Maximum Torque per Ampere Control of PMSMs Using Real-Time Numerical Optimization

This paper proposes a novel online maximum torque per ampere (MTPA) control method for permanent magnet synchronous motors (PMSMs) that addresses nonlinearities such as magnetic saturation and cross-coupling effects. A flux-linkage based nonlinear PMSM model was developed to improve the torque control under dynamic conditions. The proposed gradient projection descent (GPD) method was introduced as a computationally efficient solution that is specifically tailored for real-time applications. It is evaluated alongside the Levenberg-Marquardt algorithm (LMA), which is renowned for its high accuracy and rapid convergence. Compared to the LMA method, the proposed GPD method achieves an optimal balance between computational efficiency and control accuracy, making it suitable for embedded systems with limited resources. Notably, the method can support online experimental characterization and online parameter estimation to capture real operating conditions, such as temperature and load variations, thereby providing a significant advantage for dynamic PMSM applications. Extensive simulations and experimental validations using a 1.3 kW interior PMSM prototype and TMS320F28375S control platform demonstrate the effectiveness of the proposed methods. The results demonstrate that GPD ensures real-time feasibility while maintaining acceptable levels of torque accuracy. These findings highlight the potential of GPD as a practical solution for high-performance motor control applications, especially in resource-limited embedded systems. Future work should explore online characterization and estimation for enhanced adaptability.