Study On Multi-objective Optimization Of Regional Groundwater Pollution Monitoring Well Network

As one of the primary water sources,groundwater has important ecological service function.Groundwater monitoring provides necessary information for the management and maintenance of groundwater ecological security.In recent years,as the construction of the groundwater monitoring well,the information obtained increases and certainly,expense explodes too.How to obtain the most abundant groundwater information at the lowest cost has been one of the hot issues that scientists and managers pay attention to.Based on simulated annealing algorithm,random forest algorithm is used to select the most indicative monitoring index for groundwater quality evolution in this paper.Considering the risk of groundwater pollution,with the objective function to be the lowest cost,the highest precision and the least uncertainty,the multi-objective optimization model of groundwater pollution monitoring well network is established.The rationality of the optimization results is evaluated by the hydrogeological analysis method,voronoi map and geostatistic approach,and the demonstration application is carried out in the plain area of Pinggu Basin in Beijing.（1）The random forest algorithm is used to select the indicative indicators of groundwater quality monitoring,and the relative importance of 17 water quality indexes is ranked.The results of random forest algorithm evaluation show that NO₃^--N,Mn,Cl^-,Na⁺,As,NH₄⁺-N and total hardness are relatively important monitoring indexes.Combined with the groundwater quality and random forest algorithm screening results,NO₃^--N,Mn,Cl^-,As,Na⁺,NH₄⁺-N were determined as optimization indexes in this paper.（2）Taking groundwater vulnerability as the decision variable,we construct a multi-objective optimization model by simulated annealing algorithm.The results show that the optimal number of monitoring wells in unconfined aquifer in the study area is 29,and the redundant monitoring wells are PG-47、PG-18 and PG-58,including 1 regional monitoring well（PG-18）and two pollution source monitoring wells.At the same time,in view of the shortage of monitoring wells in some areas,the preset scheme of adding 5 wells is put forward.（3）Hydrogeological analysis,Voronoi maps and geostatistic assessment results show that three redundant wells are located in areas with lower pollution risk and five new wells are mainly located in areas with higher pollution risk.The relative error of PG-47 is 44.3%,which exceeds the specified threshold（20%）,and the relative error of PG-18 and PG-58 monitoring wells is within the threshold range.Optimized monitoring well network monitoring range is wider.