Research On Multi-objective Detailed Optimal Operation And Decision-making Methods Of Cascade Reservoirs In The Middle And Lower Reaches Of Yalong River

Cascade reservoir is an important engineering measure to develop and utilize clean renewable energy-hydropower resources.Through the planned regulation and storage of runoff in a period of time,cascade reservoirs can fulfill many important tasks such as prevention of flood and drought disasters,benefit growth of hydropower enterprises,safe and stable operation of power grid,etc.In recent years,China has introduced a series of policies to encourage the consumption of clean energy.The hydropower industries supported by hydropower resources have become the key to the transformation of China’s energy structure.At present,with the gradual completion of of the thirteen hydropower bases,such as Jinsha River,Yalong River and Dadu River,the scale of cascade reservoirs in each basin is expanding,and the focus of hydropower development is transferred from construction to operation management.The concept of intelligent water conservancy and related work highlight the importance of strengthening the management of water engineerings represented by cascade reservoirs.Therefore,it is urgent to carry out the research on multi-objective optimal operation and decision-making method of cascade reservoirs.It is expected that under the complex external environment and engineering background,the scheduling schemes for maximizing the utilization rate of water resources in cascade system can be developed,the compensation and regulation capacity between the upstream and downstream reservoirs can be fully explored,and the multiple objectives of the cascade system can be effectively coordinated.This paper gives full consideration to the theoretical research and practical production of cascade reservoirs optimal dispatching,and focuses on the problems of precise operation,the treatment of dimensional disaster,multi-objective operation and multi-attribute decision-making.Based on the theories and methods of mathematical programming,parallel acceleration,intelligent optimization,probability statistics and so on,the model construction,algorithm improvement and operation scheme decision of short-term and mid-long-term optimal operation are studied deeply,and the following main achievements are obtained:(1)Daily power generation planning for cascade hydropower stations under the requirement of precise management.In view of the calculation of optimal output of the traditional models is not accurate enough,resulting in the deviation of the scheduling scheme in the actual implementation,a precise optimization operation model of inter-plant and intra-plant integrated operation is established to achieve the unification of the optimal distribution of hydropower resources between and within power stations.On this basis,a nested multidimensional dynamic programming is proposed to solve this nested optimization model.And a case study of the Jinxi-Jindong cascade system in the Yalong River basin is carried out to verify the superiority of the refined optimization model and the corresponding algorithm.(2)Performance optimization of the algorithm based on memory footprint reduction and GPU parallel acceleration.In view of the serious dimensionality curse of nested dynamic programming in solving precise optimization operation model,data compression and database technology are used to effectively reduce the memory taken by the calculation program,and the computational efficiency is greatly improved by GPU parallelism through the OpenACC standard.Then,an improved nested dynamic programming is proposed to solve the dimensionality curse.By monitoring the changes of memory consumption and computation time before and after the introduction of the optimization strategies,the proposed improvement strategies are both reasonable and effective.(3)LMPSO algorithm and its application in multi-objective optimal operation of cascaded reservoirs.For the defect of multi-objective dimensionality reduction into single-objective optimization and the deficiency of classical multi-objective evolutionary algorithms in solving large-scale high-dimensional problems,the performance evaluation index,the hypervolume indicator,is introduced as the selection standard,and the problem transformation strategy is adopted to reduce the dimension of the search space,thus effectively reducing the difficulty of solving the multi-objective optimal operation models.On the basis of SMPSO,LMPSO algorithm is proposed,and the improved algorithm is applied to the three-reservoir joint operation of Yalong River,which verifies the computational advantages of the proposed algorithm compared with other methods in this large-scale high-dimensional problem.(4)Improved interval number grey target decision model and its application in operation scheme decision.Considering that there is forecast error existing in the inflow process of cascade reservoirs,it is more reasonable to express the index value of the operation schemes by interval number,so the grey target decision model with the interval number theory is selected to make the operation scheme decision.Based on the traditional interval number grey target model,the weight vector determination based on set value theory and the expected closeness calculation based on multi-dimensional joint sampling are designed,and the corresponding improved model is proposed.The standard and the improved decision model are used to realize the multi-attribute decision-making for the cascade reservoirs of Yalong River,and the results verify that the improved method is more reasonable in dealing with the interval number and can effectively avoid the information distortion in the calculation process.