A Novel Non-intrusive Arc Fault Detection Method for Low-Voltage Customers

Researches show that serial arc fault current waveform holds unique information of downstream appliance loads. This paper proposes a novel arc fault detection method for low-voltage customers. Besides arc fault detection, it also points out the circuit branch on which the arc fault occurs. By collecting current samples of electrical appliances both in normal operation and under arc fault conditions, we construct a sample feature set of appliance load current waveforms. We propose an arc fault detection system framework combining with supervised nonintrusive load monitoring (NILM) technologies for low voltage customers, which can help to eliminate the fire and safety hazards of the arc fault timely. First, a NILM module identifies the load appliances in operation in customer premise while an arc fault detection module screens the total current signal for arc at the same time. Second, in case of arc fault detected, an optimization module will find the best fit arc fault waveform of one specific appliance by aggregating it to the normal operating waveforms of other appliances to match the total current waveform from which the customer can be alerted with the arc resided branch (the circuit supplies the associated appliance). We build an arc fault simulation device according to the UL1699 standard and carry out lab experiments for low-voltage customer scenarios. A proof of concept example is presented to demonstrate the effectiveness of the proposed method.