Fault and PQ Event Signature Repository

Lack of specificity in fault identification is a known issue - academic research is constrained without comprehensive fault data for pattern matching and machine learning. A large and well-defined database of fault signatures (directly correlated to wildfire ignition risk) would enable researchers and specific industry sectors to eventually distinguish between types of faults in real time, allowing protective relaying and other sensor enabled hardware to respond precisely to the situation at hand. This is especially important to simultaneously recognize what caused a fault, determine its physical location, and to reduce the amount of ignition energy applied to ground fuels near the fault.

EPRI seeks to support a U.S. National Lab endeavor to update and to enhance a usable fault signature database for research by academic institutions and specific subsets of industry such as relay vendors. The project is envisioned to enable future wildfire prevention algorithms by performing real world fault tests at the EPRI Lenox, High Voltage test facility, and by concurrently capturing relevant high resolution power monitor data for each instance. Once the power data has been properly curated with both video recording and all associated electric power monitor data, these replicated fault incidents would be added to the Grid Event Signature Library (GESL) currently hosted and maintained by Oakridge National Lab. This is an extensive open-access collection of labeled grid disturbance data, which has already gained traction in the power system community with over 400 registered users.

The EPRI contributions would focus on creating real-world faults scenarios on a full-scale electric test circuit to include:

  • Conductor Slap
  • Live Downed Wires
  • Electrical Arcing, and
  • Vegetation Faults

A key attribute of EPRI’s full scale fault test range is the ability to monitor and record the associated electromagnetic events data both upstream and downstream of the fault location and to capture the full data streams for the entire event duration. This is especially important considering some of these events can last up to ten minutes with milliampere current levels while other incidents may only last tens of milliseconds and have current magnitudes in the thousands of amperes. Another key attribute of the test range is the ability to accommodate new and emergent line monitors and data acquisition equipment as identified. Toward the stated approach, the following activities are proposed:

Configure the Test Range and the Data Acquisition Hardware

This task is focused on creating and describing the specific electrical line configurations for each fault type to be replicated. The objective is to create a complete specification that could be used at any high voltage (1KV and above distribution system) test facility to consistently replicate, to electrically document, and to digitally curate any of the four fault event categories from the bulleted list above. The monitoring and data acquisition nodes would be designed to accommodate additional data acquisition equipment from emerging vendors or from the U.S. National Labs. For each fault event category, the developed protocol will include at a minimum:

  • Line and source configurations and test range requirements for the physical design
  • Electrical parameter minimum criteria and documentation criteria
  • Recommended practice for protection, isolation, and duration between sequential tests
  • Recommended monitor/sensor points, monitor options, trigger settings, and sampling criteria

Development of Replicable Test Protocols

After configuring the test range setups, this task defines the step-by-step procedures necessary for conducting the actual faults and for digitally documenting the outcomes for each of the four fault event types.

Once the test protocols have been developed and vetted for completeness, two full test weeks are planned for each of the four fault types as follows:

Vegetation Fault Testing

This effort facilitates two full test weeks (80 hours) with different vegetation samples where the vegetation under test has been prepared and measured to document the known moisture condition, estimated electrical impedance and other relevant condition information. At least four tests for each vegetation test will be repeated to document variabilities before moving on to the next vegetation sample set.

Live Downed Wires Testing

This effort facilitates two full test weeks (80 hours) with different ranges of downed wire lengths and different ground conductivity conditions. Both wire to surface contact area and ground resistance have previously been observed to influence both fault current levels and the amount of intermittent arc energy. Therefore, varying these parameters will provide are good range of useful data. At least four tests for each test condition will be repeated to document variabilities before moving on to the next live downed wire scenario.

Arcing Hardware Testing

This effort facilitates two full test weeks (80 hours) with different examples where electrical arcing from an intermittent connector or an electrical asset in pre-failure condition. EPRI will replicate at least six known conditions where an electrical failure could result in vegetation ignition. At least four tests on the same component will be repeated to document variabilities before moving on to the next sample.

Conductor Slap Testing

This effort facilitates two full test weeks (80 hours) with different overhead line configurations where the wires under test have been designed, installed, and tensioned to optimize the conductor swing and to produce the fault induced conductor slap phenomena. At least ten tests for each unique line design will be repeated to document variabilities before moving on to the next conductor slap test set.

Contributions to the Grid Event Signature Library

For each of the four event types replicated over their respective test intervals, video and digitized data will be collected for each individual test and these will be uploaded to the GESL.

This refresh of the industry fault signature repository with new fault entries and additional types of sensor data will assist considerably in helping researchers develop better algorithms with higher success rates. Another useful addition to the repository is the useful nature of multiple sensors that can capture the same fault event from a different location (upstream, downstream or laterally on the system) and they see a different voltage or current profile depending on where they are relative to the fault location.

Which 2030 Future States are Impacted by this Work?

  • Full fault count/cause visibility and history for all power lines of interest
  • A Unified fault and power quality (anomaly event) signature repository
  • AI enabled monitors and sensors at protective and transition nodes