A Line-Selection Method of Data Fusion Based on Neural Network

In distribution network, the single-phase ground fault accounted for 80% of all distribution networks’ fault. How accurately, rapidly to diagnosis faults lines are practical issues that electrical engineers often meet with. The current of small-current system is so little that could be easily affected by the environment changes and the influence parameters of itself. The single method of line selection cannot conform to the requirements. In this paper, the fifth harmonic of zero-sequence current, line's power and energy of the zero-sequence current are used as the fault feature. Based on the three characteristics and BP neural network, the multi-information line selection was operation. The results of simulation indicate that the multi-information line selection method is much better than the single line selection methods.