考量生態流量之水庫最佳操作策略

Abstract

近年來水資源的需求日益提高，但受制於台灣自然條件因素之影響，水源供給量相對有限；雖然透過水利設施的興建與營運能改善水資源不足的問題，但對於河川水域環境卻可能造成負面的影響，加上水資源的開發過程中均以人類的用水需求為主要考量，往往可能滿足人類對水的需求，卻失去了生態環境系統的平衡。本文旨在探討考量河川下游生態流量下研議水庫對下游地區供水的最佳營運策略；研究區域以石門水庫後池堰至鳶山堰段之大漢溪河道為主；利用數理規劃模式建置水庫操作於滿足各項用水需求與環境限制之操作策略，採用懲罰機制之限制型遺傳演算法優選水庫的最佳蓄放水歷程，以滿足水庫長期供水與下游河段的生態需求，將河川水域環境受水利工程構造物之影響降至最低。研究結果顯示，過去石門水庫歷年供給下游水量，大致上能達到前人研究所提的最低生態流量(4 cms)，惟乾枯水年仍有少數旬別低於生態最低需求；而本研究所研擬之數理規劃模式的最佳操作結果，不僅能適度提高河川環境生態之流量，減緩工程構造物對河川的影響，亦改善了下游地區缺水之情形；研究結果亦證實，限制型遺傳演算法架構之水庫操作模式，能有效提高優選的效率，並獲致良好的優選結果。

In recent decades, due to increasing on water demand by coupled with unbalanced temporal and spatial distribution of rainfall in Taiwan, water supply becomes more difficult. To overcome the problem, attention has been focused on improving water resources management and reservoir operation. Nevertheless, such actions frequently result on the degradation of river environmental condition. In an attempt to minimizing these negative impacts, environmental related aspects must be considered on the operation and management of water storage facilities, such as reservoirs. In this study, we propose a novel optiming technique for reservoir operation to deal with multiple water users, specially emphasizing downstream ecological flow demand. The artificial intelligence techniques, such as genetic algorithm (GA), have been successfully applied for the optimization of complex hydrosystems. The GA is applied for the optimization of water resources management. To improve the capability of the GA, a penalty function is proposed, resulting in a restricted GA model. This penalty function is cooperatively considered with the reservoir operational objectives, which also includes water quantity and ecological aspects. To investigate the applicability of the proposed methodology, the downstream area of Shihmen reservoir from Hochih Weir to Yuanshan Weir in the Dahan River is used as the case study. Based on several previous studies, the water supply of Shihmen reservoir for dowmstream could reach a base ecological flow to about 4 cms in most of time. By implemented the proposed restricted GA for optimizing the Shihmen reservoir operation in last 20 years, the results demonstrated model can provide much better the performances, in term of small GSI and large ecological flow, for most of years than the historical approach. We conclude that the restricted GA approach can improve the efficiency and effectiveness of reservoir optimization operation for mutilpe water users.