Water usage and treatment network design using genetic algorithms

A mathematical programming model has been developed in this work to identify the cost-optimal and least-consumption water usage and treatment networks (WUTNs). The network configurations of WUTNs were generated on the basis of a superstructure embedded with the existing water-using units, the repeated water-treatment units, and the mixers. The benefits of using this model are clearly demonstrated in the examples. The genetic algorithms were used in this study to solve the WUTN optimization problem. Several techniques have been developed to enhance convergence in the evolution processes. First of all, by adopting the split fractions from splitting nodes as the design variables and devising an encoding strategy accordingly, the search space can be significantly reduced. Second, by cascading a series of evolution processes according to a set of "inducing" parameters, the appropriate ranges of the design variables can be efficiently determined. As a result, the need to obtain a good initial guess can be eliminated. The total number of generations to reach optimum can also be lowered to an acceptable level even for a large WUTN design problem.