Intelligent Power Device Modeling using Optimal Deep Learning Approach for AlGaN/GaN HEMTs

AlGaN/GaN high-electron-mobility transistors (HEMTs) are the most recent sort of power devices leveraging the characteristic features of gallium nitride (GaN) and its aluminum gallium nitride (AlGaN) alloy. This device is well-known for its robustness in harsh environments, high-power capabilities, and efficient operation at high frequencies. Modeling AlGaN/GaN HEMTs includes comprehensive simulation to capture the behavior of different operational conditions, which address features such as reliability, carrier transport, gate control, and thermal effects. Accurate modeling is vital for enhancing design efficiency, optimizing device performance, and advancing their integration in RF applications and next-generation power electronics. Lately, deep learning (DL) has become an effective strategy for modelling AlGaN/GaN HEMTs, especially in the field of power devices. By using complex neural network architectures and vast datasets, deep learning (DL) approaches can efficiently capture the complicated relationships between material properties, device structure, performance metrics, and operating conditions. This manuscript develops an Intelligent Power Device Modeling using an Optimal Deep Learning Approach (IPDM-ODLA) for AlGaN/GaN HEMTs. The presented IPDM-ODLA technique aims to properly anticipate the switching behaviors of the devices without delving into the physics and geometry of the devices. In the IPDM-ODLA technique, an initial stage of power device modelling using bidirectional long short-term memory (BiLSTM) model is used. For enhancing the performance of the BiLSTM model, the IPDM-ODLA technique applies grasshopper optimization algorithm (GOA) for hyperparameter tuning process. The experimental validation of the IPDM-ODLA technique can be examined using distinct aspects. The simulation results pointed out that the IPDM-ODLA technique achieves better performance than other models.