Performance Improvement of Model-Predictive Current Control of Permanent Magnet Synchronous Motor Drives

TLDR

MPCC is a high‑performance control strategy for PMSM drives, but its use of a single voltage vector per period limits performance, especially with two‑level converters. This paper proposes an improved MPCC that estimates back EMF from past stator voltage and current, uses it in current prediction, and introduces two methods for optimal voltage vector selection and duration. The improved MPCC employs a cost function to choose the voltage vector that minimizes current error, applies a null vector alongside the active vector, and offers a six‑prediction method requiring current differentiation or a single‑prediction method enabling efficient duty calculation, all compared to conventional MPCC and deadbeat control with SVPWM. Simulation and experimental results confirm that the proposed methods achieve superior steady‑state performance while preserving the quick dynamic response of MPCC.

Abstract

Model-predictive current control (MPCC) is widely recognized as a high-performance control strategy of permanent magnet synchronous machine (PMSM) drives due to its quick response and simple principle. It uses a cost function to select the best voltage vector minimizing the current error between the reference value and the feedback value. However, as only one voltage vector is applied during one control period, it fails to give satisfactory performance due to the limited voltage vectors, especially in the case of two-level converters. This paper proposes an improved MPCC strategy for PMSM drives, which first estimates the back electromotive force (EMF) based on the past value of stator voltage and currents and then applies the estimated EMF in the stator current prediction. To achieve steady-state performance improvement, a null vector along with the active vector obtained from conventional MPCC is applied during one control period. Two methods are proposed to achieve optimal vector selection and vector duration. The first one requires six predictions and the calculation of current differentiation, while the second one only requires one prediction to obtain the best voltage vector and its optimal duty can be obtained in a very efficient way. The proposed methods are comparatively studied and compared to conventional MPCC and deadbeat control with space vector modulation. Both simulation and experimental results confirm the effectiveness of the proposed methods in achieving good steady-state performance while maintaining quick dynamic response.