基于代理模型的不对称极永磁电机优化设计  

Optimization Design of Asymmetric-Pole Permanent Magnet Motors Based on a Surrogate Model

作　　者：洪佳琪 季宇 卢琴芬[1] HONG Jiaqi;JI Yu;LU Qinfen(College of Electrical Engineering,Zhejiang University,Hangzhou 310027,China)

机构地区：[1]浙江大学电气工程学院,杭州310027

出　　处：《微特电机》2024年第8期14-19,共6页Small & Special Electrical Machines

摘　　要：针对不对称磁极V型内置式永磁同步电机进行有限元优化耗时长的问题,提出了基于BP神经网络代理模型的优化算法。基于有限元模型对采样点进行性能计算,构建样本数据库,将转子相关结构参数作为优化变量,以高平均转矩和低转矩波动为优化目标。基于样本数据使用BP神经网络得到代理模型,采用NSGA-Ⅱ算法进行了结构优化。结果表明,优化后的结构参数具有较高的精确度,采样过程比直接基于有限元优化减少了90.7%的有限元模型调用次数。Aiming at the problem of long time consuming in finite element optimization for the V-shaped interior permanent magnet synchronous motor with asymmetric poles,can make the torque enhance,but also make the torque ripple increase,so the multi-objective optimization of the structure was very important.Since the direct optimization based on the finite element model exists the problem of long time consuming,the proposes an optimization algorithm based on the surrogate model of BP neural network was proposed.The performance of the sampling points was calculated based on the finite element model,and the sample database was constructed.The rotor-related structural parameters were taken as the optimization variables,and high average torque and low torque ripple were taken as the optimization objectives.Based on the sample data,the surrogate model was obtained by BP neural network,and then the NSGA-II algorithm was used for the structural optimization.The results show that the optimized structural parameters have high accuracy,and the sampling process reduces the number of finite element model calls by 90.7%compared with the direct finite element-based optimization.

关 键 词：不对称极内置式永磁同步电机 BP神经网络 多目标优化 非支配排序遗传算法-Ⅱ 

分 类 号：TM351[电气工程—电机]