A probablisitic multi-objective management model for groundwater remediation under uncertainty

The optimal design of groundwater remediation systems are often subject to uncertain hydrogeological parameters and multiple uncertain objectives, involving minimization of remediation cost, and minimization of contaminant mass remaining in the aquifer. To design a robust and reliable groundwater remediation system, the stochastic simulations (Monte Carlo simulation) with multiple realizations of uncertain parameters, which are generated by Sequential Gaussian Simulation (SGSIM), are applied to tackle the uncertainty analysis of an synthetic remediation site. In the present study, we propose a probabilistic multi-objective evolutionary algorithm, named probabilistic improved niched Pareto genetic algorithm (PINPGA). PINPGA was improved by using stochastic simulation for objection function evaluations and incorporating probabilistic Pareto ranking and niche technique into INPGA for multi-objective selection operator. The proposed algorithm is then applied to the synthetic groundwater remediation test case. The performances of the methodology generating the reliability of the Pareto-optimal solution are assessed and compared using Monte Carlo analysis. The optimization results indicate that using such an uncertainty-based multi-objective optimization scheme can give reliable solution to groundwater remediation design, giving decision makers a practical and robust optimization tool.