Multi-objective Optimization Of Reservoir Operation Considering Multiple Information

The contradiction between water supply and water demand in China is becoming more and more prominent.Water resources management is moving toward to the comprehensive utilization and the optimal coordination development on miltiple objectives including electricity generation,water supply,environment protection,agricultural irrigation and other objectives.Reservoir is the most important control project to realize mutual coordination between multiple objectives.Scientific and reasonable reservoir optimal operation of reservoir is critical to make full use of the reservoir storage capacity,and is also critical to storage and discharge natural runoff in a planned way to meet the demand of the user goals.With the increase of the number of objectives,it is becoming more difficult to rationally optimize on reservoir operation problems.How to optimize efficiently,how to decision reasonably,and how to further improve the benefits of multi-objective reservoir optimization become more and more prominent.Considering that many objectives of reservoir operation have preferences under the influence of national policies and water supply priorities,the use of the preference information can guide for efficient optimization.Considering that the competition exists between multiple targets of the reservoir,the quantification of the competition information is helpful to grasp the competition law of water users and make rational decisions.Moreover,considering that forecast information with higher accuracy and longer forecast period can be obtained,the rational use of the forecast inforemation in multi-objective reservoir operation is helpful to correct the operation decision as soon as possible and improve the operation benefits.Based on the above consideration and analysis,this article takes the Nierji reservoir as a case study and carries out a research on multi-objective reservoir operation considering multiple information.The research which uses the preference information,the competition information and the forecast information focuses on three aspects,that is,the optimization algorithm,the decision making method and the correct use of forecast information in reservoir operation.The main research contents and results are as follows:(1)The Nierji reservoir is described and a universal multi-objective reservoir operation model is constructed.First,some basic data of the Nierji reservoir including physical characteristics of the reservoir,inflow,and water demands are presented.Then,the basic form of operation rule curves is established to guide the reservoir operation,and the multi-objective operation model is constructed to obtain the operation rule curves.This provides the model foundation for the subsequent research.(2)A multi-objective optimization method based on preference information is proposed.First,three preference strategies based on reference point,i.e.,modified crowding degree,r dominance and g dominance are applied to modify standard algorithm NSGA-II.Next,different reference point scenarios are set according to different preference.Finally,Pareto optimal solutions are obtained by the three modified algorithms and the standard algotithm respectively,and the sesearch efficiency and quality of the Pareto optimal solutions are compared on several aspects,including the convergence and diversity,the closeness degree toward the preference region,and the numbers of the preferred solutions.The results show the modified crowding degree and r dominance can guide the search direction according to the reference point,narrow the search space and improve the optimization search efficiency.The convergence and diversity of their solutions in the concerned region are better than NSGA-II,and the closeness degree to the reference point can be increased by 42.8%,and moreover the number of the preferred solutions can be increased by more than three times.However,the convergence of the Pareto optimal solutions is difficult to be guaranteed by g domination.(3)Conflict evaluation indexes for quantitative many-objective tradeoff analysis is proposed and a decision-making method based on the indexes is given.First,based on the Pareto optimal solutions,the 2-D Pareto frontier of each two objectivs is obtained by using non-dominant sorting.Based on the corresponding relationship among the 2-D Pareto frontier,the full Pareto optimal solutions and the general conflict degree of two objectives,a conflict evaluation index called CEI for two objectives conflict degree evaluation is established.Thereafter,a total conflict evaluation index,i.e.,TCEI,which is the summed CEI value of one objective with all other objectives is defined as the total conflict degree of the objective.Next,the TCEI value of each objective is regard as the judgment of importance to determine the corresponding weights for multi-objective decision making.Finally,the indexes and the decision making method are applied in the Nierji reservoir under four water demands scenarios from low to high respectively to quantify the competition relation and recommends reasonable solutions.Results show that the application of the indexes which realizes the quantification of the competition relation among multiple objectives,characterizes the competition variation with water demands change from low to high,and recognizes the objectives with intensive conflict as recommended solutions.(4)A medium-term inflow forecast method considering multiple time series factors is proposed.First,the suitability of a rainfall product and the accuracy of a rainfall forecast product are assessed.Next,multiple time series factors which reflect the process differences and independent factors are screened out from the former rainfall factors,the former inflow factors and the rainfall forecasting factors based on BP neural network.The various factors are combined to predict the future first 10 days' inflow forecasting,and the effectiveness of the proposed method is verified by comparison with the prediction results under the combination of factors with the maximum correlation coefficient.Finally,the future second 10-day inflow forecast factors is determined based on the practical situation of the Nierji basin and the first 10-day runoff forecast factors,and the future second 10-day inflow forecasting is given accordingly.Results show that the future first ten-day runoff forecast accuracy is higher,and the future second 10-day runoff forecast accuracy satisfies the standard for using with the proposed inflow forecast method.(5)The influence mechanism of inflow forecasting uncertainty on reservoir operation is revealed,and a modified application method of medium-term inflow forecasting in reservoir optimal operation is proposed.First,considering that the influence of forecast error on reservoir operation benefit is different due to the difference of the storage capacity and water use target,the influence of two 10-day runoff forecast error on reservoir optimal operation is studied.Next,based on the analysis of the influence of runoff error and the evaluation results of inflow forecast accuracy,the method of using correction coefficient to correct runoff forecast information in reservoir optimal operation is proposed.Finally,the results of reservoir optimal operation with the modified inflow forecasting are compared with those without the inflow forecasting,and without modified inflow forecasting in order to verify its validity.Results show the modified application method can effectively improve the effect of multi-objective optimization of reservoir operation and helps obtain better Pareto optimal solutions.At the same amount of power generation,the total amount of water shortage can be reduced by 72×106m3 compared with the conventional optimization,and the maximum amount of power generation can be increased by 12×106kWh under the same amount of water shortage.Finally,the conclusion and innovation of the paper are summarized and condensed,and the problems to be further studied are prospected.