Design Optimization Of Liquamatic Groundwater Circulation Well Based On Machine Learning

Groundwater circulation well(GCW)is a new in-situ remediation technology which is low-cost,convenient with small disturbance on the site.GCW has been widely used in groundwater remediation area.However,the research of GCW technology is still in initial stage in China,there is no mature application case of site.With the rapid development of global economy,groundwater pollution is becoming more and more serious.The application of environmental remediation technology is particularly important,in order to further promote the practical application of GCW technology,optimize the technical parameters of GCW,a series of researches are carried out,it is significant to the application of GCW in our country.this paper proposes an optimization method of GCW based on machine learning algorithm.Firstly,the groundwater numerical simulation package Flo Py is used to build the groundwater flow model and particle tracking model,the operation effect of GCW under typical conditions is simulated to obtain the data sets.Determine anisotropy of permeability(K_H/K_V),hydraulic gradient(I),vertical hydraulic conductivity(Kv),aquifer thickness(M),the length of middle tube(L),the total length of the screen sections(a)and pumping rate(Q),as input variables,determine longitudinal influence radius(R_L)particle recovery rate(P_r)and transverse influence radius(R_T)as objective functions.Models for predicting the operation effect was obtained by three machine learning algorithms:multiple linear regression(MLR);support vector machine(SVM);artificial neural networks(ANN).Based on the prediction models,particle swarm optimization algorithms and non-dominated sorting genetic algorithms are used to carry out single objective and multi-objective optimization calculation,realize effective optimization of GCW structure and operation parameters,provide a new method for GCW optimization design,the main research results are as follows:(1)Database of GCW operation is developed.The numerical model of GCW is set up to simulate the operation state of GCW,a database,which can fully reflect the actual operation state of GCW,is constructed.(2)Prediction model of GCW operation effect is developed.The machine learning algorithm(MLP,SVM,ANN)is used to train the models,obtain models that can predict operation effect of GCW.(3)Single object optimization of GCW is carried out for a test site in Shaanxi.Substitute hydrogeological parameters of test site into the prediction models,then PSO algorithm is used to optimize R_L and P_r respectively,the single objective optimization design schemes is obtained.(4)Multi-objective optimization of GCW is carried out for a test site in Shaanxi.Substitute hydrogeological parameters of test site into the prediction models,employ NSGA-Ⅱand TOPSIS,the multi-objective optimization scheme is carried out to improve the comprehensive operation effect of GCW.