An Effective Approach of AC Signal Detection in DC Power System Based on Wavelet Neural Network

This paper develops and experimentally demonstrates an AC signal detection method in DC power system with a combination of a novel detection circuit and the wavelet neural network method. Aims at dealing with the travelling wave and fault signal cannot be detected accurately with the uncertainty of signal velocity when AC signal injected. In which, the injected AC signal in DC power system is detected via the current transformers and the voltage transformers distributed in different current loops. The acquired signal is taken as original fault signal while the sub-band energy function of wavelet packet decomposition is used as secondary characteristic and the minimum distance is used as the criterion in WNN method. Simulations are carried out to demonstrate the correctness and validity of the proposed method and which gives a good consistency with experimental test. Results show that the AC signals can be detected and located in DC power system accurately with the proposed method applied.