A new NLP-DE hybrid optimisation model for water distribution system optimisation

Abstract

This paper proposes a novel NLP-DE hybrid optimisation method for the water distribution system (WDS) optimisation. In the proposed hybrid optimisation approach, graph theory is first used to find a shortest-distance tree for a looped WDS. Then non-linear programming (NLP) is employed to optimize the shortest-distance tree in order to obtain an approximately optimal solution for the original whole water network. Finally, a seeding table is created based on the NLP optimal solution and a differential evolution (DE) is seeded by the seeding table to optimize the original whole network. Two WDS case studies with the number of decision variables ranging from 30 to 314 are used to verify the effectiveness of the proposed hybrid optimisation method. A standard differential evolution (SDE) is also applied to the two WDS case studies to enable the performance comparison with the proposed approach. The results obtained show that the proposed hybrid method is able to find better quality optimal solutions with significantly improved efficiency than the SDE.