Research on Data Mining-Based Precision Flood Control Scheduling Strategy for Reservoirs

The accurate assessment of the relationship between reservoir outflow and downstream floods is often challenging in flood control scheduling of upstream reservoirs aimed at downstream flood protection. In this research, the Fengshuba Reservoir in the Dongjiang River Basin, China, is taken as the subject of study. Utilizing a dataset encompassing 62 years of daily measured flood processes, the MIC coefficient is employed to determine the correlation between the reservoir outflow process at different lag times and the flow at the downstream section. The flood propagation time is determined by identifying the lag time associated with the maximum MIC value. By utilizing the BPANN model, which incorporates the reservoir outflow process and the interval flood process as inputs, an accurate prediction of the downstream flood process is achieved, resulting in a closer approximation to reality in flood estimation at the downstream section. The model has been validated in the district between Fengshuba and Longchuan, exhibiting a certainty coefficient of 97% and a prediction qualification rate of nearly 90%. In comparison with the conventional Maskingen evolution method, the calculated outcomes provide enhanced support for flood control safety, enabling precise hourly control of downstream flood processes and upstream reservoir outflow processes.