Multi-objective Optimization of Permanent Magnet-assisted Synchronous Reluctance Machine Based on Dual-Driven Model with Search Space Reduction Method

<p dir="ltr">Permanent magnet-assisted synchronous reluctance machines (PMa-SynRM) are widely used for low cost, high efficiency, and have great potential for widespread application in the pump application. However, the design optimization of PMa-SynRM will be time-consuming by using conventional optimization algorithms due to complex geometric structure with a large number of parameters. In this paper, a novel design optimization method is proposed to improve the optimization efficiency while securing accuracy, by utilizing a physics-data dual driven model. The proposed method employs the low-fidelity simplified magnetic equivalent circuit to rapidly locate promising subregions in the global search space. Besides, high-fidelity (HF) finite element analysis cases are used to establish the surrogate model to accurately predict the optimization results in the local search space. Furthermore, a constrained space Latin hypercube sampling method is proposed for sampling in constrained local space to ensure the feasibility of sample to reduce the global search space. The proposed dual-driven model can reduce over 50% required time compared to traditional HF surrogate models with similar prediction accuracy. Finally, a 15 kW prototype is designed by the proposed optimization method, and fabricated and tested to validate the final optimization results.</p>