Optimization and Implementation of Three-Phase PMSM Current Harmonic Decomposition Technique 2019-01-0604

With the development of electric vehicle (EV), permanent magnet synchronous motor (PMSM) has received more and more attention. PMSM torque ripple suppression is one of the core technologies of PMSM control. Current harmonic injection method is a commonly used torque ripple suppression method. In order to accurately control the injecting current harmonics, it is necessary to quickly and efficiently decompose the three-phase PMSM current harmonics first. In this paper, an existing instantaneous harmonic decomposition method based on multiple reference coordinates is adopted. First, the causes of the analysis error of the harmonic decomposition technique are analyzed which are divided into internal factors (e.g. analysis errors generated during the discretization of continuous functions) and external factors (e.g. sampling errors). Analysis errors will directly affect the decomposition result. Then, a method that evaluates the analysis errors of the harmonic decomposition technique based on matrix condition number is established. The matrix condition number reflects the anti-interference ability of the analytical method. In this paper, the genetic algorithm (GA) is used to optimize the target filter parameters while the matrix condition number is the optimization goal. The effect of internal analysis errors caused by PMSM speed changing is considered and the required sampling frequency under different speed is calculated. At last, the harmonic decomposition method is implemented on 32-bit DSP (TMS320F28379D) and an online decomposition test is carried out on the experimental bench. The harmonic decomposition method can run fluently on the 32-bit DSP at an operating frequency of 10K, and can effectively decompose the three-phase PMSM current.