Surrogate-Model-Driven Staged Sequential Multi-Condition Optimization Method of Permanent Magnet Vernier Motors

Permanent magnet Vernier motors (PMVM) exhibit great potential in direct-drive application scenarios. As no-load and on-load performance both affect their overall operational smoothness and reliability, both conditions must be considered during optimization to meet the specific application requirements. However, the high-order harmonics of PMVM make the electromagnetic calculation more time-consuming than permanent magnet synchronous motors (PMSM). Therefore, in this article, a surrogate-model-driven staged sequential multi-condition and multi-objective optimization (MOO) method for PMVM is proposed, aiming to comprehensively optimize the motor performance under both no-load and on-load conditions with reduced computational cost. (1) Unlike traditional methods that involve redundant calculations for both no-load and on-load cases, this article employs a surrogate model-driven staged sequential optimization strategy to sequentially optimize the no-load and on-load performance. This method optimizes the on-load condition electromagnetic performance while maintaining near-optimal no-load performance, thereby achieving multi-condition optimization with reduced computational resources. (2) The optimization results are validated using the finite element method (FEM) and PMVM prototype experimental verification. The optimization results indicate that the proposed optimization method achieves ∼10% reduction of the computational cost compared with the traditional surrogate-driven optimization method, confirming the effectiveness of the proposed method for PMVM optimization. The saved computational cost varies with the complexity of the motor topology. The proposed optimization method can be extended to the optimization problems of different types of motors with more complex motor topologies, considering multiple operating conditions or multiphysics optimization.