Sensitivity to crossover and mutation probabilities in Levee's setback optimization

• Authors: M. Shafiei;O. Bozorg Haddad
• Venue: EC'05 Proceedings of the 6th WSEAS international conference on Evolutionary computing
• Year: 2005

Floods are one of the major natural disasters that often threaten human lives and cause significant economic losses around the world. The history of mankind is filled with the stories of our struggles with floods to protect human races and to sustain the progression of our civilizations. Flood defense systems are designed and constructed to protect low-lying areas against flooding. Engineering design often is at the final stage for finding technical means to best accomplish the project goals. Over the years, risk-based design and optimization methods have proven to be useful tools to obtain economic design of protection systems. The most common economic framework for floodplain management is minimization of expected annual damages and flood management expenses, structural and nonstructural flood control options. Levee systems have been built for flood protection in numerous rivers, lakes and coasts in the world over the long human history. Economic design of a levee system for flood protection involves balancing costs of levee building (height), the losses of land value sacrificed for floodway expansion (setback) and flood damages from inadequate channel capacity. The application of GAs to water resources problems has been increased in recent years. The study of genetic algorithms (GAs) has developed into a powerful optimization approach. GAs have so far had very little applications in flood defense systems optimization. In this paper some new different approaches to GA formulation are considered, along with a range of sensitivity analysis. The object has been to present GAs as a practical tool in levee design optimization and to examine the potential of different GA formulations for solving the problem. The results obtained indicate that there is potential for application of GAs to levees optimization problems, where the objective function is nonlinear and other optimization techniques may be difficult to apply.