本篇論文中，我們提出一種改良式的粒子群演算法，名為切換式自我學習粒子群演算法(Switching Self-Learning Particle Swarm Optimization, SSLPSO)，利用切換式的方式，在不同時期運用不同的速度更新公式，使運算量降到最低，並加入了上升及下降函數，使其收斂速度能更快，前期粒子的多樣性足夠，後期粒子運用自我學習機制，學習群體中表現最佳者，讓解不落入區域最佳解，搜尋到全域最佳解。我們最後利用16個測試函數進行模擬，與其他已提出的19種改良式粒子群演算法做比較，實驗結果得知，本論文提出新的改良方法，能夠在大部分的測試函數中有較優越的表現。並且比較其中4種演算法利用MATLAB進行運算搜尋解的時間，分析可得我們論文不但具備「精確」，更擁有「快速」的優點，也就是省時。;In this thesis, we propose a new particle swarm algorithm called Switching Self-Learning Particle Swarm Optimization (SSLPSO), which switches to different velocity updating formulas in different stages(periods), so the amount of calculation can be minimized. By adding "the rise and fall functions", the convergence rate can be faster. While the diversity of the particles are abundent at the beginning, the particles apply self-learning method at the later stage to learn from those who have the best performance, thus not falling into local optimum but reaching the global optimal. The 16 benchmark functions are used in the simulation of our proposed. The experimental results show that the proposed SSLPSO performs better on most of the functions, compared with other 19 improved particle swarm algorithms. We also analize 4 algorithms with MATLAB to see which can find the solution faster, and the result shows that the proposed SSLPSO is not only more accurate but also more efficient, which means more time can be saved.
