類神經網路為基礎之非線性系統的最佳化模糊控制器設計

Abstract

在這篇研究中，我們提出了一個對未知的非線性系統，整合其模糊模型建立與最佳化模糊控制之方法。首先我們可以藉由線性的自建類神經模糊推論網路 ( linear self- constructing neural fuzzy inference network ) 得到非線性系統的Takagi-Sugeno (T-S) 模型。有了非線性系統的輸入與輸出的資料後，線性的自建類神經模糊推論網路可以動態的增加模糊規則的數目，並且調整各規則的參數，使得輸出的誤差最小化。如果每個子系統都是完全可控制與完全可觀察，我們便可將 [24]-[26] 中的最佳化模糊控制的設計方法，應用在所建立的T-S 模型上。當系統的模型無法得知的時候，這篇研究可以提供一個方法來穩定並最佳化控制這個物理系統。我們用四個例子來說明這個方法的設計流程。

In this work, we propose an integrated approach to fuzzy modeling and optimal fuzzy control for unknown nonlinear systems. We first obtain the Takagi-Sugeno (T-S) fuzzy model of the nonlinear plant by linear self-constructing neural fuzzy inference network (linear SONFIN). With training input and output data of the nonlinear system, linear SONFIN can dynamically increase the number of fuzzy rules, and also adjust the parameters of each rule to minimize the output error. Then, if each fuzzy subsystems is completely controllable and completely observable, we can apply the optimal fuzzy controller design scheme [24]-[26] to the proposed linear T-S fuzzy model. In the case of system model is unavailable, this approach can provide a way to stabilize and optimal control the physical system. Four examples are given to demonstrate the design procedure of this approach.