Neural network model-based training algorithm for transient signal analysis

The quality of electricity supply has become a major concern of electric utilities in power system. In order to acquire the improved power quality, it is required to know the sources of power quality disturbances and find ways to mitigate them. This paper presents a new approach to detect and classify power quality disturbances based on wavelet transformation and neural network. By means of wavelet analysis, the properties of time-frequency domain demonstrate that the complex wavelet transformation is an excellent tool for processing the transient signal of power system disturbances. The feature extraction of disturbance is explored to obtain dynamic parameters, in which the combined information can be obtained from both magnitude and argument coefficients to extract the desired band of the transient signal and detect the disturbance source. The improved training algorithm is utilized to complete the neural network parameters initialization, acquiring good convergence. The simulation results and analysis demonstrate that the proposed method is effective for transient signal feature extraction.