On Temperature-Dependent Small-Signal Behavioral Modelling of GaN HEMT Using GWO-PSO and WOA

The accuracy and convergence of Artificial Neural Network (ANN) based models are contingent on initial weights as they employ backpropagation algorithm. To address these issues, this paper investigates and develops accurate Global Optimization (GO) assisted ANN based small-signal behavioral models for Gallium Nitride (GaN) High Electron Mobility Transistor (HEMT). Two potent explorer-exploiter frameworks-based optimization algorithms namely Grey Wolf Optimization-Particle Swarm Optimization (GWO-PSO) and Whale Optimization Algorithm (WOA) are utilized to fine-tune the initial weights of ANN. Thereafter, ANN, GWO-PSO- and WOA-assisted ANN based behavioral models are thoroughly examined and evaluated for various regression metrics. A strong correlation between the predicted and measured scattering parameters across the entire frequency spectrum is observed. We found both GO assisted ANN based models produced accurate models. However, GWO-PSO-ANN based models have shown better convergence behavior and at the same time the most accurate among all the tested models in this paper.