Monitoring and Detection of Low-current High-Impedance Faults in Distribution Networks

TL;DR

This paper presents a new sensing and deep learning-based detection system for low-current high-impedance faults in distribution networks, addressing a critical challenge in monitoring extensive unmonitored power systems.

Contribution

Developed a physical test facility, characterized fault behaviors, and implemented a pole-mounted sensing system with deep learning for HIF detection in distribution networks.

Findings

Deep learning classifier accurately recognizes fault types

Successful outdoor pole-mounted sensing system deployment

Characterized unique fault signatures for tree-branch HIFs

Abstract

Faults in electricity distribution networks have the potential to ignite fires, cause electrocution, and damage the system itself. High current Low Impedance Faults (LIF) are typically detected and mitigated via over-current, distance, directional relays, fuses, etc. In contrast, while High Impedance Faults (HIF) are equally hazardous, they are much more challenging to detect due to the fault current being much lower than load currents and their time-varying and nonlinear behaviour. Moreover, New Zealand distribution networks are extensive and largely unmonitored beyond the substation, and suitable HIF detection schemes are still an ongoing research challenge. To date, we have built a physical test facility for power system fault analysis and developing and evaluating our sensing and fault detection system. We have simulated LIF and HIF with different fault surface materials and load-switching events. From the data collected, we have characterized the unique fault behaviour for both LIF and HIF in 400V networks and trained a Deep Learning classifier to recognize the type of fault present from its unique signature. We have developed an outdoor pole mountable sensing system and have installed this in Wellington Electricity's network for ongoing data collection and evaluation. This paper will describe the test facility and our experience developing and implementing the sensing system. The widest range of HIF phenomena observed was in the fault experiments involving the tree branch. For brevity, therefore, this paper reports on the results of just these tree-branch experiments. HIF faults on other surface materials will be reported elsewhere. Finally, we will detail the pole-mountable sensing system installed in Wellington Electricity's network and the outcomes thus far.