Multiobjective ensemble surrogate-based optimization algorithm for groundwater optimization designs

Simulation technique is an increasingly focused method for conveniently evaluating a solution or scenario in the field of groundwater. However, traditional evolutionary algorithms require at least thousands of simulation executions when addressing groundwater simulation-based optimization problems to find reasonable solutions. Intensive simulations usually yield a prohibitive computational burden if the simulation involved is time-consuming. To defeat the issue, this paper proposes a multiobjective ensemble surrogate-based optimization algorithm named MESOA for groundwater optimization designs. Surrogate models can reduce the continual usage of expensive-cost models in a way that approximates objective functions. Unlike existing surrogate-assisted evolutionary algorithms, MESOA employs surrogates with various basis functions (RBFs) and Kriging as available surrogates, and presents an adaptive switching technique to construct surrogate models in an online way. In addition, MESOA involves three sample infill criteria and a novel population filter. With the assistance of these techniques, MESOA can fully depict the outline of the true Pareto front, although the times of invoking simu-lation are limited. Some representative benchmark cases are provided to test the applicability and effectiveness of the proposed algorithm at first. Afterward, MESOA is applied to solve some practical groundwater multi -objective optimization designs, such as groundwater remediation and requirement optimization. All empirical results indicate that the proposed algorithm obtains more availability and effectiveness than other algorithms and has wide universality for groundwater optimization designs.