The Study Of Mid-long Term Runoff Forecasting Model And The Design & Application Of The System

The Yulong Kashgar River is located in the western frontier region of China.The climate in the basin is dry,rainfall is low,and sunshine hours are long.In recent years,due to the overexploitation and waste of freshwater resources,the river basin ecosystems have been deteriorating,and river floods have caused disasters.At present,the hydropower monitoring module for the hydrology and water control automation system of the Xinjiang Dakarqu hydropower station has low prediction accuracy for mid-and long-term runoff.The redesign of the mid-and long-term runoff forecasting system module is of great significance to the southern Xinjiang where water resources are extremely lacking,and the rational dispatch of water resources to provide information to solve the contradiction between irrigated agricultural water use,ecological water use,and hydropower development.The traditional tele-relational statistical analysis method directly analyzes the correlation between tele-correlation factors and forecast variables(such as runoff,rainfall,etc.).The drawback is that tele-correlation factors are often large-scale meteorological factors,and their decisive role in predicting variables can only be The more obvious rules appear on larger scales,and the relationship with local variables is susceptible to various contingent factors.Based on this,consider the introduction of temperature,evaporation and other data to ensure the credibility of the forecast.The paper is based on a data-driven model forecasting method.The BP model and GA-BP model were studied.The two methods are applied to the runoff forecasting of the Dakokake Hydropower Station on a monthly time scale.The selection of the calculation parameters and the structure of the neural network are analyzed.Then,a genetic algorithm was introduced to improve the global search ability to optimize the weights,GA-BP was established,and the accuracy was evaluated to meet the application goals.The nonlinear and local optimization ability of the wavelet basis function itself is stronger than the S-excitation function,and the BP algorithm has the characteristics of nonlinear mapping,self-organization,self-organization,self-learning ability,strong robustness and fault-tolerance,etc.Construct a wavelet neural network to increase the rigor and convergence speed of the architecture.On the basis of GA-BP,the GA-WNN model was built using intelligent combination of models,taking into account the advantages of the three models,and achieving good results from the prediction accuracy.Comparing these three kinds of prediction algorithms,using the analysis of the identification standard,the GA-WNN prediction model has the lowest average relative error rate and is of great value.The paper analyzes the application of various models in the water quality control and monitoring module of the Dacokq hydropower station,realizes the application of some models in the water control system,and sends the data to the National Grid Power Dispatching Control Center.As an important technical support for the dispatch system.At the same time,it implements a host system for reinforcing the servers in the network security protection system of the power monitoring system to ensure that the irrigation water supply,flood prevention and disaster reduction,and hydroelectric power generation tasks provide reliable decision-making basis.