Simulated Annealing法の多点探索型への改良と最適化例  [in Japanese] Modification of Simulated Annealing to Multi-Point Search Type and Example of Optimization  [in Japanese]

This paper proposes an effective modification of simulated annealing method. Simulated annealing method has the ability to avoid being trapped in a local minimum. However, simulated annealing method is a sequential search type optimization method and the search point may be moved away from the global minimum. And so, we modify simulated annealing method so that the search point is split into two search point when the objective function is made a change for the worse. Subsequently, multi search points exist on the search space and the searchable patterns are increasing. Firstly, we attempt the optimization of Schwefel function which has multi extreme values. As a result, this proposed method makes better results than the ordinary simulated annealing method. Next, we implement the electrode shape optimization for the reduction of electric filed strength at power cable joint. The difference between the best result and the worst result in the case of the proposed method is smaller than that in the case of the ordinary simulated annealing method. Moreover, the proposed method makes better results than the ordinary simulated annealing method.