The Framework of Real Time Sediment Flushing Operation

台灣受地形及降雨特性影響，水資源利用困難，為增加可用水量，主要透過興建水壩截蓄水源以供利用，然經過長年開發，再加上先天地形限制，能新建水庫之位址已相當缺乏，再加上近年環保意識高漲，水庫興建更顯困難。也因此，現今政策多新建水庫開拓水源轉向現有水庫的更新與有效利用。水庫淤砂一直是永續水庫利用之最大挑戰。隨著氣候變遷、極端降雨事件頻繁，使得水庫上游集水區土石崩落、水流常浹帶大量泥砂入庫，造成水庫淤砂日益嚴重，大幅減低水庫使用效益並減少水庫壽命。如何維持水庫之庫容為一重要議題，雖現今已發展出不少清淤方式，在實務上有關於水庫即時排砂操作之研究相當缺乏，排砂之水庫操作目標即希望透過適當操作以排除濁水、蓄存清水，然操作之困難仍在於入流量與入砂量之不確定性。 本研究旨在解決不確定入流下，水庫即時排砂操作策略之擬訂，建立即時最佳操作模式以決策水庫排砂操作。為取得水庫於排砂與蓄水兩大目標間之平衡，模式操作目標包含有(1)使水庫排砂量最大化及(2)使水庫最終蓄水量達所訂定之目標蓄水量。為考量入流之不確定性，本研究透過序率規劃找出最佳操作策略，亦即決策水庫操作期間內每一時刻之放水量，並考量不同因子對操作結果之影響。為簡化濃度分布所造成之複雜計算，本研究假設濃度於空間上為均勻分布。首先建立兩期兩階段序率規劃模式(2-period 2-stage stochastic programming)，研究目標函數特性及主要影響因子，發現放水決策主要受入流濃度影響。接著進一步將模式延展至長時間操作，並研究不同入流量-入流濃度關係之入流對於水庫操作之影響，發現水庫排砂操作之決定因素主要為入流濃度與水庫濃度間之關係。而即使在不同入流-濃度關係之入流條件下，水庫之排砂操作仍可歸納為同一操作策略，即排砂操作於水庫濃度極高時集中在短時間內大量排砂。最後，將模式修正為多階段序率規劃模式(multi-stage stochastic programming)，並透過簡化後之目標函數進行最佳決策以提高操作之效益與計算之效能。可發現水庫操作仍依循同一操作策略：水庫於事件初期先進行蓄水、並維持高水位，當水庫濃度達到最大值時便進行大量放水以排出泥砂，最後再將水庫蓄回至目標蓄水量。其中，當入流事件之入流濃度尖峰早於洪峰時，可得到最佳之水庫排砂效益，這是由於此種入流事件下，水庫能夠於早期排除濁水、再蓄存清水之緣故。 最後，本研究將模式應用於莫拉克颱風、凡娜比颱風、蘇拉颱風及蘇力颱風於石門水庫之案例分析，可發現除了凡娜比颱風之外，其排砂效益可高達五至七成，並確保水庫蓄水量於颱洪事件結束後仍可高達八至九成。其中由於凡娜比颱風之入流量極少且當時水庫水量低，故入流多被蓄存而使得排砂量少、排砂效益低。將模式於四場颱風之操作結果與歷史操作之結果相比較，在能夠達到相同最終蓄水量之條件下，透過模式進行水庫排砂操作能夠提高兩倍以上之排砂效益，成效顯著，惟模式操作之成效好壞仍受制於入流預測之精度影響。Reservoir sedimentation is a serious problem with considerable environmental and economic implications. The sedimentation significantly decreases the reservoir capacity and reduces benefits. To achieve the sustainability, it is critical to maintain the reservoir capacity. Although the idea is not complex, the real-time sediment flushing operation is still not well studied yet. The goal of this kind of operation is obvious. The key point is how to store clear water and release the muddy water in an appropriate manner. But the difficulty of sediment flushing operation is from the uncertainty of both inflow and sediment discharge. The two objectives of real-time sediment flushing operation are 1) maximization of the total flushing sediment volume, and 2) reaching the target storage after the flood event. The decision variables are water release during operating period with the assumption of uniform spatial distribution of concentration. This study considered the uncertainty of both inflow and sediment discharge, and developed a two-period two-stage stochastic programming to study the relationship between reservoir concentration and release operation. The result suggests that the release decision is mainly influenced by the inflow concentration. In the end, a reservoir operation optimization model was developed for operational decision in two ways: (1) time period extension of two-stage stochastic programming model, and (2) multistage stochastic programming model. These two models were tested with different inflow relations between inflow concentration (CI) and inflow discharge (I), and further examined the influence of inflow information on the operational policy. As a result, the key factor of the operational decision is the relationship between reservoir concentration and inflow concentration. With different inflow relations, the reservoir operation can be generalized to the same strategy: taking the flushing operation in short time as the reservoir concentration rises to the extreme value. Among the inflow relations, the flushing efficiency performs the best when the inflow concentration peak is earlier than inflow peak. After all, this study took Typhoon Morakot, Typhoon Fanapi, Typhoon Saola and Typhoon Soulik in the Shihmen Reservoir to investigate the efficiency of operational policy. Results show that the outcomes of the operational model are better than that of the historical operation according to the two targets: flushing efficiency and final storage