Accurately detecting and quickly mitigating high-impedance (HiZ) faults has long been an industry-wide challenge. HiZ faults can be caused by trees falling on powerlines, downed powerlines and arcing through a failed insulator on a power pole. When these HiZ faults are undetected, these can be public safety hazards and may lead to arcing resulting in initiating wildfires,
With small fault current traits, these HiZ faults are similar in nature to normal loads on the grid. Unlike typical electric (short-circuit) faults that are generally cleared by protective devices (like circuit breakers or reclosers) within a few power cycles due to their high-fault current magnitude, HiZ faults are hard to detect by traditional protective devices.
The characteristics of HiZ faults depend on a variety of conditions, adding further complexity and difficulty in detecting them. Specific tree variety, utility grounding practice, soil conditions, humidity, power system topology, system voltage, weather conditions and load type impact the fault current and characteristics. For example, vegetation-related faults are known to develop gradually over time as the material slowly chars until it suddenly ignites.
As a result, utilities have resorted to public safety power shutoffs (PSPS), or temporary outages, to reduce risk. By turning the power off in an area during high-risk conditions, utilities can eliminate the risk of the HiZ faults during the outage, but that comes with its own challenges.
Military bases and critical infrastructure are also at risk of outages and wildfires. Military installations are dependent on uniquely long stretches of electrical lines in remote places. So, it is highly critical to be able to quickly identify these HiZ faults and de-energize powerlines to mitigate the risk of spreading wildfires.
Government, industry and research labs teamed up
Eaton, the U.S. Army Corps of Engineers and the National Renewable Energy Laboratory (NREL) teamed up to develop a data-driven HiZ fault detection technology that uses sensing technology and advanced machine learning algorithms.
The team conducted hundreds of experiments and performed simulations to develop an understanding of the electrical signature of HiZ faults.
Extensive tests were performed at Eaton’s Thomas A. Edison test center in Franksville, Wisconsin, which is one of only five short-circuit distribution class test labs in North America. The Eaton lab has high-power and high-voltage labs equipped with 8.3 kilovolt (kV) and 25 kV power feeders, along with a 500 megavolt-amperes (MVA) motor generator set. It is an accredited lab for electrical testing with special capabilities in high-fault current and load testing.
The extensive tests included downed-conductor events on different surfaces such as concrete, dry grass, asphalt, sand and more, with varying levels of moisture and other external conditions. Additionally, the tests were performed with tree contact to live powerlines using various tree species found in North America; these tests were vital to analyze unique fault characteristics that are specific to tree variety.
Eaton used the data collected from the testing and simulations to develop HiZ Protect technology, a novel, AI-based solution designed to detect and de-energize powerlines during these events using common edge devices on the grid, such as Form 7 recloser control. In lab-emulated tests at the high-power lab in Franksville, Wisconsin, the novel analytics are detecting HiZ faults with greater than 90% accuracy.
Now, the team is in the next stage of validating this wildfire mitigation strategy. Working closely with multiple North American utilities, Eaton is piloting the HiZ Protect technology on the grid to obtain more data and better understand the nuances of wide-scale deployment. Across these pilot projects, the HiZ Protect technology is being deployed in distribution grid using edge device without communication capabilities because many powerlines are in remote locations with extremely limited connectivity. To reduce complexity, the technology is running on low-cost, processing and memory-constrained hardware without needing communication infrastructure.
These utility pilots are providing the opportunity to further validate this solution and gain a deeper understanding of the corner cases (like false positives) to help refine the technology. Once the utility pilots are completed, the HiZ Protect technology will be commercialized and available to market for utilities, aiming to solve a longtime challenge faced by electric utilities and help significantly reduce the risk of wildfires.
AI-based technology
The AI-based HiZ Protect technology currently being tested in the field is composed of three primary novel elements: integrated sensing, machine learning and edge-based implementation.
The fault detection algorithm is built on the latest advancements in machine learning technology that have recently solved long-standing challenges in AI and cognitive systems. The success of these approaches relies on a rich set of data to develop, train and validate the machine learning models. Therefore, Eaton and its project partners created a comprehensive library of the signature of HiZ fault patterns by leveraging Eaton laboratories and test facilities, NREL’s grid simulation capabilities and field data from multiple utilities in North America.
The HiZ Protect technology holds promise and versatility for utility and military applications, as it can be integrated into grid-edge monitoring and control devices with access to high-fidelity data. Currently, most utilities rely on traditional threshold-based protection, which lacks the necessary observability and protection capabilities to enable swift action by grid operators. Once this technology with HiZ signature recognition and proactive protection solution is commercialized, it can be implemented across existing edge platforms—such as recloser controls (currently implemented in Form 7 recloser control), line sensors and capacitor bank controls—to significantly enhance grid visibility and bolster wildfire prevention efforts.
