Incipient Fault Detection in Distribution Systems Using Line Sensor Data: A Real-World Study

Incipient faults lead to very short voltage/current disturbances that may not be detected by protective devices, and do not result in a line tripping. However, detecting incipient faults in distribution systems is crucial due to their potential to escalate into permanent faults, equipment failures and potentially even fires. This paper proposes a novel detection approach for incipient faults based on event-triggered processing of current and voltage waveforms reported by medium-voltage (MV) line sensors. First, a window of waveform data is fed to enhanced phase-locked loop (EPLL) that estimates the magnitude response and then search algorithms find the transient boundaries that determine the feature interval. Incipient faults are distinguished from other events by analyzing the post-descent magnitude difference along with the maximum current and minimum voltage values within the feature interval. Effectiveness of the proposed method is demonstrated using field data measured by line sensors in a real-world distribution feeder. The experimental results corroborate that under-ground cable faults are properly detected and distinguished from other types of faults.