Research On Fault Prediction Of The Power Electronic Devices Based On Improved Gray System

Along with the development of power electronic technology, power electronics devices get fastdevelopment and be widely used in life applicance, traditional industry and high technology industriesetc. The failure of the power electronic equipments will lead to electric products damage, traffic jamsand the loss of industrial and mining enterprise.What is more,it could cause significant casualties andcalamity accidents. The fault predition of the power electronics devices can help to monitor the statusand the trend of the system fault, realize the maintenance in advance and avoid major accident. So thefault prediction technique of electronic equipments has the vital significance for improving thereliability and safety of the system.The paper presents as:(1) The research of extraction algorithm of capacitor characteristic parameters. Thecharacteristic parameters which reflect the performance of the capacitors is determined through theanalysis of the structure, failure mechanism and the equivalent circuit model of aluminum electrolyticcapacitors. The time domain method, power method, recursive least square method and finithedifference-genetic algorithm mehod were introducted and used in the electrolytic capacitorcharacteristic parameters extraction experiment in Buck-Boost circuit.(2) The research of the Buck-Boost circuit fault model. The mathematical model of the outputcharacteristics is established through the analysis of the Buck-Boost circuit which works incomponents parameters failure situation. The output voltage mean and output voltage ripple of thecircuit were identified as the characteristic parameters and extracted in Buck-Boost circuit whichworked in components failure mode.(3) The research of the power electronic circuit fault prediction method based on the greymodel. The gray system, GM(1,1) model, was studyed and,based on this, the paper puts forword theGAGM and the PSO-NGM fault predicion models. The three models were used to predicit thecharacteristic parameters of circuit level and component level in Buck-Boost circuit to verify thefeasibility and effectiveness of the improved models.