Multiscale Strategies for Solving Water Resources Management Problems with Genetic Algorithms

Genetic Algorithms (GAs) face major computational bottlenecks when numerical models are used for estimating the fitness of the objective function. Especially in large-scale water resources design problems, where the scale of the spatial grids is important in determining the numerical accuracy of the design, a tradeoff exists in the precision of the fitness function and the computational expenses endured. This paper discusses multiscale strategies that can be utilized for improving the performance of GAs when working with spatial grid dependant fitness functions. The strategy uses fine grid and coarse grid fitness functions strategically to maintain the accuracy and computational speed of the problem and drive the GA towards better and more accurate solutions faster. The algorithm's efficacy is tested using a groundwater remediation design case study.