A novel two-step approach for optimal groundwater remediation by coupling extreme learning machine with evolutionary hunting strategy based metaheuristics

Abstract

We propose a simulation-optimization (SO) model based on a novel two-step strategy for the optimal design of groundwater remediation systems. The SO models are developed by coupling simulation models directly or through the extreme learning machine (ELM) with evolutionary hunting strategy based metaheuristics (EHSMs). In the first step, EHSMs with a combinatorial optimization technique are used to obtain optimal pumping locations by minimizing the percentage of contaminant mass that remained in the aquifer while keeping the pumping strategy as constant. In the second step, the optimal pumping locations are directly used as input, and a composite function is employed to minimize the sum of the water extraction rates and the percentage of extracted contaminant mass by constraining hydraulic heads and contaminant concentrations. The performance of the two-step strategy is found to be slightly better and computationally more efficient than the alternate approach. Moreover, various statistical measures suggest the superiority of EHSMs over other metaheuristics for groundwater remediation.

Keywords: Extreme learning machine (ELM); Grey wolf optimization (GWO); Groundwater remediation; Harris hawks optimization (HHO); Whale optimization (WO).