複合演算類神經－模糊推論模式應用於洪水預測

Hybrid Artificial Neural-Fuzzy Inference Model for Flood Forecasting

梁晉銘(Jin-Ming Liang) ； 張斐章(Fi-John Chang) ； 陳彥璋(Yen-Chang Chen)

《中華水土保持學報》 31卷3期 (2000/09) Pp. 183-191

https://doi.org/10.29417/JCSWC.200009_31(3).0002

摘要

本研究介紹團塊理論與類神經-模糊推論模式之組合架構與推理程序。其中團塊理論之分類程序，係使歷史事件之特徵得到掌握並進而界定出系統內含之不同特性；其優點除可使推論式依據之規範數量客觀得到外，並能藉由模糊理論之觀念，結合已知事件之特質對未知事件予以預判，俾使水文序列之推估較爲精確；此外，研究亦利用改變推論函式型態的方式使模式結構進一步簡化。有關模式成效之檢測，係利用（1）定率函數輸出模擬及（2）清水溪暴雨事件流量序列之推估來進行；透過模式建構並應用顯示，對上述定率函數與水文序列之檢定與驗證，其結果皆呈良好，足見此模式之實用性。

關鍵字

團塊理論 ； 類神經-模糊推論模式

並列摘要

This study presents a novel structure and reasoning process of an improved Artificial Neural-Fuzzy Inference Model (ANFIM). The ANFIM has the hybrid learning scheme, unsupervised and supervised schemes. The fuzzy min-max clustering is introduced to extract information from the input data. An advantage of the fuzzy min-max clustering is it s ability to determine the characteristics of each cluster. The efficiency and accuracy of the model are first tested by using a deterministic function. Then, the model is employed to build the rainfall runoff model of the Chingshuin River for fore casting the one-hour ahead flood. The results show that the ANFIM can simulate the deterministic function and predict flood accurately.

並列關鍵字

clustering theory ； Artificial Neural-Fuzzy Inference Model