Non-intrusive Series Arc Fault Detection Based on V-I Trajectory Features

Abstract

The arc fault is one of the main causes of electrical fires. Different from the parallel arc, the amplitude of the series arc fault current tends to be close to that of the normal load current, so it is difficult to be detected. Most of the existing methods are mainly designed for the simple single appliance scenarios with detection equipment installed in or embedded in the appliance sockets intrusively, which are uneconomical and difficult to implement. To this end, the non-intrusive series arc fault detection method is studied in this paper. Firstly, the experimental platform of arc fault is built, where the physical arc fault generator and real appliances are connected to simulate the series arc faults. When multiple appliances operate in parallel, the aggregated current and terminal voltage data samples in the power supply entrance under different scenarios are collected. Secondly, the V-I trajectory is introduced for series arc fault detection for the first time, which could make full use of current and voltage waveform characteristics, and a new arc feature extraction method is proposed based on the asymmetry of adjacent waveforms’ V-I trajectory of series arc fault, making the arc characteristics more distinguished in the aggregated current. Then, integrating multiple V-I trajectory features, a non-intrusive series arc detection method based on the support vector machine model is proposed. The experimental results on the real measured dataset show the detection accuracy for different scenarios is above 93%, which verifies the feasibility of our proposed method.