運用改良粒子群演算法於大型最佳化問題

飛鳥演算法(又稱為粒子群演算法)常用於連續型最佳化問題的求解過程。這個演算法主要是將飛鳥覓食的行為作為演算法的模擬機制。從1995年發表至今，廣泛的被應用於生活中各種不同最佳化連續型的問題。因為是模擬飛鳥的覓食行為，若領頭的飛鳥無法找到最多得食物，整群的飛鳥就有陷於區域最佳解的可能，所以演算法常有落於區域最佳解的危險中。因此，如何改善搜尋的多樣性，以防止提早收斂到區域解的危險，就變得非常的重要。本研究將演化式策略加入原始的演算法中，並開發適當機制來增強搜尋的多樣性以改善演算法的正確性與準確性。經過不同的實測，相對於不同版本的飛鳥演算法，本方法在精確度上都有優良的表現。

關鍵字

最佳化；粒子群演算法；演化式策略

並列摘要

Particle swarm algorithm (PSO) is a popular method used for continuous optimization problems. It simulates the social behavior of bird flocking for food. This algorithm was first published on 1995, and it is commonly used to solve the real world problems with all the different evolutions. Since this algorithm mimics the behavior of bird hunting food, it is very likely the whole flock traps in local minima if the leading bird fails to reach the global optima. How to increase the diversity to avoid the premature condition so it can obtain the global optimization has been a research problem. This study improves the original PSO algorithm by blending the evolutionary strategy to enhance diversity so that the robustness and accuracy can be ensured. Various benchmark functions have been tested and compared to verify the algorithm fidelity.