Construction And Application Of Reservoir Flood Control Forecast Operation Rules Using The Decision Tree Algorithm

In recent years,floods caused by extreme weather have occurred frequently,resulting in huge social and economic losses,especially the 2018 Shouguang flood incident has aroused widespread concern for flood control safety.Reservoir is an important flood control engineering measure,scientific reservoir flood control operation mode is the key to achieve reservoir design benefits,it can play a significant flood control benefits and ensure the sustainable development of social economy.Conventional operation and optimal operation are two kinds of methods for reservoir flood control operation.The former lacks consideration of the latest operation data and the complex nonlinear relationship between discharge and its influencing factors,and it is difficult to obtain the optimal operation scheme.The latter encountered many difficulties in practical application because of the problems such as prematurity caused by local fast convergence,poor local search ability caused by a large number of global search and long iterative time.Therefore,this paper takes Bashan Reservoir in Shandong Province as the research object,and integrates random simulation technology,modeling technology,multi-objective fuzzy optimization model and data mining technology into the design of reservoir flood control operation rules.a method of using decision tree C5.0 algorithm to extract reservoir flood control operation rules from flood operation data is proposed and evaluated.The main results are as follows:(1)taking the upstream watershed of Bashan Reservoir as the research object,the control forecast and ensemble forecast of five forecasting models ECMWF,NCEP,CMA,JMA and UKMO in TIGGE system are selected.From the point of view of rainfall classification forecast test and rainfall forecast test,as well as Brier score and Talagrand distribution,the accuracy of control forecast and ensemble forecast data from May to September 2007 to 2019 is analyzed.The results show that the prediction effect of UKMO and NCEP is the best.Based on this study,high-precision rainfall forecasting products can be selected to serve flood control operation.(2)since few historical flood operation data have been collected,stochastic simulation technology and hydrological model are combined to simulate a large number of flood processes and optimize their operation.First of all,on the basis of 19 measured flood processes of Bashan Reservoir,eight groups of feasible parameter schemes of Xinanjiang model of Bashan Reservoir are optimized by SCE-UA algorithm,and the optimal parameter scheme is selected by multi-objective fuzzy optimization model,based on which the historical flood is simulated to verify its accuracy.Then,the stochastic simulation technique is used to generate 104 rainfall processes with different frequencies,which are substituted into the Xinanjiang model to obtain the flood process.Finally,the particle swarm optimization algorithm is used to solve the optimal operation model of flood control in Bashan Reservoir.The results show that the optimal operation scheme meets the control and operation mode of the reservoir and is reasonable.(3)the flood operation data set is generated from the optimal operation scheme of Bashan Reservoir,and the decision tree c5.0 algorithm is substituted to generate the decision tree of operation rules of Bashan Reservoir,which is used to dispatch four measured floods.The comparison between the measured scheduling scheme and the decision tree scheduling scheme shows that the scheduling scheme made by the decision tree is feasible,and in some cases,it is better than the measured scheduling scheme.