Study On The Regulation Mode Of The Three Gorges Reservoir In Drawdown Period For The Protection Of Four Major Fish Spawning

The construction of the Three Gorges project has brought a negative impact that can not be ignored on the water ecosystem in the middle and lower reaches of the Yangtze River.The use of reservoir operation,a non engineering measure,is an important means to carry out the practice of water ecosystem protection and restoration.Ensuring the sustainability of fish,a key aquatic organism,is a key link in the restoration practice.In recent years,in order to promote the spawning and reproduction of fish aquatic organisms dominated by four fish,the Three Gorges reservoir has carried out many ecological regulation experiments and achieved certain results,but it is still faced with problems such as weak research foundation and unclear regulation strategy.The research and practice of ecological regulation is still in the primary stage.In order to explore the specific regulation mode of spawning protection of four major fish during the ebb and flow period of the Three Gorges reservoir,this paper takes "demand—scheduling—decision" as the main line and carries out research work.The specific contents are as follows:(1)Aiming at the problem that the environmental demand for spawning resources of the four major fish species has not formed a unified quantitative value.Through literature research,this paper summarizes the relevant indicators and recommended ranges(values)of the existing literature on the environmental requirements of the spawning resources of the four major fish.At the same time,it summarizes the ecological regulation experiments carried out in the Three Gorges from 2011 to 2019 to promote the spawning of the four major fish.Based on the literature and the practical results of ecological regulation,a set of ecological regulation flood rules for the protection of the spawning of the four major fish is proposed,mainly involving the duration of flood Daily increase of flow,total increase of flow,minimum rising flow and maximum rising flow.On the other hand,based on the Three Gorges reservoir regulation and other documents,the current operation mode of the Three Gorges reservoir is summarized to support the follow-up research.(2)The traditional scheduling process optimization method is difficult to obtain the scheduling process that accurately reflects the ecological water demand,and the rule scheduling is difficult to accurately determine the optimal starting time of ecological scheduling.This paper deduces the mathematical process of the coupling between the dispatching process and the dispatching rules,and then puts forward the cooperative optimization method of the dispatching process and the dispatching rules.The ecological dispatching rules reflecting the rising water process required by the spawning of four large fish are coupled to the reservoir optimal dispatching model with the maximum power generation as the goal,and an ecological dispatching model of the cooperative optimization of the dispatching process and the dispatching rules is constructed.Taking the Three Gorges Project as an example,according to the 1959～2008 long series of inflow runoff data and the proposed flood regulation,the dispatching calculation is carried out.The conclusion shows that the ecological dispatching model can better complete the flood process required for the spawning of four fish;The flood duration is 7 days and the daily flow increase is 3000 m~3/ S this water rising rule has not been successfully implemented in 50 years.The frequency of successful implementation of the other eight water rising rules in 50 years is 90%,90%,86%,90%,82%,52%,90% and 62%.The implementation of ecological dispatching can completely control the loss of power generation within 3%.(3)Aiming at the problem that it is difficult to determine the starting time of ecological regulation for the protection of spawning of four large fish in practice and what kind of ecological regulation flood rules to start.In this paper,three machine learning models,k-nearest neighbor,support vector machine and convolution neural network,are used to obtain the running state sample set and start-up state sample set according to the long series scheduling calculation results and the incoming runoff in the next 7 days.The two sample sets are used to support the identification of the optimal start-up time of ecological regulation and the identification of the flood rules of ecological regulation.The results show that the AUC values corresponding to the ROC curves of k-nearest neighbor,support vector machine and convolution neural network are 0.89,0.91 and 0.95 respectively,and the accuracy of the test set is 82.88%,85.60 and 89.11%.Convolution neural network is recommended to support the decision-making of ecological dispatch;In the judgment of ecological regulation flood rules,the kappa values of the three models of k-nearest neighbor algorithm,support vector machine and convolution neural network are 017,0.46 and 0.66 respectively,and the discrimination accuracy of the three models on the test set are 27.69%,52.58% and 70.77% respectively.It is recommended to give priority to convolution neural network in the judgment decision of ecological regulation flood rules,and use the prediction results of support vector machine model to verify.