Application of Proxy Simulators in Groundwater Simulation Optimization: ANN and SVM proxy simulators for optimal pumping policy assessment

Simulation optimization techniques are often used to address many problems of groundwater management. In these techniques an aquifer simulation model is coupled with an optimization algorithm to get the optimal solution. Simulation model normally engages a finite difference, finite element or a mesh free technique. ANN and SVM based proxy simulators are used in this study to replace the FEM simulation model. When the objective function in the study is implicit, its gradient cannot be determined with respect to the decision variables which restrain the use of gradient based optimization methods. In this study differential evolution optimization model is used to minimize the collective drawdown for obtaining an optimal pumping policy for the aquifer system. The ANN and SVM based simulation optimization models result in noteworthy saving of computational time.