Research On The Groundwater Dynamic Method

The water level dynamics are the performance of the groundwater system stimulated by various factors from the external environment and the response of the groundwater system to maintaining its own balance under the influence of both natural and human factors..The paper analyzes the dynamic change law of shallow groundwater level and eatablishes related models by using GM(1,1),ARMA,BP neural network model combined with MATLB7.0,SPSS and EVIEWS7.2 software on the premise of mastering many years of dynamic data of the Mai Ling water source such as precipitation,exploitation and groundwater level.(1)According to the collected data of water level,precipitation and exploitation in 2000~2015,this paper analyzes the dynamic change of shallow groundwater level in the Mai Ling water source and the main factors such as exploitation and precipitation that affect the dynamic change of groundwater level are determined.The change of groundwater level during the year and the year of the groundwater level is analyzed,and the change of the groundwater level in the wet,normal and dry years is analyzed.The change of the groundwater level is mainly related to the precipitation.The groundwater level is the highest in the wet year,and the groundwater level is relatively stable.The annual change of groundwater level in shallow water source is mainly due to the increase of artificial exploitation and the decrease of precipitation,so that the overall groundwater level is declining.(2)Respectively make the water source xiang3?xiang14?xiang20 Well's underground water level,set up GM(1,1),ARMA(2,1)and BP neural network prediction models.The average relative error of GM(1,1)model is 0.74%,0.66% and 0.51% respectively,and the prediction effect is normal.The average relative error of ARMA(2,1)model is 0.49%,0.44% and 0.39% respectively,the prediction effect is better.The BP neural network has the factors that affect the exploitation and precipitation of groundwater level,the average relative error of prediction is 1.02%,1.37% and 1.14% respectively,and the prediction accuracy is general,which is the lowest in the three models.The reason is that BP neural network is slow to learn.easy to fall into the local minimum,can not get the global optimal solution and other shortcomings.(3)The advantages of each model are presented through the combination of three kinds of single model,The establishment of GM-ARMA-BP neural network,makes up for the lack of a single model.The results show that the average relative error of the model is 0.19%,0.29% and 0.24% respectively.The prediction accuracy is higher than the other three single models.If there is only a sequence of groundwater level samples,the ARMA model used in this paper can be used to forecast the groundwater level of the future year.If there are multiple sets of sample data such as precipitation,mining volume and groundwater level,The GM-ARMA-BP neural network model can be used to predict the groundwater level of the future year.In this paper,four different models are established by using the Mai Ling source of shallow groundwater,which can predict the future trend of groundwater,evaluate the groundwater resources,and also provide scientific basis for groundwater resources management and decision-making of related department.