An Improved Gradient-Based Optimization Algorithm for Synchronous Optimal Modulation of High-Speed PMSM Drives

Abstract

By employing offline optimized pulse patterns (OPPs), synchronous optimal modulation (SOM) can achieve minimum current total harmonic distortion (THD) in high-speed PMSM drives when switching-to-fundamental frequency ratio is low. However, the pulse pattern optimization can be very time consuming and has to be tailored for each high-speed drive. This paper presents an improved gradient-based optimization algorithm for SOM of high-speed PMSM drives. It incorporates analytical gradients of the nonlinear constraints, and employs both predictive and random initial values in two steps of optimizations. The proposed method can greatly reduce the computation time of the OPPs and avoid tedious and complex re-optimizations as required in conventional methods. Additionally, based on the optimization results, the total harmonic currents of a high-speed PMSM drive with SOM in real-time control can be predicted, which can serve as an efficient performance evaluation tool. The numerical optimization results, extensive time-domain simulation and experiment results on a prototype high-speed PMSM drive have been obtained and the effectiveness of the proposed method validated.