Abstract: In order to study the parasitic parameters of SiC MOSFET power devices comprehensively, a variable temperature parameter model of silicon carbide MOSFET is proposed. The parasitic inductance, internal parasitic capacitance, parasitic diode and other parasitic parameters are considered in this model. According to the different stages of the switching process of the device, the conductive channel of power device is equivalent to different circuit models. In order to obtain the required parameters in the proposed model, a test platform based on Agilent B1505A power semiconductor analyzer and vector network analyzer (VNA) was built to measure the static characteristic parameters of SiC MOSFET power devices comprehensively. Then an improved BP artificial neural network based on Bayesian regularized LM algorithm is used to fit the measured nonlinear data. The results of fitting show that this method has higher accuracy and generalization ability than the traditional fitting methods, which provides an important basis for the modeling and parameter analysis of SiC MOSFET power devices.