Fault Arc Detection Based on Time and Frequency Domain Analysis and Radom Forest

The Low voltage AC power grid is the most widely used power grid, but also the most day-to-day life-oriented power grid. The damage caused by fault arcs in low voltage AC networks is also incalculable. By using the self-made arc generator to simulate the generation of low voltage fault arcs, the current signals of 8 kinds of fault arcs are obtained. Time domain features such as average and variance of each load current waveform are extracted, and then the current waveform is passed through the Fourier transform to obtain such features as variance frequency ratio in its frequency domain. Through the analysis of decision tree and random forest, random forest is selected as the classification algorithm of fault arc detection, then the accurate load identification and fault identification are carried out by using random forest. A total of 21907 current state information is collected. The results show that the recognition rate of this algorithm is 100% for light bulb and electric fan, and 99% for other electric apparatus. Experimental results show the effectiveness of the proposed detection algorithm.