Rainfall Characteristics Analysis And Rainfall Prediction In The Rainy Season Of Shiyang River Basin

Shiyang River Basin is located in the arid area of Northwest China,which is greatly affected by global warming,with little rainfall and large evaporation.The river is mainly supplied by rainfall in the atmosphere,and the water resources are relatively scarce,which is prone to drought.There have been several floods in history.The population of this area is concentrated in the oasis basin,the demand for water resources is great,and the problems of water environment pollution and water resource distribution inequality are particularly serious.The rainy season is four months from June to September.It is very important to analyze the characteristics of rainy season rainfall and forecast the rainfall for the rational development,utilization and allocation of water resources in this area.Many scholars at home and abroad have done a lot of research on the change of rainfall characteristics and the prediction of rainfall.However,the research on rainfall characteristics of Shiyang River Basin in China mainly focuses on the analysis of rainfall trend,and there is no accurate rainfall prediction model and method.Therefore,in order to obtain a more comprehensive rainfall characteristics of Shiyang River Basin and accurately predict rainfall.This paper has done the following work:(1)Using linear regression,R / S analysis,wavelet analysis,MK mutation,sliding ttest and other methods,the trend,mutation and period of rainfall in different grades in rainy season and between months and years at 10 meteorological stations were analyzed.Among the five selected meteorological factors(average relative humidity,evaporation,average sunshine duration,average air pressure and average temperature),the one with the strongest correlation with rainfall was the average relative humidity,and the other four meteorological factors are of medium or weak correlation.(2)A BP neural network model was established to predict the total rainfall in rainy season with meteorological factors as prediction factors.The feasibility of the model was verified.The average relative error between the predicted value and the real value ranges from 8% to 19%,and there was a large optimization space.In this paper,genetic algorithm was used to optimize BP neural network to find the optimal initial connection weights and thresholds.The average relative error range between the predicted value and the real value obtained by GA-BP model is reduced to 6%-15%.The change trend of the predicted value of each station was basically the same as the real rainfall,which was more consistent with the change trend obtained by the original model.Therefore,the prediction accuracy of GABP model was higher and it could better describe the original Trends of rainfall data.In the future,other algorithms could be considered to continue optimization.