A remaining useful life prediction method of IGBT based on online status data

Abstract

Power electronic devices are very important component of power processing circuits. However, sometimes under circuit overstress, they may face abrupt failures. Lifetime prediction is needed to prevent these sudden failures in power devices. However, the random noise and errors in the measurement data make the existing methods have large prediction errors. This paper proposes a fusion method based on Least Squares Support Vector Machines (LSSVM)-Particle Filter (PF) that can accurately and stably predict the Remaining Useful Life (RUL) of Insulated gate bipolar transistor (IGBT). First, the method uses the LSSVM model to extract the degraded non-linear feature. Then, the linear regression model is used to extract the degraded linear features. Finally, the PF algorithm is used to fuse the two features to obtain more accurate prediction results and uncertainty expression. The method of feature extraction and fusion is used to effectively eliminate the interference of random noise and errors, so it has more accurate and stable prediction results. The online aging data of the IGBT is used to verify the algorithm, and the results prove that the algorithm can more accurately and stably predict the status or life of IGBT. This method provides a new perspective to solve the problem of life prediction.