張麗秋張斐章Chang, Li-ChiuLi-ChiuChangChang, Fi-JohnFi-JohnChang2009-02-232018-06-292009-02-232018-06-291999-12http://ntur.lib.ntu.edu.tw//handle/246246/139370傳統水庫操作系統具以下幾項特徵:蒐集歷史流量資料、藉助系梳分析方法尋求水 庫最佳操作策略(規線制定)、隨著新觀測資料的增加而予以逐步校正;但因規線的範圍大,無 法進行精確操作以有效利用水資源,且未考慮入流量預測與極端事件之影響,往往限制了水庫 的防洪與救旱機能。為改進傳統操作上的缺失,本研究以調適性網路模糊推論系統(ANFIS)模 擬控制水庫操作之放流量。 ANFIS係將模糊邏輯理論架構在類神經網路系統上,以便於處理數值資料與定性(語意)的知 識;本文以模糊減法聚類作為模糊推論的架構,並建立相關隸屬函數,再經由ANFIS學習與自 我調適求得函數參數最佳解;此模式運用在水庫操作系統的主要步驟有二:(一)資料前處理: 收集歷史流量資料與建立適當的水庫操作目標函數,套用遺傳演算法求得最佳放流歷程作為 類神經網路的訓練標型;(二)建立ANFIS控制模式:從網路訓練時期調整適當的參數後,以推 求未來時期水庫之最佳放流量;本研究以石門水庫操作為例,結合上述方法,由歷史流量模擬 操作結果,並與傳統的石門水庫M-5規線操作結果相比較。A common strategy in the traditional reservoir operation is to search rule curves through system analysis according to historical inflow data, then, step by step, to adjust rule curves with coming data. This strategy is easy and convenient for reservoir operation; however, the ranges between rule curves are too large to precisely operate outflow of reservoir for water usage. Moreover, the information of inflow and extreme events are not taken into account. Consequently, the rule curves, in general, could not work very well for the prevention of flood or drought. From the aforementioned points, we propose an adaptive network-based fuzzy inference system (ANFIS) to enhance the efficiency of reservoir operation. By combining fuzzy inference systems and neural networks, ANFIS not only can handle both quantitative (numerical) and qualitative (linguistic) knowledge, but also can successfully deal with the control laws for a complex system. Two main procedures are performed in order to implement the model to reservoir operation system. First, the genetic algorithm is used to search the optimal reservoir operating histogram which is recognized as the training pattern is the next step. Second, the ANFIS model is built to create the fuzzy inference system, to propose the suitable parameters, and to estimate the optimal water release. The proposed model is intended to investigate its practicability and efficiency by using the Shihmen reservoir. The M-5 rule curves are also performed for the purpose of comparison. The results show that the ANFIS model has better performance than the M-5 rule curves.en-US適應性網路模糊推論系統水庫操作模糊邏輯理論遺傳演算法類神經網路Adaptive network-based fuzzy inference systemReservoir operationFuzzy inference systemsGenetic algorithmNeural networksjournal articlehttp://ntur.lib.ntu.edu.tw/bitstream/246246/139370/1/智慧型水庫即時操作控制系統.pdf