Optimal Rehabilitation of Water Distribution Systems using a Cluster-based Technique

This is the author accepted manuscript. The final version is available from [American Society of Civil Engineers via the DOI in this record.Optimal rehabilitation of large water distribution system (WDS) with many decision variables, is often time-consuming and computationally expensive. This paper presents a new optimal rehabilitation methodology for WDSs based on graph theory clustering concept. The methodology starts with partitioning the WDS based on its connectivity properties into a number of clusters (small sub-systems). Pipes which might have direct impact on system performance are identified and considered for rehabilitation problem. Three optimisation-based strategies are then considered for pipe rehabilitation in the clustered network: 1) rehabilitation of some of the pipes inside the clusters; 2) rehabilitation of pipes in the path supplying water to the clusters; 3) combination of strategies 1 and 2. In all optimisation strategies, the decision variables for rehabilitation problem are the diameters of duplicated pipes; the objective functions are to minimise the total cost of duplicated pipes and to minimise the number of nodes with pressure deficiency. The performance of proposed strategies was demonstrated in a large WDS with pressure deficiencies. The performance of these strategies were also compared to the full search space optimisation strategy and engineering judgement based optimisation strategy in which all pipes and selection of pipes are considered as decision variables respectively. The results show that strategy 3 is able to generate solutions with similar performance that are cheaper by around 53% and 35% in comparison with the full search space and engineering judgement based optimisation strategies respectively. The results also demonstrate that the cluster-based approach can reduce the computational efforts for achieving optimum solutions compared to the other optimization strategies.The work reported is supported by the UK Engineering & Physical Sciences Research Council (EPSRC) project Safe & SuRe (EP/K006924/1)