Search Space Analysis for the Combined Mathematical Model (Linear and Nonlinear) of the Water Distribution Network Design Problem

摘要

This paper presents an experimental study of the solutions space generated by the mathematical model of the Water Distribution Network Design Problem by using Two-Looped network benchmarks to find the feasible solutions space. It shows how the performance of a typical Evolutionary Algorithm (EA) can be improved by considering the importance of working with a feasible population and carrying out repetitive mutations and crossovers to generate new feasible offspring with better fitness. The replacement of parents represents the mortality index of a population at each generation of EA. Aiming to compensate the mortality index, EA is forced to maintain a constant population size by increasing the number of descendants with the crossover operator. The experimental results show both the feasible solutions space and the results of the algorithm when using feasible solutions and varying population size.