Research On Fault Prediction Of The Power Electronics Based On Models

Power electronic circuits are strong nonlinear circuit and system which involves very richnonlinear phenomena. Therefore, the working stability of power electronic circuits get a greatchallenge. After nearly50years of development, the structure of the power electronic equipmentis becoming more and more complex, the function is more and more perfect, and the degree ofautomation is more and more high. Now, the electronic equipment has been widely used inpeople’s life, the traditional industry and high-tech industry, etc. Once the device becomemalfunction, not only will it cause great economic loss, but also may endanger personal safety,cause environmental pollution, and bring serious social problem. The fault prediction of the powerelectronics devices can help to monitor the status and the trend of the system fault, realize themaintenance in advance and avoid major accident. So the fault prediction technique of electronicequipments has the vital significance for improving the reliability and safety of the system.The paper presents as:(1) The research of the Buck circuit fault model. Learn about the bond graph theory ofmodeling, analyze the failure mechanism of power electronic circuit key devices and establish thefailure model of the equivalent circuit model. At last, the bond graph model is set up combinedwith Buck circuit working mechanism and output characteristics, each device failure model andthe bond graph modeling theory. The output voltage and output voltage ripple of the circuit wereidentified as the characteristic parameters and extracted in Buck circuit which worked incomponents failure mode.(2) The research of extraction algorithm of capacitor characteristic parameters. According tothe structure of capacitor and failure mechanism, the capacitance equivalent circuit model isestablished. The FFT frequency domain analysis method, Kalman filtering method and leastsquare algorithm were introduced and used in the electrolytic capacitor characteristic parametersextraction experiment in Buck circuit with bond graph simulation data.(3) The research of the power electronic circuit fault prediction method based on theimproved neural network algorithm. The neural network, Genetic algorithm and grey systemtheory were studied and, based on these, improved neural network models were put forward. Thethree models were used to predict the characteristic parameters of circuit level and componentlevel in Buck circuit to verify the feasibility and effectiveness of the models.