A genetic algorithm approach for water system optimization

The progress made in past years in genetic algorithms led to a general interest in the possibility of applying these techniques to real Water Systems (WS). As is well known, this kind of problem typically generates computationally expensive models involving a large number of alternatives. In order to reach a more adequate level of adherence to the physical system some more detailed models are resolved taking into account non-linearity in objective function and constraints. An adequate approach for the design and maintenance optimization of pipe networks for water supply distribution would consider the non-linear relation between head-loss in each pipe, its diameter, length and hydraulic property. A large literature exists on optimization procedures of pipe networks. The traditional method is by trial and error guided by single iteration results and by user experiences. Linear programming (LP) approximation or procedures that use LP in partition model have also been developed for this problem. Standard non-linear optimization procedures (such as generalized reduced gradient technique) frequently identify only local optima for this kind of problem. Recent works have been on application of genetic algorithms even for complex water distribution networks including new design, rehabilitation and maintenance of water systems. Efficient resolution techniques exploiting the peculiarity of the matrix constraints structure in the global optimization model can be adopted in the resolution of non linear system during a genetic iteration. Applications, results and perspectives are presented too. From first applications, the problem seems to be well modeled using genetic algorithms and results show that this technique can be used for a wide variety of water resources optimization problems. Some applications, results and perspectives are presented.