Transformer Health Analysis Using AMS Data

Presented by Steven Dennis and Charles Douglas III May 04, 2017 SWEDE 2017 Conference

~ 630 miles

~ 390 miles

Texas’ largest regulated transmission and distribution utility – 6th largest in the U.S.

117,000 miles of transmission and distribution lines

54,000 square miles of territory

3.4 million meters

Oncor Electric Delivery

Motivations For AMS Data Analysis

Electric utilities have recognized that analyzing voltage and loading data associated with transformers can help detect the likelihood of transformer failure. Analyzing Advanced Metering System (AMS) historical data allows us to develop insights into transformer health and, subsequently, to take proactive measures to plan the remedy instead of reacting to an outage after a failure.

AMS Data Analysis Use Cases

Identification of transformer winding damage issues • Proactive replacement • Employee safety e.g., daylight vs. dark in back yards • Scheduled replacement for added customer convenience • Improved service quality – reduced outage time Identification of overloaded transformers • Proactive load management • Labor load leveling • Employee safety e.g., daylight vs. dark in back yards • Scheduled replacement for added customer

convenience • Reduced environmental impact Identification of connectivity and phasing issues • Increase accuracy in system modeling for predictive

analysis

Data Sources Advanced Metering System (AMS) • KWH

• 15 minute data (5 year availability) • Instantaneous voltage

• 2 samples per day that are spaced to approximate the peak and valley of the daily load cycle

• Average voltage • 15 minute interval • About three million meters

Electrical Model

Proof of Concept for Shorted Coil Transformers - March 2016

Process • Sent out a spreadsheet to service centers with identified

locations • Issues were validated by field technicians and the

transformers were replaced • Verified that the voltage from AMS returned to a normal

range • Damaged transformers tested at the Oncor repair facility

• Transformer Turns Ratio (TTR) • Visual inspection • Arrester testing

• Break over voltage • Current leakage

Proof of Concept Results

205 transformers identified through the process and were validated

8 regulation-related issues addressed

• 84 transformers affected by the regulation issues

121 transformers tested and retired at the Oncor transformer shop Every transformer identified and tested had damaged coils

• Proof that the identification process works

Transformer Voltage Profiles

Coil Damage

Regulation

Connectivity Identification Example

Transformer A

Transformer B

Meter 1

Meter 3 Meter 2 Meter 1

Meter 3 Meter 2

Phasing Issue Identification

Lightning Analysis – Overvoltage Began Sept 24, 2016

Voltage Analysis Tool

Intranet Based Field Interface

Status Update for AMS Exceptions

This automated email is sent out to all service centers once a week

The top chart is all unresolved issues by service center

The bottom chart shows all work that was completed in the past week

Daily Targeted Email Notification

An automated email sent on a daily basis to the responsible service center anytime there is a new issue waiting for resolution in the interface

Project Results as of April 18th, 2017

Resolved Issues

Issue Year to Date Project to Date Damaged Winding Transformers 131 557 Regulation (Issue Count) 133 268

Benefit Realized

Customer Interruption Minutes (CIM) Avoided

112,360 399,885

Overloaded Transformer Analytics Using AMS Data

When a transformer is connected to a load that exceeds its KVA rating for a sustained period of time, excessive heat is generated in the windings.

Sustained overloading has an accumulative destructive effect on the integrity of the insulating fluid, winding and its insulation.

Once the insulating properties of the windings are compromised, the transformer will ultimately fail with an open or shorted coil.

Our goal here is to use AMS data to predict the end of life for a transformer and proactively address the overload condition before it fails.

Overload Project Target Selection Criteria

Single phase 120/240V secondary buses

Residential meters

Used daily kWh to select initial targets

Mean to peak kWh coefficient

Low-voltage correlation to overloaded transformers

Targets taken from 140-200% secondary bus kVA capacity

Low Voltage to Loading Correlation

Overload Project – March 2017

101 transformer buses were selected from the target population zone • Between 140 and 200% projected peak load

Intranet interface was developed and rolled out for communicating issues and receiving status feedback from the field • Transformer size validation • Meter connectivity validation • Connectivity corrective information (where applicable)

If the bus connectivity was valid (field matches the electrical model) the overload issue was proactively resolved • Upsized the transformer • Split up secondary bus and added an additional transformer

Typical Reactive Transformer Failure Rate by Month

6.49%

6.51%

8.02%

8.09%

10.60%

11.18%

11.99%

10.74%

8.07%

6.41%

5.48%

6.42%

Dec

Nov

Oct

Sep

Aug

Jul

Jun

May

Apr

Mar

Feb

Jan

Thre

e Ye

ar A

vera

ge

One Year Load Chart Using 15 Minute Load Data

Next Steps

Reduce connectivity issues sent to operations through the use of statistical voltage profile validation

Utilize historical 15-minute average kWh data to validate targets after the voltage profile analysis has validated the likelihood of correct model connectivity

Calculate top oil, hot spot temp., and transformer loss of life to further target near term failures • IEEE C57.91-2011 formula

Final Thoughts

Access to AMS data opens up a new window of opportunity for the discovery of efficiency improvements – most may not have even been envisioned at this point in time

The value of AMS is difficult to extract without connecting it with the electrical model to form a complete picture and this is where proficiency in data mining skills coupled with model knowledge is essential

Model connectivity errors is an industry-wide issue that is critical for the accurate analysis of AMS data but, ironically, AMS data also offers a solution for identifying connectivity issues

Questions?