DFA may be able to advise the nature of the trouble, when a customer calls to report outage or flicker. This would enable better diagnosis, increase crew efficiency, and decrease the number of “no cause found” incidents requiring subsequent call-backs.


Examples:Other Anomalies

• Case 1: Line burned down past line recloser and auto-sectionalizer. DFA helped determine which device failed.

• Case 2: In multiple cases, utility has used DFA to determine whether questionable reclosers were functioning properly, without removing them for offline testing.

• Case 3: DFA report is helping ongoing investigation to diagnose why a self-healing circuit did not respond properly.

• Case 4: On multiple occasions, DFA has helped a utility identify that a recloser stopped mid-sequence, failing to reclose or lock out, creating hazard for a crew that assumes that an open recloser is locked out.


Summary

• DFA’s fundamental premise is that electrical waveforms reflect feeder activity, and that analysis of these waveforms by sophisticated on-line algorithms can improve the utility’s awareness, and therefore operation, of the system.

• Scope of DFA capabilities and value has expanded greatly beyond original focus on just anticipating failures.

• Utilities can use DFA website to learn of impending failures, misoperating equipment, protection-system anomalies, etc, and thereby make decisions and take corrective action.

• Current and anticipated efforts will examine integrating DFA information with other utility information systems and emailing alerts to utilities.

• DFA’s implementation anticipates and provides for future enhancements, as new waveform signatures are identified and algorithms are implemented.

Presentation to the

IEEE Power and Energy Society, Phoenix ChapterRadisson Hotel Phoenix Airport

Phoenix, Arizona, March 22, 2012

Carl L. Benner, PESr. Research Engineer

Assistant Director, Power System Automation LaboratoryDepartment of Electrical and Computer Engineering

Texas A&M UniversityCollege Station, TX 77843-3128

(979) 845-6224, [email protected]

Bringing Visibility to Power Distribution System Operations and Engineering

Presentation to the IEEE Power and Energy Society, Phoenix Chapter Radisson Hotel Phoenix Airport Phoenix, Arizona, March 22, 2012 Carl L. Benner, PE Sr.

Documents

university system

system events

dfa overview contd dfa

early dfa research

diagnostic system

highfidelity waveforms

engineering slide

dfarelated patents