Junction temperature estimation of a SiC MOSFET module for 800V high-voltage application in electric vehicles

Silicon Carbide (SiC) power devices have significant advantages on power density and energy efficiency, and are widely accepted as promising solutions for future electric vehicles (EVs) with high-voltage fast charging systems. However, there still exist some noteworthy thermal problems to be addressed. High-efficiency heat dissipation and junction temperature management are determinant factors for successful application of SiC devices. Due to the high integrating level and thermal coupling of multiple chips, a practical estimating method of device junction temperature is especially necessary for SiC power modules. In this paper, a junction temperature estimating method is proposed. Firstly, the SiC MOSFET's power losses during inverter operation are analyzed and modeled. Secondly, internal structure of a power module EAB450M12XM3 is introduced and a finite element analysis (FEA) model is constructed. Based on the models, temperature field and heat dissipation features of the SiC device are investigated. Subsequently, an innovative junction temperature estimating method is proposed based on the methodology of digital twin (DT) and neural network (NN). Both feedforward NN and compensation mechanism with thermistor signal feedback are designed and implemented. Accuracy of the NN-based DT is validated via benchmark test of a prototype motor controller with the studied power module. Real-time computational burden of the NN-based DT is also studied with a 32-bit dual-core micro control unit (MCU), which confirms the practicability of the proposed method.