Real-time Inverter Semiconductor Die Temperature Estimation Using Inverter Operating Condition-based Gate Recurrent Unit

Real-time semiconductor die temperature estimation for switching power semiconductor is essential for power converter thermal management and reliability monitoring. This research proposed a method for real-time power semiconductor die temperature estimation using recurrent neural networks. This paper first investigates the thermal dynamic model of the inverter as a physical model reference. The structure of the recurrent neural network is based on the thermal dynamic model of the inverter. The inverter operating parameters are used as the inputs of the neural network. In this research, the inputs and the outputs of the network will be normalized based on the actual operating range of the inverter. The gate recurrent unit network is trained, cross-validated, and tested with the inverter experimental records. The estimated die temperature are compared with the measured results.