LightGBM-Based Short-Term Inflow Forecast And Reservoir Dispatching System Research And Development

In recent years,China's natural disasters have occurred frequently,especially floods,which have caused major threats to the personal safety of the people and have a major impact on China's economy.Therefore,flood control and dispatching work are particularly important.Short-term inbound flow is an important basis in flood control and dispatch,and its prediction accuracy directly affects the effectiveness of flood control and dispatch.In order to reduce flood hazards and achieve the purpose of flood prevention and benefit,it is necessary to improve the accuracy of short-term inbound flow prediction.In view of the low prediction accuracy of the current main model when processing data in Changde,Hunan,the loss function huber loss of the machine learning algorithm light GBM is improved by increasing the regularization factor.The model is applied to the real data of a reservoir in Changde,Hunan Province,to predict the short-term inbound flow of the reservoir.The model is compared with the current popular short-term inbound flow prediction models such as SARIMA,LSTM and AN-GA,The results show that this model has a better effect on short-term inbound flow prediction.Its Root Mean Squard Error is 9.8% lower than the original light GBM model,and Compared with SARIMA,LSTM and ANN-GA models,the reduction was 19.7%,10.4% and 14.6%,respectively.Using C/S Framework,C# language on the Visual Studio platform to develop Win Form form reservoir short-term inflow flow prediction and reservoir scheduling system,using Entity Framework for Entity relationship mapping,with the help of Dot Net Bar optimization Win Form UI.The system has many functions,such as meteorological data management,reservoir basic information management,rain-water information management,rainwater forecast,dispatching plan pre-generation,etc.Through the integrated and improved short-term inflow flow prediction model,it is possible to forecast the water situation and rain situation in advance,so as to understand the flood situation in time,make reasonable plans,and improve the working efficiency and recognition of the reservoir units.