Dynamic scenario selection in optimal design problems and evolutionary optimization with uncertain system knowledge

The design of water resource management and control systems have provided a promising space for evolutionary algorithms. In many cases a system for managing a water resource requires a large degree of planning and design before implementation and many stake holders perceive different objectives with different importance. Multiobjective evolutionary algorithms inherently provide a tool that can best satisfy the desires of many stakeholders (many objectives) through computation of a non-dominated solution set. However, the performance of an optimal solution provided by a multiobjective evolutionary algorithm is likely to deteriorate during real-world implementation if design conditions of the optimization framework are not identical to those imposed on the system in practice. This paper focuses on evaluating a scenario based multiobjective evolutionary algorithm for real-world design problems in which the environment where a system will operate is dynamic, and uncertain. A previously developed genetic algorithm termed the "RNSGA-II" used for water distribution system design is augmented to incorporate robust objectives and simple Monte Carlo sampling to solve the classic water quality sensor placement problem. This study aims to further develop an understanding of scenario based optimization methods for optimizing solutions to perform well in the face of uncertainty.