Intelligent Memetic Algorithm Using GA and Guided MADS for the Optimal Design of Interior PM Synchronous Machine

Abstract

Optimal design of an electric machine based on finite element analysis (FEA) calls for much longer computation time for maintaining high accuracy. In order to compensate for the excessive computation time and guarantee the reliable convergence to a global optimum, an intelligent memetic algorithm is newly implemented by combining a genetic algorithm (GA) and the guided mesh adaptive direct search (MADS) that employs an extension search step after the poll step. The effectiveness of guided MADS (GMADS) alone has been verified through the function optimization, and the proposed memetic algorithm is applied to an optimal design of an interior permanent magnet synchronous machine (IPMSM), of which the cost function has many local minima. Optimization results confirm that the proposed method locates an acceptable solution more effectively maintaining the reliable accuracy.