Aiming at the issue of fault prediction technique of power electronic circuits’ a method based on characteristic parameter data and particle swarm optimization nonhomogenous grey model (PSONGM) for the prediction of power electronic circuits was proposed. The proposed PSONGM model can predict the development trend of fault feature parameters prediction, and judge the reliability life. Based on the buckboost circuit experiment, the new method can trace the characteristic parameters’ trend and can be effectively applied in fault prediction of power electronic circuits.