ANN-GA-Based Model for Multiple Objective Management of Coastal Aquifers

Abstract

A linked simulation-optimization model using artificial neural networks (ANNs) and genetic algorithms (GAs) is developed for deriving multiple objective management strategies for coastal aquifers. The GA-based optimization approach is especially suitable for externally linking a numerical simulation model within the optimization model. However, the solution of a linked simulation-optimization model is computationally intensive, as a very large number of iterations between the optimization and the simulation models are necessary to arrive at an optimal management strategy. Computational efficiency and feasibility for such linked models can be enhanced by simplifying the simulation process by an approximation. A possible approach for such approximation is the use of an ANN model. In this paper, an ANN model is developed initially as an approximate simulator of the three-dimensional density dependent flow and transport processes in a coastal aquifer. A simulation-optimization model is then developed by linking the ANN model with a GA-based optimization model for solving multiple objective saltwater management problems. The performance of the optimization models is evaluated using an illustrative study area. For comparison of the solution results, a multiple objective management model is also solved using embedded formulation and classical nonlinear optimization technique. The comparison of results shows potential feasibility of the proposed methodology in solving multiple objective management model for coastal aquifers.