Research On Dynamic Control And Decision-making Of Cascade Reservoirs' Water Level In Flood Season Based On Risk Sharing

China is rich in hydropower resources,but there are significant differences in water resources distribution among seasons.Reservoir is an effective engineering measure to tap the potential of hydropower resources and alleviate the uneven distribution of water resources in the region.Reservoirs are often faced with the arduous task of flood control in flood season,which requires the operation water level of reservoirs to be controlled below the flood control limit water level,which is in sharp conflict with the demand of power generation and water supply.With the global warming,the meteorological and hydrological conditions in each basin have changed significantly,and with the improvement of dispatching technology,risk analysis capability and emergency response mechanism,the flood limit water level designed in the planning stage can no longer meet the comprehensive utilization requirements at present.Under the condition of controllable flood control risk,raising the operating water level properly in flood season is of great practical significance for improving the comprehensive utilization benefit of reservoirs and realizing the efficient utilization of hydropower resources.Taking Xiluodu-Xiangjiaba cascade reservoirs in Jinsha River basin as the research object,this paper comprehensively uses black box model,big data,intelligent algorithm,prospect theory and other methods,and focuses on flood control and power generation.Based on the "spatial risk sharing" model for calculating the dynamic control domain of cascade reservoirs' combined operation water level,the flood control risk and power generation benefit of the combined scheme of cascade reservoirs' dynamic control water level are further analyzed,and the scheme is optimized through the multi-objective group decision-making model,thus realizing the systematic combination of forecasting,dispatching,risk analysis and decision-making,aiming at improving the power generation benefit of cascade reservoirs in flood season under the condition of controllable risks,perfecting the theory and method of cascade reservoirs' water level control in flood season operation,and providing technical support for stable and safe operation of reservoirs.The main achievements are as follows:(1)Medium-and long-term runoff forecasting model based on runoff division and forecasting factor screening.Firstly,MK test and RS test are used to analyze the change trend of Xiluodu historical inflow runoff series.A runoff division method based on runoff series feature clustering is proposed.The historical runoff is divided into three typical categories by K-means clustering method.According to the characteristics of runoff to be predicted,the previous runoff series of the corresponding category is taken as the forecasting factor,and the highly correlated forecasting factors are screened out by MIC method as the input of BP artificial neural network.Compared with traditional forecasting methods,the above methods can improve the quality of neural network input side and improve the accuracy of medium and long-term runoff forecasting.(2)Dynamic water level control and risk analysis method of cascade reservoirs in flood season based on spatial risk sharing.Aiming at the problem that the asynchronous storage of cascade reservoirs may lead to the occurrence of system risk ahead of time,the principle of "equal proportion storage" is put forward to optimize the flood control storage capacity allocation mode of cascade reservoirs to reduce system risk.On the basis of Xiluodu-Xiangjiaba flood regulation rules,considering the principle of "equal proportion of water storage",the dynamic control domain of water level for joint operation of the two reservoirs in flood season is deduced,and then the combination schemes are worked out,and the flood control risk analysis of different combination schemes is carried out,which lays a foundation for realizing flood resource utilization.(3)Combined optimal operation model of cascade reservoirs based on improved electronic search algorithm.Taking different water level combination schemes of Xiluodu-Xiangjiaba in flood season as constraints,the optimal operation model of joint generation of two reservoirs is established.In view of the problems of over-limit search space and low search efficiency of electronic search algorithm in solving cascade reservoir optimal operation problem,this paper puts forward a feasible domain search strategy to ensure that individuals in each iteration are feasible solutions,and adopts parameter adaptive method to improve the early global search speed and later local search ability of the algorithm.The superiority of the algorithm in solving efficiency is verified.It is applied to solve the optimal dispatching model of Xiluodu-Xiangjiaba combined power generation,so as to optimize the dispatching process and obtain higher power generation benefits.(4)Group decision making model with maximum satisfaction of expert group based on cumulative prospect theory.The combination weighting optimization method is used to obtain the index weight which takes into account the ranking degree and importance degree of the index;The individual decision-making results close to the actual decision-making psychology are obtained through the cumulative prospect theory.On this basis,a group decision-making model is established according to the principle of maximum satisfaction of experts,and the most appropriate scheme for individual decision-making results of all participating experts is solved as the group decision-making result.The optimized scheme weighs the risks and benefits,and can provide reference for the efficient utilization of water resources.