Multi-objective Cascaded Reservoirs Optimal Operation Rule Considering Streamflow Uncertainty

With the rapid development of society and economy,hydropower energy has played a major role in improving the energy structure,dealing with with climate change,and achieving the "carbon peak and carbon neutral" planning objective.Scientifically optimizing the management of cascade reservoirs in the river basin is an effective way to improve the utilization rate of water resources and reduce coal consumption.The optimal operation of cascade reservoirs in a river basin is affected by many factors such as hydrological forecast uncertainty,stable operation of power grid,flood control,water supply demand and other factors.It is a high-dimensional,multi-objective,multi-constrained nonlinear optimization problem with uncertainty.Therefore,this article focuses on the key scientific and technical problems faced by the identification and optimization of multi-objective reservoir operation rules under the uncertainty of inflow forecasts,and aims to improve the comprehensive utilization of water resources of cascade reservoirs through hydrometeorology,probability statistics,machine learning technology and multi-objective optimization theory.The main research contents and achievements of this paper are as follows:(1)To solve the problem of the quantification of the uncertainty of streamflow forecasting,this paper intriduces the hidden sate of streamflow,which can be regarded as dry or wet climatic conditions,and establishs the streamflow hidden Markov model.For each hidden state,there is a corresponding observation probability model,which is a joint Gaussian model.Then,the paper extends the Gaussian mixture regression method in the hidden Markov model,and propose the hidden Markov regression probability forecast method.The hidden Markov regression probability forecast method is applied to the monthly streamflow probability forecast.The experimental results show that the method has excellent performance with high accuracy and high reliability,and can effectively quantify the "heteroscedasticity" and "Nonnormality" the of the streamflow forecasting uncertainty.(2)Reservoir operation rules play a key role in real-time reservoir operation,the main factors affecting operation decisions are the current reservoir status and the future inflows.To study the influence of inflow uncertainty on reservoir operation rules,this paper proposes a Bayesian Deep learning method that considers both model parameter uncertainty and inflow uncertainty.In the model,the Monte Carlo integration is used to convert the complex integrals of inflow probability into a summation form.Variational inference is employed to obtain the posterior distribution of model parameters.The proposed method is applied to a real-world application of Three Gorges-Gezhouba cascade reservoir on the Yangtze River.The results show that the extracted dispatching rules can not only provide deterministic operation decisions,but also provide decision water level intervals under different confidence intervals.The proposed method can effectively reduce the impact of inflow forecast uncertainty on operation decisions,and provides richer and more reliable infoamation for cascade reservoirs operation.(3)To solve the "dimensionality disaster" problem faced by traditional multi-objective algorithms when solving high-dimensional object optimization problems.This paper paoposes an evolutionary algorithm based on a region search strategy(RSEA)to deal with different kinds of benchmark problems.In the proposed algorithm,each solution is associated with a region,and the region search strategy is applied to constrain the updating process;this strategy will enhance the diversity of population without losing convergence.A series of test functions with different characteristics are used to verify the algorithm.The experimental results show that the region search evolutionary algorithm can effectively deal with multi-objective optimization problems such as linear,mixed concave/convex,multimodal,discontinuous,biased,and non-normalized.and obtains good convergence.It has high computational efficiency while obtaining good convergence and diversity,which is proved to be a multi-objective evolutionary algorithm with excellent performance.(4)Focusing on the key issue of optimization of multi-objective operation rules for cascade reservoirs,the papper proposes a hierarchical flood operation rule.Based on the rule,a multi-objective flood control optimal rule model considering flood control,power generation,and shipping is established.In addition,by comprehensively considering inflow forecasts and their uncertainties,this paper proposes a hierarchical pre-discharge flood operation rule and establishs a multi-objective hierarchical pre-discharge flood operation rule optimization model which considering flood control objective for upstream reservoirs,flood control objectives for downstream reservoirs,power generation objective,and navigation objective.The RSEA is used to solve the two models,a set of non-inferior operation rules is obtained.The results prove that the optimal flood operation rules can greatly improve the power generation during the flood period without under the flood control standers,which provides strong technical support for the actual flood control decision.