Parallel evolutionary multiobjective optimization of water distribution system design

A master-slave parallel version of epsilon-NSGAII, in conjunction with the use of a supercomputer, is used to solve the Battle of Water Networks II (BWN-II) multiobjective design problem. In order to minimise the computation time, the optimisation problem is decomposed into multiple simplified problems. These problems have a smaller number of decision variables and have different operation analyses performed, i.e. with or without the slow power failure operation analysis. The Paretooptimal solutions of these simplified problems are used as pre-conditioners to the optimisation of the full problem formulation that is much more computationally intensive. A set of Pareto-optimal solutions were obtained from the full search of the problem and the tradeoff relationships were analysed and used to select the final solution for the BWN-II competition. The final solution selected has a good performance in the water age and greenhouse gas emissions objectives; however, as a tradeoff, it has a relatively higher cost.