Recognition of power quality events using wavelet-based dynamic structural neural networks

Recognition of power-quality (PQ) events by analyzing the voltage and current waveform disturbances is a very important task for the power system monitoring. In this work, a hybrid of wavelet transformation and dynamic structural neural network (DSNN) approach is introduced for PQ events classification. The PQ waveform is first decomposed by four level Daubechies-8 wavelet analysis, and the decomposed waveforms then be processed by the DSNN for PQ event classification. By utilizing the DSNN, the PQ event recognition system can be implemented with minimum neurons and produces maximum performance. Moreover, the proposed method can adapt more training patterns without manually revising the neural network system. The proposed approach is implemented in a simulation program by C++ language to verify the validation and classification accuracy.