Power Quality Disturbance Detection Based on Improved Robust Random Cut Forest

Many distribution network disturbances exhibit unique electrical signatures which can be observed from voltage and current waveforms. With the continuous enhancement of Power Quality Data (PQD) acquisition capabilities, it is feasible to continuously monitor the operating status of power nodes from a more detailed perspective. In the past decade, an increasing amount of high-quality PQD has been collected accumulating a massive and unique high-resolution power grid data asset. However, how to obtain valuable information, such as Power Quality Disturbance Events (PQDEs) from massive PQD remains challenging in the research community. In this paper, a reliable PQDEs detection method is proposed based on the improved Robust Random Cut Forest (RRCF). This method achieves accurate detection of PQDEs through the adaptive improvement of pre-filtering based on Ensemble Empirical Mode Decomposition (EEMD) and redundant interpolation. Numerical test results on the synthetic PQD and the realistic pollution experiment of a silicone rubber insulator in a salt fog chamber demonstrate the reliability, efficiency, and scalability of the proposed approach in practical online detection.