Design And Implementation Of Medium And Long-term Runoff Forecasting System

Hydrological forecasting is an important part of the hydrology science,while medium and long-term runoff forecasting is an important research area for hydrological forecasting.Due to the long forecast period,medium and long-term runoff forecasting has important guiding significance for optimal management and comprehensive dispatching of reservoirs.With the passage of time,the shortcomings of the traditional medium and long-term runoff forecasting methods,such as complicated process,low efficiency,and inconvenient analysis,have become increasingly prominent and can no longer meet production and living needs.This subject comes from the Hanjiang River Water Resources Forecasting and Dispatching System Project of China Institute of Water Resources and Hydropower Research.establish a medium and long-term runoff forecasting system can effectively provide relevant departments with long-term inward water forecasts for years,months,etc.,at the same time,users can relevant water dispatching plans and water resources management plans based on historical water situation and system analysis results so as to be timely and accurate.The measures taken to implement water resources coordination arrangements effectively solved the shortcomings of the traditional methods of forecasting and improved the overall benefits of related departments.This thesis is in accordance with the software engineering specification.Based on the feasibility analysis of the system,analysising the requirements of Hanjiang River medium and long-term runoff forecasting system,including functional requirements and non-functional requirements;Introducing the software architecture design,problem domain design,persistent design,system interface design.And the Support Vector Machine for Regression(SVR)model parameter optimization design;according to the previous needs and design,to achieve the entire medium and long runoff forecast system.The choice of forecast model determines the accuracy and credibility of medium and long-term runoff forecasting,and also determines the practicality of medium and long-term runoff forecasting systems.After studying the status quo of medium and long-term runoff forecast at home and abroad,the thesis finally selects SVR model for medium and long-term runoff forecast,and aims at the three model parameters: the penalty coefficient C,nuclear parameter and insensitivity loss coefficient have a large amount of calculation in the actual assignment process,it is difficult to get the optimal values and other issues,Using Paticle Swarm Optimization(PSO)to optimize the parameters of SVR model.Establishing the PSO-SVR model and achieves automatic selection of parameters.The experimental results show that the PSO-SVR model improves the prediction accuracy compared with the SVR model,and is more stable than the Artificial Neural Network(ANN)model.At present,the system is operating normally,it worked well.It has a good effect on medium and long-term runoff forecasting for major reservoirs and sections in the Hanjiang River basin.It can provide digital decision support for rainfall and runoff forecasting in the Hanjiang River basin and improve the management efficiency of the water resources business in the Hanjiang River basin Economic Belt.It is of great significance.