基于移动最小二乘法和粒子群算法的优化算法

Optimal method based on moving least square and particle swarm optimization

针对复杂电磁装置优化问题中目标函数计算次数过多的问题,提出了一种基于移动最小二乘法(MLS)和粒子群优化算法(PSO)的快速全局优化方法.该方法利用基于MLS的表面响应模型,重构原始的优化问题,采用加权PSO算法对重构后的目标函数进行寻优,再使用拟牛顿法,对原优化问题直接寻优,从而得到优化问题最终的最优解,并对基准测试函数和实际电磁装置问题进行优化计算.结果表明,与加权PSO相比,该算法能找到优化问题的全局最优解,并能有效地减少目标函数的计算次数,节省了计算时间,提高了计算效率.

To reduce the calculating numbers of objective function needed in solving the complicated optimization problems of electromagnetic devices, a fast global optimization method based on moving least square （MLS） and particle swarm optimization （PSO） was proposed. The original optimal problem was reconstructed by using the response surface model based on MLS, and then the reconstructed objective function was solved by the weighted PSO algorithm. Quasi-Newton method was adopted to directly search the original optimal problem again, and the final optimal solutions were found. The benchmark test function and practical electromagnetic device were optimized by utilizing the method. The results show that compared with the weighted PSO, the proposed method can find the global optimal solution of the optimal problem, reduce the number of objective function calculated, save computation time and improve the calculation efficiency.