Research On Two-stage Aggregation Decomposition Method To Derive Operation Rules For Hydropower Stations

Hydropower is an important clean and renewable energy to support China to achieve energy transformation,energy conservation and emission reduction goals.With the rapid development of hydropower in China,a number of large-scale cascade hydropower stations have been built in the southwest region where hydropower resources are rich.To further tap the compensation and regulation capacity between cascades to improve the efficiency of the whole hydropower system has become a key problem to be solved.Through the long-term optimal operation of hydropower stations,the operation rules can be derived to guide the actual operation of hydropower system,which can help decision-makers to make reasonable decisions under the condition of knowing the hydrological information of the current period in order to obtain long-term stable benefits,which is of great significance.However,due to the uncertainty of inflow,the close hydraulic and power relationship between cascade hydropower stations and the large scale of the system,it is extremely difficult to derive the operating rules of large-scale hydropower systems.Aiming at the problem of deriving the operating rules of large-scale hydropower system,this paper takes the large-scale hydropower system composed of the Lancang River,the Hongshui River and the Wu River cascade power stations in southwest China as the background.The operating rules are derived by using aggregated sampling stochastic dynamic programming and a method combining aggregated sampling stochastic dynamic programming and simulation-optimization respectively.The main research contents are as follows:(1)In order to give full play to the compensation regulation ability of multi-basin hydropower stations and derive the system operating rules to guide the actual operation,an aggregated sampling stochastic dynamic programming method is used to solve the problem.In this method,while using the historical inflow scenarios to explicitly describe the uncertainty of the inflow,the cascade power stations are aggregated into a single power station,and the cascade energy storage is used as the state variable.the computational dimension is greatly reduced to overcome the curse of dimension problem faced by multi-dimensional sampling stochastic dynamic programming.It is applied to a large-scale hydropower system in southwest China,and the results show that this method can quickly find a satisfactory solution.The obtained operation rules,which can be used as a good initial solution for the simulation optimization algorithm,can be further improved.(2)Because of the complexity of reservoir operation in large-scale hydropower system,the curse of dimension problem limits the application of conventional sampling stochastic dynamic programming algorithm in large-scale hydropower system,while the traditional simulation-optimization method can only obtain the local optimal solution in most cases.A novel two-stage algorithm for solving scheduling rules of large-scale hydropower systems is proposed.In the first stage,the initial solution is generated quickly by using the above aggregated sampling stochastic dynamic programming algorithm.In the second stage,multiround genetic algorithm is used to solve the simulation-optimization model.This combination method can improve the quality of the solution without increasing the calculate time too much.The method is applied to the operation of three cascade hydropower systems in southwest China,and the effectiveness and practicability of the method are verified.