Evaluation of Direct Torque Predictive Control for SRM with Reduced Computation Resources

Switched reluctance motors are considered prime candidates for electric vehicle applications due to their numerous advantages, especially the absence of permanent magnets. However, controlling switched reluctance motors poses significant challenges due to their high levels of nonlinearity. In this work, a direct torque predictive control algorithm is proposed. This method requires fewer parameters to be tuned compared to indirect torque control and involves fewer offline steps, thereby reducing both memory and computational demands while offering better dynamic robustness. Typically a finite control set model predictive controller for a three-phase switched reluctance motor drive evaluates 27 switching states at each control instant, but the proposed approach reduces this number to 9. Consequently, the method is more computationally and memory efficient, making it practical for industrial applications. By incorporating a modified cost function, the controller demonstrates excellent performance at lower speeds across various metrics. Nonetheless, challenges at higher speeds are identified, particularly regarding the controller's limitations in achieving advanced commutation due to the generation of slight negative torque at specific electrical positions.