Resilience, which defines the ability of a system to maintain continuous customer supply, is a key driver in the management of water distribution networks (WDN). The research programme investigates the optimal design-for-control (DfC) problem which consists in simultaneously selecting new valves and pipes for installation, and optimizing control valve settings in existing WDNs, to minimize pressure induced pipe stress and leakage, and maximize resilience. The problem is illustrated with practical DfC applications on case study networks and a large-scale operational system.

First, the thesis focuses on the single-objective DfC problem for the minimization of pressure induced pipe stress and leakage, or the maximization of resilience, represented, respectively, by the average zone pressure (AZP) and the resilience index (Ir), which are both increasing functions of the pressure at network nodes. A tailored spatial branch-and-bound (sBB) algorithm is developed to compute a feasible solution of the resulting non-convex mixed-integer non-linear program (MINLP), with global optimality bounds. Iterative bound tightening is implemented, to accelerate the convergence of the algorithm, and an equivalent reformulation of Ir is introduced, to facilitate the computation of global bounds.

Next, we consider jointly the conflicting objectives of AZP minimization and Ir maximization. Using the method of epsilon-constraints, the resulting non-convex bi-objective mixed-integer non-linear program (BOMINLP) is reformulated as a sequence of non-convex MINLPs, which are solved using sBB. The approach returns a set of feasible trade-off solutions between AZP and Ir, along with global non-dominance bounds, in the form of an outer approximation of the Pareto front.

Finally, we investigate the dynamic aggregation of DMAs in a large-scale, sectorized network. Due to the computational complexity of the DfC problem, the previous global methods scale badly to large network instances, and a heuristic method is developed, based on the solution of the problem on a reduced network model. The results of the numerical experiment show that dynamic DMA aggregation improves network resilience while minimizing pressure induced pipe stress and leakage.