基于信赖域模型管理技术的永磁同步电机多目标优化

Multi-objective optimization of permanent-magnet synchronous motor based on trust-domain model management technology

针对永磁同步电机多目标优化过程中要反复调用耗时的仿真分析且在整个设计域构建代理模型精度低的问题,提出基于信赖域模型管理技术的永磁同步电机多目标优化算法,该算法将整个设计域划分为一系列信赖域,在信赖域上构建径向基函数代理模型并采用微型多目标遗传算法近似优化,降低多目标优化中代理模型的精度要求。通过代理模型与电磁仿真模型的近似程度来调节信赖域的大小,通过缩放和平移信赖域,引导各个电机目标性能逐步逼近Pareto最优前沿面,加快收敛速度。为了进一步减少有限元仿真次数,引入遗传智能布点技术,通过样本遗传策略,遗传落在下个信赖域上的样本。通过算例验证了该算法具有较强的求解永磁同步电机多目标优化问题的能力及较高的求解效率。

When the permanent magnet synchronous motor is subject to multi-objective optimization,the simulation and analysis are time-consuming due to repeated invoking and the it is inaccurate to construct the proxy model in the whole design domain.In this article,the algorithm for multi-objective optimization of the permanent magnet synchronous motor is proposed based on the trust domain model management technology,which divides the whole design domain into a series of trust domains.The radial basis function proxy model is set up in the trust region and approximated by means of the micro multi-objective genetic algorithm,in order to reduce the requirement of the proxy model’s accuracy in the process of multi-objective optimization.Through the approximation degree of the proxy model and the electromagnetic simulation model,the size of the trust region is adjusted.By scaling and translating the trust region,the target performance of each motor is gradually approximated to the Pareto optimal frontier,and the convergence speed is accelerated.In order to further reduce the number of finite-element simulation,the geneticsbased intelligent dot-distribution technology is introduced,and the samples on the next trust region are inherited with the help of the sample genetic strategy.The numerical examples show that this algorithm has strong ability in multi-objective optimization of the permanent magnet synchronous motor and high efficiency in solving such problems.