Study On Contaminated Site Monitoring Method Based On Unet-LSTM

The landfill leachate has complex components and high pollutant load.After the leakage of the polluted site,it will cause great pollution risk and potential safety hazard to the surrounding soil,groundwater and other environment.Therefore,the pollutant diffusion monitoring of hazardous waste landfill is the key to the environmental safety control of landfill.The resistivity method is based on the resistivity difference between the landfill material and leachate in the polluted site,and inverts the geological structure of the landfill site by observing and analyzing its distribution law.The traditional resistivity method obtains the real resistivity data through the linear inversion algorithm least square inversion.However,the least square inversion requires a high initial model and needs to iterate the partial derivative matrix.Because the resistivity inversion is a nonlinear inversion problem,it is easy to fall into a local minimum and cannot get the global optimal solution.In recent years,nonlinear inversion methods and artificial neural networks have developed rapidly.Artificial neural networks directly construct nonlinear mathematical models from samples,which can map the nonlinear relationship between input and output.The image segmentation model Unet adopts symmetrical structure design,integrates the low-dimensional and high-dimensional features in CNN and uses a small number of data sets for training and testing,and achieves good results.With the migration of pollutants,the apparent resistivity data changes with time.The cyclic neural network LSTM introduces the time series to extract the time representation,which can solve the correlation problem of the apparent resistivity monitoring data,and has a dual memory structure of storing spatial information and time information.Therefore,the integration of u-net and LSTM network to extract the high-dimensional features of pollutant diffusion data from the time and space dimensions can quickly realize the dynamic change of pollution diffusion in the polluted site and provide technical support for subsequent efficient remediation.Based on the above analysis,this paper mainly studies as follows:（1）The chemical transfer module and fluid flow module in COMSOL simulation software are used to simulate the dynamic change of leachate.The geological structure of the study area is from the exploration data of a landfill in Jiangsu Province,and the effects of porosity and seepage intensity on Leachate Transport are studied.With the increase of porosity,the diffusion velocity of pollutants in the direction of groundwater flow decreases in the horizontal direction;In the vertical direction,the vertical diffusion velocity is basically unchanged.With the increase of seepage velocity,the diffusion velocity of pollutants in the direction of groundwater flow increases in the horizontal direction and changes slowly in the vertical direction（2）Integrating u-net and LSTM network,the Unet-LSTM nonlinear inversion network is proposed.By changing the location of injection well,the injection concentration of Na Cl solution,the injection speed of Na Cl solution,porosity and hydraulic conductivity,the concentration of Na Cl solution under different conditions can be obtained,and then converted into resistivity through the concentration resistivity formula to obtain different pollutant concentrations The sample data of different geological structures are used as the input of Unet-LSTM,which is trained to obtain the distribution map of its apparent resistivity,so as to judge the diffusion law of Na Cl solution.（3）In order to study the migration law of pollutants,the Na Cl solution was poured into the experimental well outside the vertical impervious membrane of a landfill in Jiangsu Province,and the Wenner high-density electrical method device was used to collect the apparent resistivity data after the change of geological resistivity after salt water diffusion at different times.The Unet-LSTM neural network was used for inversion to simulate the migration law of pollutants after leakage from the impervious membrane of the landfill,and the actual data was used to verify the effectiveness of Unet-LSTM.