Fault diagnosis of power electronic circuits using artificial neural networks

A new approach for diagnosing single open circuit faults in the inverter of the advanced static var generator (ASVG) using the feed forward neural network is presented. Single open circuit faults are characterized by distinctive patterns of the output voltage waveform, which is divided into four sections in a period. This distinction appears on abnormal value in some section in a period of the output voltage waveform, it can be recognized by using the feed forward neural network. The samples, with which the feed forward neural network can be trained, are acquired by the power electronic circuit simulator (PSIM). Successful results have been obtained from tests of the feed forward neural networks on fault samples.