Study On Operation Rules Of Cascade Reservoirs Based On Stochastic Dynamic Programming

Water energy resources are a type of pollution-free and recyclable energy.The use of water energy for power generation can reduce damage to the ecological environment.When developing and using water energy resources,there are also huge benefits for power generation,flood control,irrigation,shipping,water supply,and other aspects.This article takes the Longyang Gorge-Liujia Gorge cascade reservoirs in the upper reaches of the Yellow River as the research object.Firstly,three prediction models,namely BP neural network,RBF neural network,and MLR multiple linear regression,are used to predict the medium and long-term runoff of the Longyang Gorge reservoir in monthly intervals;Secondly,historical runoff is simulated using Copula function and Gibbs sampling,and the transfer probability matrix between adjacent months is constructed using Markov chains;The interval range of inflow runoff is divided by mathematical statistics,and a stochastic reservoir optimal operation model is constructed;Extract stochastic dynamic programming(SDP)scheduling rules;Finally,develop a reservoir scheduling plan based on runoff forecasting and scheduling rules.The research content and achievements of this article are as follows:(1)Selection of runoff forecast model.Based on the historical monthly runoff data and rainfall data of Longyang Gorge Reservoir.The runoff and rainfall data from 2011 to 2015 were selected for grey correlation analysis,and the eight prediction factors with the highest correlation degree with the inflow runoff of Longyang Gorge Reservoir were screened out.The BP neural network,RBF neural network and MLR multiple linear regression runoff prediction model were established.The model parameters were trained,and the deterministic coefficient,mean square deviation and average relative error of the model during the training and test periods were counted.The prediction results show that the three models have good effect in the medium and long term runoff prediction of Longyang Gorge Reservoir,with high accuracy and small error.The prediction results are of great significance for the hydrological prediction of Longyang Gorge Reservoir.(2)Taking the inflow runoff of Longyang Gorge Reservoir as the research object,four common univariate distribution functions are used to fit the probability curve of monthly runoff density.Based on the Copula function fitting,the optimal distribution of each month is optimized.Combining the Gibbs sampling principle,the optimal Copula function between adjacent months is converted into conditional function,and the monthly runoff of 10000 years is randomly generated.The reliability of runoff is tested by correlation coefficient.(3)Aiming at the uncertainty of incoming water,based on the monthly runoff simulated by Copula function,the transfer probability of incoming runoff is described by Markov chain and transfer matrix method.Divide the months generated by simulation into different levels,calculate the transfer probability of different levels of incoming water in each month,and build the stochastic optimal operation model of Longyang Gorge-Liujia Gorge cascade reservoirs in the upper reaches of the Yellow River with the power generation as the target.The SDP algorithm is used to solve the problem and obtain the reservoir dispatching rules of Longyang Gorge-Liujia Gorge hydropower station containing the incoming water information in the future.The results can provide a direct reference for the optimal operation of the combined generation of Longyang Gorge-Liujia Gorge hydropower station;Taking the measured runoff of a certain year as an example,the annual power generation based on SDP and DP algorithm is very close,which verifies the reliability of the dispatching rules constructed in this study.Taking the runoff predicted by the runoff forecast model as the input,and combining with the constructed regulation rules,the reservoir regulation plan is formulated.