模糊範例學習推論系統於水位預測之研究

The Fuzzy Exemplar-Based Inference System for Water Level Forecasting

蔡亞欣(Ya-Hsin Tsai)

指導教授 ： 張斐章

國立臺灣大學/生物資源暨農學院/生物環境系統工程學研究所/碩士(2004年)

https://doi.org/10.6342/NTU.2004.02018

摘要

在科學與學術的研究發展上，使用模糊推論系統(Fuzzy Inference Systems，FIS)於模擬人類的知識及邏輯推論，可以處理得更為貼近及有效率。本研究將提出一模式─模糊範例學習推論系統(Fuzzy Exemplar-based Inference System，FEIS)，期望能提高對水文現象的預測能力。
FEIS結合了以往應用於分類的範例學習超矩形模式(Exemplar-Aided Constructor of Hyper-rectangles，EACH)，並引入模糊推論系統的概念，使模式於學習與模擬人類智慧的特色更加細緻與精確。而FEIS擁有三項重要的特質：由數值資料萃取知識、知識規則庫、模糊推論過程。本研究首先應用一函數資料及一組混沌時間序列，做為驗證此模式的可行性及預估能力的正確性，並與傳統的EACH模式做比較，結果顯示EACH有良好的分類能力，而FEIS對於連續數值的推估及預測混沌時間序列的能力與精確度大為提昇。最後將此模式應用於蘭陽溪流域洪水時期水位的預測上，並與倒傳遞類神經網路(Back Propagation Neural Network，BPNN)做一比較，我們可以證實FEIS對於蘭陽溪流域的水位預測，具有優越的學習能力及推估能力。

關鍵字

水位預測 ； 神經網路 ； 模糊推論系統 ； 範例學習超矩形模式

並列摘要

Fuzzy inference systems have been successfully applied in numerous fields since they can effectively model human knowledge and adaptively make decision processes. In this paper, we present an innovative fuzzy exemplar-based inference system (FEIS) for flood forecasting. The FEIS is based on fuzzy inference system with its clustering ability enhanced through the EACH (Exemplar-Aided Constructor of Hyper-rectangles) algorithm, which can effectively simulate human intelligence by learning from experience. The FEIS exhibits three important properties: knowledge extraction from numerical data, knowledge (rule) modeling, and fuzzy reasoning processes. To explore its feasibility and predictive accuracy, a mathematical function and a chaotic time series are trained and validated by the model and also compared with the original EACH module. The results demonstrate that the EACH is suitable for categorization but cannot well present the continuous characteristic of the simulated function, while the FEIS can nicely fit the continuous mathematical function and well forecast the chaotic time series. We then apply the proposed model to predict one-hour ahead water level during flood events in the Lan-Yang River, Taiwan. For the purpose of comparison, the back propagation neural network (BPNN) is also performed. The results show that the FEIS model performs better than the original EACH and the BPNN. The FEIS provides a great learning ability and high predictive accuracy for the water level forecasting.

並列關鍵字

Neural networks ； Exemplar-aided constructor of hyper-rectangles ； Fuzzy inference system ； Water level forecasting

參考文獻

6. 梁晉銘、張斐章、陳彥璋，2000，”複合演算類神經-模糊推論模式應用於洪水預測”，中華水土保持學報，31(3)，pp.183-191。

3. 江衍銘，2002，二階段動態回饋式類神經網路於流量預測，國立台灣大學生物環境系統工程學系碩士論文。

19. Chang, F. J. and Lin, H. F. (1997), An application of the nested hyper-rectangle learning model for stream flow forecast in droughty seasons. Journal of Chinese Agricultural Engineering, 43(2): 20-45.

12. Aha, D., Kibler, D. and Albert, M. (1991), Instance-based learning algorithms. Machine learning, 6: 37-66.

14. ASCE Task Committee on Application of The Artificial Neural Networks in Hydrology, (2000a). Artificial neural networks in hydrology I: preliminary concepts. J. Hydrol. Engng, ASCE, 5(2): 115–123.