Thermal Behavior Prediction of the Silicon Carbide Power Devices Based on Time Series Decomposition

The junction temperature parameters of power devices have an important impact on the device lifetime and the overall reliability of power electronic converters. The impact is mainly reflected in the characteristic parameters such as the fluctuation and mean value of the junction temperature, which directly affect the health status of the devices. The traditional thermal management method is only based on the historical junction temperature data of the power devices, and does not predict and manage the future junction temperature behavior. Therefore, the calculation results cannot cover the future thermal behavior of the power devices. In addition, the historical junction temperature data of the traditional thermal management method is very short, and the big data of junction temperature is not introduced. Therefore, the thermal behavior state reflected has great limitations. This paper innovatively proposes to use the Prophet algorithm based on time series decomposition to calculate the future thermal behavior of the junction temperature of the silicon carbide (SiC) power devices. The algorithm introduces a large database of junction temperature and combines historical junction temperature data with future junction temperature numerical prediction data to accurately predict the lifetime of the SiC power devices. Simulation and experiments prove the effectiveness of the algorithm proposed in this paper.