曾文水庫防洪減淤操作最佳規劃模式之研究

Abstract

台灣山區地勢高聳，地質結構鬆散，且夏秋季時常發生颱風豪雨，大量洪水及土石流，除造成水庫潰壩，與下游河段溢堤等災害之風險外，大量泥砂進入水庫庫區造成嚴重淤積，將縮短水庫壽命，並影響水庫長期供水效應。因此，如何在颱洪時期藉由適當的操作手段，減輕颱風暴雨對水庫壩體與下游河道災害之影響，以及提高水庫排砂量，乃是一個當前重要的議題。 有鑒於此，本研究乃整合遺傳演算法、河道水理演算、類神經網路及庫區泥砂動態模式，並以曾文水庫為研究區域，發展水庫防洪減淤操作最佳規劃模式，利用多場歷史颱風場次進行分析，其結果可提供後續規納防洪操作規線研究時之參考依據。模式中之目標函數同時考量減災、水資源與減淤等不同標的，水庫操作則以曾文水庫運用要點之規定為依歸，本研究並針對各目標項之權重值做進一步探討，並建議出在優先保障民眾安全且兼顧水資源供應與水庫排砂下較合理之權重比例。 本研究進一步比較本研究發展之模式所得結果，與以往之實際操作結果，以2007年柯蘿莎颱風及2008年辛樂克颱風為例，就降低下游淹水而言，對柯蘿莎颱風本模式可將溢堤高度由2.7公尺降至0.9公尺，對辛樂克颱風則可將水位由溢堤0.6公尺降至堤頂以下。就減淤效果而言，對柯蘿莎颱風本模式可將排砂量由627萬噸增至840餘萬噸，約增加33%，對辛樂克颱風而言，本模式可將排砂量由160萬噸增至200萬噸，約增加25%。整體而言，模擬的結果顯示本研究發展之水庫防洪減淤操作最佳規劃模式，除可降低下游河道溢堤情況，保障民眾生命外，更可有效提升排砂量，延長水庫壽命，達永續經營之目的。

Due to fragile geology, steep topography, and frequent storm events, reservoirs are threaten by sedimentation and dam failure in Taiwan. Moreover, flooding and water supply problems are also related to the same reasons. Therefore, this study develops a model to optimize the reservoir operation to control the flooding damage, increase the water supply, and improve the sediment sluicing efficiency. This model includes four parts: (1) genetic algorithms (GAs), (2) river simulation model, (3) artificial neural network (ANN) model, and (4) reservoir turbidity dynamic model. This developed model is applied to Zengwen Reservoir and is tested using multiple historical storm events. A multi-objective function is adopted. These objectives include flood control, water supply, and sediment sluicing. The weighting coefficients for each objective is also studied and discussed. The results show better operation efficiency in terms of flooding control and sediment sluicing. For example, the maximum river depth is reduced by 1.8m for Typhoon Krosa and flooding is completely avoided for Typhoon Sinlaku. The sluicing sediment increases 33%, which is about 1,130,000 tons, for Typhoon Krosa and increases 25%, which is about 400,000 tons, for Typhoon Sinlaku, respectively. These results show the efficiency and capability of the proposed model.