Simplifying multiobjective optimization using genetic algorithms

Many water resources problems require careful balancing of fiscal, technical, and social objectives. Informed negotiation and balancing of objectives can be greatly aided through the use of evolutionary multiobjective optimization (EMO) algorithms, which can evolve entire tradeoff (or Pareto) surfaces within a single run. The primary difficulty in using these methods lies in the large number of parameters that must be specified to ensure that these algorithms effectively quantify design tradeoffs. This paper addresses this difficulty by introducing a multi-population design methodology that automates parameter specification for the Nondominated Sorted Genetic Algorithm-II (NSGA-II). The NSGA-II design methodology is successfully demonstrated on a multiobjective long-term groundwater monitoring application. The design methodology fully exploits the efficiency of the NSGA-II to enable the solution of a new class of high order multiobjective applications in which users can balance more than two performance objectives. Using this methodology, multiobjective optimization problems can now be solved automatically with only a few simple user inputs.