The junction temperature of SiC MOSFET is a critical parameter for evaluating its health and performance. However, the current techniques, which are influenced by the degradation of the gate oxide, make it difficult to evaluate the junction temperature accurately. Therefore, a novel method is presented in this article to estimate the junction temperature of SiC MOSFET. The SiC MOSFET’s on-state drain-source voltage(VDS-on), threshold voltage(VTH), and turn-on transient dId/dt(dId/dt) are combined as parameters inputting into a backpropagation neural network(BPNN). Based on the BPNN, this method can estimate the SiC MOSFET’s junction temperature and decouple the influence of gate oxide degradation. The final experimental results indicate the estimation error is approximately 4.02 ℃ and demonstrate this method has a high accuracy.