Research On Groundwater Level Prediction Based On Wavelet Transform And GRU Deep Neural Network

The prediction of groundwater level is an important subject of hydrology research.It plays an important role in the management and utilization of water resources,and is the premise and guarantee of the control of the groundwater level.Accurate groundwater level prediction helps to maintain reliable water supply and maintain ecological balance in life,agriculture and industry.In order to manage groundwater resources effectively,it is very important to record and predict groundwater level accurately.Using groundwater level prediction results to provide basis for decision-making departments,we can make use of groundwater in a sustainable and effective way.In this paper,the wavelet transform and GRU deep neural network are used to predict the change of groundwater level.The main contents and achievements are as follows:(1)A grasshopper optimization algorithm(GOA)based on group intelligence is studied.The original grasshopper optimization algorithm was improved from three aspects,and an improved grasshopper optimization algorithm(IGOA)was proposed.Subsequently,the IGOA is used to optimize the parameters of the GRU model,and the IGOA-GRU prediction model is established.(2)Using the wavelet transform to deal with the groundwater level data and reconstruct each component after wavelet decomposition,the WT-IGOA-GRU groundwater level prediction model is established,and the WT-IGOA-GRU model is first used to predict the groundwater level in the field of groundwater level prediction.(3)The prediction results of the WT-IGOA-GRU model are compared with the prediction results of the GRU model and the IGOA-GRU model when the WT-IGOA-GRU model is used to predict the real groundwater level.The prediction accuracy is evaluated by four indicators,such as root mean square error.The results show that the IGOA algorithm can improve the performance of GRU model in predicting groundwater level,and combined with wavelet transform can further improve the prediction accuracy.Subsequently,the groundwater level was forecasted using BP and RBF models optimized by IGOA.By comparing the prediction results of different network models,it is found that the predictive performance of the WT-IGOA-GRU model is better than the other two neural network models,and the conclusion can be used as a reference in the future research of the prediction of groundwater level.