New Cost Function Design for Robustness Deadbeat Model Predictive Control of a Five-Phase PMSM

To suppress parameter mismatch in the model predictive control (MPC) system, this paper proposes a novel robustness deadbeat MPC method for a five-phase permanent-magnet motor considering d-q axis inductance and flux linkage parameter uncertainty. Firstly, the parameter sensitivity of MPC is analyzed, producing the predictive current error between the normal parameter and the mismatched parameter. The key of this proposed method is to design a new cost function based on the compensation of the predictive current error. Then, the newly designed cost function replaces the traditional one in the proposed robustness deadbeat MPC. Moreover, to further simplify the candidate vectors, a switching table of voltage vectors is developed to reduce the calculation complexity. Therefore, the total number of prediction vectors is two instead of ten in traditional MPC, and the computation time can be significantly reduced. Then, the duty cycle modulation technology is applied in the proposed method, and the optimal duration time is derived from the new cost function. Hence, the proposed method offers good robustness to parameter variation and lower computation time. Finally, the experimental results verify the validity and correctness of the proposed method.