透過您的圖書館登入
IP:3.142.201.214
  • 學位論文

結合模糊類神經與遺傳演算法 於水庫即時操作

Real-Time Reservoir Operation Using Fuzzy Neural Network and Genetic Algorithms

指導教授 : 王安培

摘要


台灣雨量豐沛,然因川陡流急、降雨時空分佈不均,伴隨而來的往往是人民生命財產的損失。又因台灣人口密度高達世界平均人口密度的 14 倍,枯水期常因颱風期間水庫洩洪過多造成農工商業及民生用水不足之問題,此誠為台灣水資源主要問題所在。興建水庫以增加河川蓄水、疏洪及供水能力實為根本解決之道,但優良壩址多已開發殆盡,另尋新址並不容易,因此,如何妥善利用現有水庫資源,更有效地發揮水庫防洪蓄水之功能,實乃一重要課題。 本研究期望建立一水庫最佳放流預測模式,在颱風期間事先推估三小時之最佳放流量。首先分別嘗試以類神經網路(Artificial Neural Network,簡稱ANN)與適應性網路架構模糊推論系統(adaptive network-based fuzzy inferencesystem 簡稱為 ANFIS)建立石門水庫上游降雨 ─ 逕流模式與下游放流 ─水位模式,最後再利用兩模式結果與遺傳演算法(Genetic Algorithms,簡稱GA)優選符合水庫本體安全、下游河道不溢堤與洪後盡量達到理想蓄水量三大目標之水庫最佳放流量,作為最佳放流預測模式建立之依據,最後再以ANFIS 建立最佳放流預測模式,以供水庫操作者於颱風期間操作之參考。 本研究在建立最佳放流優選模式時,考慮後續 ANFIS 預估流量中,可能產生低估之情形。為避免因低估放流使得水庫水位超過滿水位 245 公尺,故調整 GA 優選模式之限制水位,使 ANFIS 模式即使產生流量低估,亦能使水庫水位維持在滿水位下,保持水庫安全。結果顯示,此觀念是可行的。

並列摘要


Although Taiwan has rich precipitation, it usually brings damage to people’s lives and properties due to steep rivers and fast flow. Also, the population density in Taiwan is 14 times in the world; the water supply demand of domestic, industrial and agricultural would be not enough if excess floodwater is released during the typhoon period, which is the major problem of water resource management. Building reservoirs to advancement the hydraulic ability of rivers is the basal approach, but it is hard to find the new places. Therefore, how to allocate water resource well and operate existing reservoir effectively become the main issue. The purpose of this study is to simulate the optimum reservoir release model, which predicting the optimum releasing volumes 3 hours in advance. At first, the Artificial Neural Network (ANN) is used to simulate the rainfall-runoff model and the adaptive network-based fuzzy inference system (ANFIS) is used to applied to analyze the water stage increment due to Shihmen Reservoir’s releasing. Secondly, Genetic Algorithms (GA) is applied to select the best release of reservoir during flood period. Finally, the optimum reservoir releasing model is built up; the objective conditions are dam safety, flood control and optimal reserving volumes. It is expected that results of this study could be used for online reservoir operation in the future. In this study, since the results of ANFIS model may underestimate reservoir release, GA model needs to adjust the limited water level for ensuring reservoir’s high volumes even model underestimates. The results show that the proposed approach is accurate and feasible.

並列關鍵字

Optimum reservoir operation Fuzzy theory ANFIS GA ANN

參考文獻


37. 宋政遠(2007),「河道變遷分析─以大甲溪下游為例」,中原大學土木工程學系研究所碩士論文。
16. 張淑勤(2006),「模糊決策在防洪之應用」,中原大學土木工程研究所碩士論文。
35. 黃泳樹(2006),「西北颱降雨分析」,中原大學土木工程學系研究所碩士論文。
36. 邱繼弘(2007),「河道沖淤分析─以大甲溪下游為例」,中原大學土木工程學系研究所碩士論文。
14. 張大元(2003),「類神經網路在水庫放流對河川水位增量之研究」,中原大學土木工程研究所碩士論文。

被引用紀錄


蘇嘉惠(2010)。多變量時間序列模型應用於入流量預測〔碩士論文,中原大學〕。華藝線上圖書館。https://doi.org/10.6840/cycu201001003
紀昀(2013)。輻狀基底函數類神經網路結合遺傳演算法於河川水位預測之應用〔碩士論文,中原大學〕。華藝線上圖書館。https://doi.org/10.6840/CYCU.2013.00388
吳建宏(2009)。模糊決策在洪水期間橋梁安全預警系統之應用-以新海大橋為例〔碩士論文,中原大學〕。華藝線上圖書館。https://doi.org/10.6840/CYCU.2009.00838
林偉立(2009)。結合ANFIS模式與Web GIS技術於雨水下水道水位預測〔碩士論文,中原大學〕。華藝線上圖書館。https://doi.org/10.6840/CYCU.2009.00799

延伸閱讀