- 學位論文
模擬退火演算法於區間第二類模糊類神經網路控制器設計
Design of Interval Type-2 Fuzzy-neural Network Controllers Using Simulated Annealing Algorithms
若您是本文的作者，可授權文章由華藝線上圖書館中協助推廣。
摘要
本文提出利用模擬退火演算法(SA)於區間第二類模糊類神經網路控制器設計，該模擬退火演算法可以適當的調整模糊類神經系統內部的參數，並應用於函數近似與非線性系統之適應控制設計。此模擬退火演算法應用在適應控制器設計，不需要事先的離線學習程序和複雜的數學運算，相較於傳統的非線性系統適應控制器，可以有效減少適應控制器需要的複雜數學運算，模擬退火演算法的參數調整機制主要是依據閉迴路系統的穩定性，這樣一來可增加線上系統之穩定性。傳統上在非線性系統的適應控制過程中，模糊類神經控制器的權重値是透過模擬退火演算法做即時的線上調整，產生所想要的控制輸入。為了立即評估閉迴路系統穩定的趨勢，我們從Lyapunov函數的推導過程中，提出一個能量成本函數於模擬退火最佳演算法中，藉此來獲得更好的閉迴路系統的穩定度。此外，為了防止模擬退火法可能在控制過程中使系統狀態進入不穩定不安全的區域，我們加入監督控制器來限制，使閉迴路系統的狀態維持在安全穩定的區域。 本文利用電腦來模擬與實驗所提出方法的可行性與效果。最後將模擬退火演算法於區間第二類模糊類神經網路應用於具有直流轉換器之馬達控制實驗。
並列摘要
We propose a framework that can be applied into the controller design by integrating simulation annealing (SA) algorithm into interval type-2 fuzzy-neural network control. The SA algorithm can adjust the parameters of a fuzzy-neural network (FNN) so that the FNN system would be more adapted to functional approximation and adaptive controller design. Typically, the weighting parameters of an FNN controller are on-line adjusted by SA algorithm. Compared with traditional nonlinear adaptive controllers, the proposed SA-based adaptive controller design has a lower computational time complexity because it requires no more beforehand off-line learning processes and the relevant complicated operations. Because the adaptation mechanism of SA algorithm relies on the stability of the close-loop system, the stability of the proposed on-line system can then be enhanced. Moreover, in order to simultaneously evaluate the tendency of the stability of a close-loop system, a cost function which is derived from Lyapunov function is proposed and is applied to SA to obtain a better close-loop performance. Also, in order to prevent circumstance that SA algorithm might make our system unstable, an additional controller which acts as a supervisor is added to confine the close-loop system in a stable working state. Finally, the feasibility and efficacy of the proposed method are measured by computer simulations. This SA-based type-2 FNN control strategy is also experimented on the motors with DC-DC buck converter.
參考文獻
[4] W. Y. Wang, Y. H. Chien and I. H. Li, “ An On-Line Robust and Adaptive T-S Fuzzy-Neural Controller for More General Unknown Systems,” International Journal of Fuzzy Systems, vol. 10, no. 1, pp. 33-43, 2008.
[5] C. T. Lin and L. Siana, “An Efficient Human Detection System Using Adaptive Neural Fuzzy Networks,” International Journal of Fuzzy Systems, vol. 10, no. 3, pp. 150-160, 2008.
[95] 李承洲，黃克勤，DC/DC轉換器控制器設計之研究，國立臺灣師範大學工業教育系，2010。
[1] K. Hornik, M. Stinchcombe and H. White, “Multilayer feedforward networks are universal approximators, ” Neural Networks, no. 2, pp. 359-366, 1989.
[2] L. X. Wang, Adaptive Fuzzy Systems and Control, Prentice Hall, 1994.