本文提出一個針對T-S模糊模型的新式建模方法,首先利用模糊c-平均值(Fuzzy C-Means, FCM)演算法將資料點做分群並找出合適的分群數,再將此數目當作規則數,把未知系統分成數個線性系統。接著結合粒子群最佳化演算法(Particle Swarm Optimization, PSO)與模糊c-回歸模型(Fuzzy C-Regression Model, FCRM)演算法來找出資料點與這些線性系統的模糊關係,並建立出模糊規則參數初始值,最後再用權重遞推最小二乘方法得到系統參數的最佳值,以建構出T-S模糊模型。 最後,我們使用本文的方法針對一些例子進行建模,並與其他方法做比較,經由實驗結果證明了本文提出的方法能更精準的建模。
A novel modeling method for T-S fuzzy model is proposed in this thesis. Firstly, fuzzy c-means algorithm is adopted to classify the data points and determined the numbers of the cluster. In addition, by defining the cluster numbers as the rule number, several linear subsystems can be divided from unknown system. Moreover, particle swarm optimization (PSO) algorithm and fuzzy c-regression model (FCRM) algorithm are adopted to find the fuzzy relationship between the data points and these linear subsystems, and construct the initial value of the fuzzy rule parameters. Finally, the weight recursive least squares method is adopted to obtain the optimal values of the system parameters and establish the T-S fuzzy model. Some models are illustrated to demonstrate that our modeling method can provide the more precious model than some well-known methods.