Gradient Descent-Based Model Predictive Current Controller for Dual Three-Phase PMSMs

Abstract

This paper proposes a gradient descent-based model predictive current controller, referred to as the three-voltage-vector gradient descent model predictive current controller (TVV-GD-MBPCC), for dual three-phase permanent magnet synchronous motors. The controller utilizes the gradient descent method to iteratively compute the optimal duty cycle by minimizing a cost function, with five iterations performed per control cycle. Furthermore, the optimal switching state is identified through a secondary minimization of the same cost function. To ensure a fair comparison, both the proposed TVV-GD-MBPCC and existing predictive current controllers are implemented on the TMS320F28379D microcontroller developed by Texas Instruments. The feasibility, accuracy and performance advantages of the proposed controller are validated through experimental results, demonstrating its effectiveness and practical applicability.