Defect Detection in Power Electronic Circuits by Artificial Neural Network Using Discrete Wavelet Analysis

摘要

Power electronics occupies a major section of industrial drives and systems in terms of power converter and nonlinear circuits for running and controlling three-phase or single-phase machine. Three-phase controlled rectifier and inverter are the most important analog circuit in power electronics. These circuits have also gained immense importance in modern grid-connected system synchronized with renewable energy sources. In this context, it requires maximum attention for smooth operation of these devices at minimum recovery time during faulty condition. And hence detection of faulty component during running condition becomes extremely important. Considering these particulars, this paper presents a proficient defect-oriented parametric test method for two power electronic circuits like three-phase rectifier and inverter based on artificial neural network using discrete wavelet decomposition as preprocessor for feature extraction. Two types of feed forward neural network such as BPMLP and PNN are employed here for fault event detection. Results are found to be very promising with utmost of 99.95%.