Prediction Of Medium And Long-term Incoming Water In Reservoirs Of Water Conservancy Power Stations

With the development of social economy and the increasing consumption of resources,energy issues have become an important issue that needs to be solved urgently.As a new type of energy,water energy has the characteristics of being clean,pollution-free,cheap,and renewable.Economic development plays an increasingly important and important role.As an important facility for the development of hydropower,hydropower stations have seen more and more influencing factors in their reservoir dispatching work.The significance of water supply prediction in hydropower plants is mainly based on economic benefits and safety benefits.Only accurate water supply forecasting can make a good decision for power generation operation.Water forecasting is an important indicator for evaluating the operation and scheduling of hydropower plants,and improving its forecasting accuracy is an important reference for the economic benefits of hydropower stations and comprehensive safety benefits.value.In this paper,a comprehensive linear prediction method and a cyclic neural network time series prediction method are proposed to solve the problems of water supply prediction in hydropower stations.According to the existing prediction methods,several prediction models are integrated to propose a new idea for mediumand long-term water prediction in hydropower station reservoirs,and the advantages of existing algorithms are fully integrated to achieve higher-precision prediction of long-term sequences.In order to further verify the actual performance of the proposed method,and further explore the impact of the sequence of prediction methods on the results,nine groups of models were constructed and tested.Firstly,a single LSTM,ARIMA and SVM models were constructed,and the real water data from the Xiaoxi hydropower to Zhexi hydropower stations were used for prediction.Then two methods are selected,one method is used as the basis of the other method,and six groups of models are constructed for each case.Finally,the prediction results of the single model and the mixed model are compared based on the mean square error.And draw the final conclusion.