Abstract
Water demand increases in urban zones, and water scarcity is associated with maintenance problems, such as leakage and pipe ageing, which inevitably lead to complex water distribution systems (WDS) management. In this scenario, intermittent operation emerges as an alternative for system operation. This is not the most desirable solution from a social point of view since many consumers cannot be supplied many times for days. Several works have been developed to help decision-makers improve water system efficiency during the last decades. Nevertheless, few works have considered the combination of several structural interventions, such as pipe replacement, new pump station installation, fixing leaks, or installing and controlling pump stations and valves. This happens because the several alternatives for recovering the hydraulic capacity in decision-making processes are computationally burdensome, mathematically complex, and, sometimes, even physically incompatible. Considering the problem stated by the Battle of Intermittent Water Supply, this work proposes a methodology for optimal operation and recovery of a WDS. The Battle problem is presented in two stages: year zero and the following five years. For year zero, only operational optimisation is allowed. Consequently, optimal operation of pumps and valves is proposed for this initial year to maximise the number of nodes being supplied. Since many leaks and negative pressures appear on year zero, slowing the hydraulic simulation, a skeletonisation process is addressed before applying any optimisation algorithm for finding the optimal operation of the system. For the rest of the years, since implementing structural changes is allowed, the proposal suggests applying a search space reduction process based on the cost-benefit and the hydraulic relevance of each structural alternative evaluated individually in terms of the nine indicators proposed in the Battle statement. Those alternatives that better improve the indicators are then considered in a multi-objective optimisation setting. For every year, a set of structural changes is selected followed by related changes in the operational setup. The alternatives are selected year by year and evaluated considering the past selected alternatives to consider the effects during the five evaluation years. This is done in a dynamic programming process, ensuring to achieve the best solution by the end of the last five years.