Cost Optimization Method for Less-Rare-Earth Permanent Magnet Synchronous Motor Based on Surrogate Model

To solve the problem of increasing the manufacturing cost of all-rare-earth permanent magnet synchronous motor (ARE-PMSM) due to the rising price of rare-earth permanent magnet (PM), two kinds of less-rare-earth permanent magnet synchronous motors (LRE-PMSM) are proposed. Firstly, the material cost and electromagnetic properties of LRE-PMSM and ARE-PMSM are compared and analyzed. Then, before optimization, sensitivity analysis is performed. Thirdly, the sample data is collected by using the space filling Latin hypercube method, and the surrogate model are constructed by using BP neural network (BPNN). With the size of PM, the proportion of ferrite PM and the air gap as optimization variables. The electromagnetic torque and core loss as constraint conditions. Furthermore, the genetic algorithm is employed to optimize the cost of PM, resulting in a 15.59% reduction in the cost of PM. Finally, the correctness of proposed optimization method based on BP neural network surrogate model is proved by the results of finite element simulation.