Global Optimal Predictive Control of PMSM Using Dynamic Programming: An Offline Benchmarking Tool

Abstract

Predictive Permanent Magnet Synchronous Motor (PMSM) control with a long preview horizon significantly improves motor transient response and reduces torque and current ripple. Finding the global optimal control actions is challenging due to the hybrid nature of the continuous armature current dynamics and discrete inverter states. This research utilizes Dynamic Programming (DP) to find the global optimal inverter switching strategy for PMSM control with a previewed torque demand. The proposed DP-based PMSM control serves as an offline study tool for predictive motor control strategies, addressing a significant gap in the existing literature. The proposed method takes advantage of the limited number of inverter states, significantly reducing computational load compared to conventional state-space DP. Through three case studies using a two-level inverter, the offline generated DP results effectively benchmark other PMSM control strategies. The first case study optimizes non-predictive PMSM control by tuning the cost function, including field weakening actions for high-speed operation. The second case study explores the benefits of a receding horizon Model Predictive Control (MPC) setup with previewed torque demand. Performance improvement is observed with increased preview horizon length, plateauing after around 10 steps or 5 ms. The last case study benchmarks a previously published Explicit Model Predictive Control (EMPC) based PMSM control strategy using the proposed DP approach. Comparison indicates similar performance between EMPC and DP when sampling frequency and weight tuning are consistent.

Keywords

• PMSM
• inverter control
• dynamics programming
• model predictive control
• hybrid system control