Research On Three Gorges Reservoir Operation For Non-flood Season Considering Runoff Forecast Uncertainty

Reliable reservoir inflow forecast is the basis of scientific long-term reservoir scheduling decision.However the forecast horizon of medium and long term flow forecast model is limited and flow forecasting is uncertain,Quantitative description of runoff forecasting uncertainty and analysis of the influence on reservoir decision making are the basis of proposing more perfect,practical and reliable scheduling strategies.In this paper,the optimal operation of reservoirs under the condition of runoff forecast uncertainty is studied,the sources and correlations of runoff forecast uncertainty are analyzed,and the stochastic simulation method of forecast uncertainty is put forward.Build the multi-objective optimal dispatching model of reservoir and propose the solution method.The optimal scheduling scheme group under the uncertainty condition is obtained,the influence of the forecast uncertainty and the error on the decision of the reservoir is analyzed,and the optimal operation strategies of the reservoir are put forward.The main contents of the study are as follows:Firstly,the paper systematically summaries suppression algal blooms strategies of the Three Gorges Reservoir in near-dam tributary.Based on the proposed suppression strategies,blooming suppression scheduling criteria is utilizing the inflow to achieve storage scheduling as far as possible when meeting the minimum discharge which is required of the downstream ecology and shipping.Secondly,the characteristics of the monthly runoff series after the mutation are analyzed.Based on the artificial neural network,propose the monthly runoff forecast model of the Three Gorges Reservoir in the middle and late non-flood season,roll operation forecast and analyze the correlation between the relative error sequences.The stochastic simulation method of runoff forecast uncertainty based on Copula function is proposed.The results show that forecast model for reservoir monthly runoff based on artificial neural network with runoff in the last four continuous months as the inputs can effectively obtain monthly runoff in non-flood season and can be used to scroll to job forecast.The four-variables t-Copula function can describe the correlation between prediction relative error sequences well,the gap of correlation coefficients,statistical characteristic values and empirical distributions between the simulated sequences and the observed sequences is small,the stochastic simulation model can be used for describing the uncertainty of runoff forecast.Thirdly,based on the determined criterion of restraining algal blooms,considering the constraints of reservoir water balance,reservoir water level,discharge capacity and power station output,this paper constructs the ten-time scale multi-objective optimal dispatching model of the Three Gorges Reservoir in the middle and late non-flood season to maximize the power generation and suppression algal blooms,and to solve the model,the Genetic Algorithm is combined with Pareto optimality.Fourthly,select the typical dry,normal and wet inflow situations,and use the constructed stochastic simulation model of runoff forecast uncertainty to simulate the monthly runoff process in the mid-late period of non-flood season as the inputs of the optimal dispatch model and get all the possible solutions,according to the classification of forecasting flow grade,the paper analyzes the influence of uncertainty and error of runoff forecasting on reservoir scheduling decision.Finally,the paper proposes the non-flood season scheduling strategy of Three Gorges Reservoir,that is to say,under the conditions of meeting the requirements of downstream ecological water demand and navigation,it is suggested that the Three Gorges Reservoir adopt the precipitation level operation in March and May 10 to June 10 at the ten-day scale.When the forecast error of the general flow from March to June and the forecast error in March and April are large,that is,the total forecast flow is greater than and close to the actual flow,and forecast flow of March and April is larger,take the lift level operation in April at ten-day scale,and take the precipitation level operation from May 1 to May 10 at ten-day scale;When the forecast error of the general flow is large and the error of the flow forecast in March and April is small,that is,the total forecast flow is larger than and close to the actual flow and the forecast flow in March and April is smaller,from April 1 to April 20 and from May 1 to May 10,the water level is to be taken lift according to ten days and from April 20 to April 30,the water level is to be taken down at ten days.From March to June,when the forecast error of the general flow is small,that is,the total forecast flow is less than the actual flow,in April to take precipitation level operation,May 1 to May 10 to take lift level run.When the forecast error of flow is relatively large,the corresponding forecast flow is more,and the rise and fall range of each ten-day water level is more greatly.The Three Gorges Reservoir can be operated in accordance with its typical situation and operational forecast flow corresponding to the typical situation in this paper and its predicted flow grade,with reference to the average water level and discharge of the class and the water level rise and fall regularities summarized.