Research On Comprehensive Utilization And Regulation Rules Extraction Of Cascading Reservoir Groups In The Falling Period

In recent years,with the continuous construction and commissioning of cascade reservoirs in river basins,the main task of China’s water conservancy projects has shifted from the construction of reservoir projects to the scheduling and management of cascade reservoirs in river basins.Cascade reservoirs need to take into account the multiple dispatching needs of power generation,water supply and shipping in the basin during the fading period.How to reasonably allocate the limited hydropower resources of the cascade reservoirs during the fading period to meet the water demand of various aspects of the river basin has become one of the important contents of reservoir dispatch management.The traditional reservoir declining schedules mostly aim to determine the maximum power generation in the incoming water process,and can no longer meet the scheduling needs of multiple incoming water scenarios and multiple optimization targets under today’s complex constraints.There is an urgent need to carry out research on the comprehensive utilization of cascading stages and the extraction of dispatching rules in order to give full play to the capacity compensation benefits of upstream and downstream cascade reservoirs.Therefore,based on the multi-objective optimization theory and data mining theory,this paper takes lower Jinsha River cascade as the research area,and researches on the selection of cascading reservoirs’ timing,multi-objective comprehensive utilization and extraction of reservoir scheduling rules.The main research contents and innovative achievements are as follows.(1)Aiming at the problem of water abandonment caused by the gradual fading of cascade reservoirs before flood,the study first gives the combination of all the timing of cascading reservoirs.Then,for each combination of cascade fading opportunities,cascade simulation scheduling is carried out by using the control method of high water level operation of the reservoir in the early stage of fading,and evenly fading to the limited water level in the later stage.Then,the cascade power generation amount and the waste water amount corresponding to the set fade time are obtained.Finally,the weighting method is used to obtain the best combination of the timing of each cascade library under the three different goals of maximum power generation,minimum water consumption,and cascade power generation and waste water under different incoming water scenarios.Provide some reference for the actual fading of the cascade.(2)In response to the multi-purpose comprehensive utilization needs of the cascade reservoirs during the decline period,a multi-objective dispatch model considering the power generation benefits,shipping benefits,and power generation stability of the cascade reservoirs during the decline period was established.After that,the multi-objective genetic algorithm NSGA-Ⅲ was used to solve the multi-objective model,and the non-inferior solution front of the model was obtained.Then,according to the distribution trend of the non-inferior solution frontier in the target space,the competition between the scheduling targets in the cascading stage is analyzed.After that,the relationship is judged according to the key scheduling indicators selected by the scheduling decision maker and the mutual importance between the indicators,AHP is used to prioritize the non-inferior solution frontiers of multi-objective scheduling to obtain multiple relatively better solutions with higher rankings.For decision-makers to make reference choices when they disappear.(3)Limited by the accuracy of mid-long-term runoff forecasting,the deterministic multi-objective optimal dispatching results of the cascade reservoirs’ dip period are difficult to apply directly in actual dispatching work.Research based on implicit stochastic optimization theory and data mining theory,using support vector machine model(SVM)to extract rules from long history series of multi-objective optimal scheduling results.Then use genetic algorithm(GA)to optimize the parameters of the SVM model.Obtain the reservoir scheduling rules hidden in the long series of samples.When setting the independent variable factor of reservoir scheduling rules,in addition to the initial water level and the flow into the reservoir,two energy factors of cascade energy input and cascade energy storage are added in this paper.Characterize the effect of the energy state of the cascade on the current reservoir decision.The fitting results show that the two additional step energy factors can significantly improve the fitting accuracy of the SVM model.Applying the GA-SVM method proposed in this paper to the extraction of scheduling rules in the cascade of the lower reaches of the Jinsha River,and using historical runoff to simulate scheduling.The results show that the simulated water level process of the cascade reservoir is very close to the optimization results,indicating that the extracted scheduling rules have high accuracy.It can well guide the actual decline and dispatch of cascade reservoirs.