Premium
Using genetic algorithms to solve a multiple objective groundwater pollution containment problem
Author(s) -
Ritzel Brian J.,
Eheart J. Wayland,
Ranjithan S.
Publication year - 1994
Publication title -
water resources research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.863
H-Index - 217
eISSN - 1944-7973
pISSN - 0043-1397
DOI - 10.1029/93wr03511
Subject(s) - mathematical optimization , genetic algorithm , crossover , pareto principle , population , selection (genetic algorithm) , multi objective optimization , containment (computer programming) , set (abstract data type) , computer science , mathematics , programming language , demography , artificial intelligence , sociology
The genetic algorithm (GA), a new search technique, is applied to a multiple objective groundwater pollution containment problem. This problem involves finding the set of optimal solutions on the trade‐off curve between the reliability and cost of a hydraulic containment system. The decision variables are how many wells to install, where to install them, and how much to pump from each. The GA is an optimization technique patterned after the biological processes of natural selection and evolution. A GA operates on a population of decision variable sets. Through the application of three specialized genetic operators: selection, crossover, and mutation, a GA population “evolves” toward an optimal solution. In the paper, simple GAs and GAs that can solve multiple objective problems are described. Two variations of a multiple objective GA are formulated: a vector‐evaluated GA (VEGA) and a Pareto GA. For the zero‐fixed cost situation, the Pareto GA is shown to be superior to the VEGA and is shown to produce a trade‐off curve similar to that obtained via another optimization technique, mixed integer chance constrained programming (MICCP). The effect on the VEGA and Pareto GA of parameter variation is shown. The Pareto GA is shown to be capable of incorporating the fixed costs associated with installing a system of wells. Results for several levels of fixed cost are presented. A comparison of computer resources required by the GAs and the MICCP method is given. Future research plans are discussed, including the incorporation of the objective of pump‐out time into the model and the development of parallelized GAs.