SiC MOSFET Neural Network Model And Application Research

With the development of science and technology,aerospace,automotive electronics,smart grid,and other fields urgently need a power device that can still work normally under high radiation,high power density,high voltage and high frequency environments.Among the existing semiconductor materials,silicon carbide device,a new type of wide band gap semiconductor power device,has attracted the attention of researchers with its excellent physical and electrical properties,and it is the most promising power device.Aiming at the needs of new silicon carbide MOSFET model establishment and circuit simulation design,this thesis proposes a simplified model of silicon carbide MOSFEET based on neural network multilayer forward perceptron.The rapid modeling method of silicon carbide MOSFET based on Spice language and neural network algorithm is studied.The error between the simulation results and the experimental results is less than 2.2%,which verifies the accuracy of the Si C MOSFET model.Based on the established model,the driving circuit is designed for the dynamic characteristics of the Si C MOSFET,and a dual-pulse test(DPT)platform is constructed to analyze the key parameters of the driving circuit and the influence of the loop parasitic parameters on the switching characteristics.The experimental results and simulation results have a great agreement,the maixim error is 8%,which indicates the neural network model described in this thesis could describe the transient characteristics of the device accurately.The Buck converter based on Si C MOSFET has been simulated and designed.Through the comparative analysis of converter losses based on silicon and silicon carbide MOSFETs,it is shown that the application of silicon carbide power devices can effectively improve the efficiency of Buck converters.