本文提出了一個改良式的粒子群演算法，稱為排名切換式粒子群演算法(Rank Switching Self-Learning Particle Swarm Optimization, RSLPSO)，利用排名機制及切換機制，來選擇粒子在每個迭代時期最適合的更新公式，有效的加快粒子的收斂速度，並且將低了運算量，並加入了學習機制，讓粒子之間的交流可以更頻繁，有效的做經驗交流，增加粒子的多樣性，使得粒子不易落入區域最佳解中。本論文使用16個測試函數來測試提出的改良式粒子群演算法的效能，而測試結果顯示，本論文所提出的改良法能夠在大部分的測試函數下表現優異，最後應用在交流馬達控制器的參數搜索，及時的找到馬達在每個取樣時間所需要的最佳控制器參數，來提升馬達的運動表現，而從馬達的響應圖中可以分析出使用本論文所提出的粒子群演算法可以有效的找到最佳的控制器參數。;This thesis proposes an improved particle swarm algorithm called Rank Switching Particle Swarm Optimization (RSLPSO), which uses a ranking mechanism and a switching mechanism to select the suitable update formula at each iteration time, effectively speed up the convergence of particles, and will reduce the amount of computation, and add a learning mechanism, so that communication between particles can be more frequent, effective experience exchanges, increase the diversity of particles, making particles not easily fall into the local optimum. In this dissertation, 16 test functions are used to test the performance of the proposed improved particle swarm algorithm. The test results show that the improved method proposed in this paper can perform well under most of the test functions. Finally, it is applied to the parameter search of the AC motor controller to find the best controller parameters needed by the motor at each sampling time in order to improve the motor′s performance. From the motor′s response graph, it can be analyzed the proposed particle swarm algorithm can effectively find the best controller parameters.
