Double-vector model predictive current control for PMSM drive system with low calculation burden

• Published in: Dianji yu Kongzhi Xuebao = Electric Machines and Control Vol. 29; no. 5; p. 19
• Main Authors: Xu, Qiwei, Long, Xuehan, Miao, Yiru, Wang, Yiming, Tu, Yuxiaoying, Tang, Mengyang
• Format: Journal Article
• Language: Chinese; English
• Published: Harbin Harbin University of Science and Technology 01.01.2025
• Subjects: Computation; Control methods; Cost function; Predictive control; Simulation models; Steady state
• ISSN: 1007-449X
• DOI: 10.15938/j.emc.2025.05.003

The double vector model predictive control strategy can improve the steady-state control performance of the motor without significantly increasing switching losses. However, its vector determination method and duty cycle calculation process are more complex, which requires heavy computation burden. Therefore, a double-vector model predictive control scheme with low computational burden for PMSM was proposed. Firstly, when the PMSM operates in the steady-state, the candidate range of the first optimal active vector was reduced to three, including the first optimal vector adopted in the previous control period and its adjacent active voltage vectors. The cost function was sequentially substituted to determine the first optimal vector, thereby the number of comparisons decreases from six times to three times. Then, the remaining two vectors and the zero vector was considered as the alternatives for the second optimal vector, according to the deadbeat condition of the q-axis current, the duty cycle of all vector combinations was obtained separately. The vector combination and duty cycle that minimizes the cost function was obtained by substituting the vectors with their duty cycle into the cost function. Finally, the simulation models and experimental platforms were established, the stability, feasibility and effectiveness of the proposed model predictive control scheme with low computation burden were certified, respectively. The proposed model predictive control can achieve a current THD of 6.57% and a torque ripple of ±0.4 N·m with an average calculation time of only 15.3 μs. Compared with other model predictive control methods, it can obtain the best steady-state performance.