The Algorithm Reseach of Detecting Single-Phase-to-Ground Fault Line on Distribution Network with Distributed Generation

This paper studied the zero-sequence current of the distribution network with DGs. By comparison analysis, DGs do not affect the zero-sequence current measured at the initial points of each feeder. The transient component of zero-sequence current contains abundant fault information. The algorithm decomposed zero-sequence current on four scales with wavelet function db4. Fault characteristic components extracted were used to train BP neural network. At last, this paper built a 10kV model with DGs by MATLAB/Simulink, having four feeders. The results indicate that, the algorithm based on wavelet packet transform and BP neural network has a good practicability and reliability, which is not affected by the neutral grounding, grounding resistance and faults occurring time.