Application of Decision Tree to Obtain Optimal Operation Rules for Reservoir Flood Control Considering Sediment Desilting-Case Study of Tseng Wen Reservoir

Abstract:

Reservoirs are the most important water resources facilities in Taiwan.However,due to the steep slope and fragile geological conditions in the mountain area,storm events usually cause serious debris flow and flood,and the flood then will flush large amount of sediment into reservoirs.The sedimentation caused by flood has great impact on the reservoirs life.Hence,how to operate a reservoir during flood events to increase the efficiency of sediment desilting without risk the reservoir safety and impact the water supply afterward is a crucial issue in Taiwan.

　Therefore,this study developed a novel optimization planning model for reservoir flood operation considering flood control and sediment desilting,and proposed easy to use operating rules represented by decision trees.The decision trees rules have considered flood mitigation,water supply and sediment desilting.The optimal planning model computes the optimal reservoir release for each flood event that minimum water supply impact and maximum sediment desilting without risk the reservoir safety.Beside the optimal flood operation planning model,this study also proposed decision tree based flood operating rules that were trained by the multiple optimal reservoir releases to synthesis flood scenarios.The synthesis flood scenarios consists of various synthesis storm events,reservoir’s initial storage and target storages at the end of flood operating.

　Comparing the results operated by the decision tree operation rules(DTOR) with that by historical operation for Krosa Typhoon in 2007,the DTOR removed sediment 15.4% more than that of historical operation with reservoir storage only8.38×10⁶m³ less than that of historical operation.For Jangmi Typhoon in 2008,the DTOR removed sediment 24.4% more than that of historical operation with reservoir storage only 7.58×10⁶m³ less than that of historical operation.The results show that the proposed DTOR model can increase the sediment desilting efficiency and extend the reservoir life.