應用類神經網路於石門水庫防洪減淤操作最佳規劃模式之研究

Abstract

台灣每年遭受颱風侵襲之機率相當高，颱風豪雨所造成的大量洪水及土石流，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。 本研究以石門水庫為研究區域，整合遺傳演算法、河道水理演算與類神經網路等模式，發展水庫防洪減淤最佳規劃操作模式。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，水庫操作則以石門水庫運用要點之規定為依據。 由結果發現本研究所發展之水庫防洪減淤最佳規劃操作模式，相對於歷史操作結果，對減少庫區淤砂較為顯著，以各場颱風為例，薔蜜颱風本模式可將排砂量由48.08萬噸增至65.48萬噸，約增加36%，鳳凰颱風的部分，本模式可將排砂量由26.61萬噸增至38.41萬噸，約增加44%，對辛樂克颱風而言，本模式可將排砂量由178.51萬噸增至274.89萬噸，約增加54%。整體而言，模擬的結果顯示本研究發展之水庫防洪減淤最佳規劃操作模式，可有效提升排砂量，延長水庫壽命，達永續經營之目的。

In Taiwan, frequent typhoon events result in flood related damages such as reservoir sedimentation, dam failure, and uncertain water supply. 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. An optimal flood control model is developed using Genetic algorithms, a river simulation model, and an Artificial Neural Network (ANN) model. This developed model has multiple objectives including flood control, water supply, and sediment sluicing. Shimen Reservoir is selected for this study. Three historical typhoon events are used: Typhoon Jangmi, Typhoon Fung-Wong, and Typhoon Sinlaku. The results show extraordinary operation efficiency improvement in terms of flooding control and sediment sluicing. This model increases the sluicing sediment efficiency for 36%, or 174,000 tons; 44%, or 118,000 tons; and 54%, or 96300 tons for Typhoon Jangmi, Typhoon Fung-Wong, and Typhoon Sinlaku respectively. These result shows that the developed model can be a very useful tool for optimal flood control operation for reservoirs.