GaN Device Modelling Using Machine Learning Approach for Power Electronic Converters

This paper introduces a deep learning-based approach for modelling various GaN -based power electronic devices, crucial for designing and simulating diverse power electronic circuits and conducting performance analysis. The deep learning model is trained using different voltage, current levels, gate resistances, and parasitic inductances. The training dataset is obtained by simulating a realistic Double Pulse Test (DPT) circuit, considering parasitic elements. A many to many Long Short-Term Memory based deep learning model is designed to effectively capture complex switching behavior of GaN devices. This model enables efficient analysis of switching voltage and current, along with reliability assessment of GaN-based power converters. The accuracy of the proposed model is verified by comparing the LTspice simulation results.