Applying Multi-objective Niching Co-evolutionary Algorithm to Generate Insight for Water Resources Management Problems

Real-world engineering problems typically involve multiple objectives that should be addressed simultaneously. A set of Pareto-optimal solutions that represents the trade-off among conflicting objectives can be identified to provide knowledge about the performance of alternative solutions for design and management problems. Many multi-objective evolutionary algorithms (MOEA) have been developed and designed to efficiently identify a set of nondominated solutions, and these algorithms have been successfully applied for realistic engineering problems. Realistic design problems may require more analysis and solution generation capabilities than provided by a typical MOEA. The fitness landscapes for realistic design problems are often nonlinear, complex, and multi-modal, and, in addition, water resources planning and management problems typically involve a diverse set of stakeholders with a set of preferences that are not represented mathematically and included in an optimization model. Identification of alternative sets of nondominated solutions that are similarly Pareto-optimal can address the problem of multi-modality in the decision space and provide additional insight to problem solution and options for implementation. The Multi-objective Niching Co-evolutionary Algorithm (MNCA) was designed to use a multi-population optimization search to evolve multiple nondominated solution sets. MNCA is demonstrated here for solution of two illustrative water resources management problems, including a water supply network design problem and a water quality management problem. Results are analyzed to demonstrate the use of MNCA to generate new insight and options for addressing difficult problems.