Multiobjective Approach for Water Distribution Network Design Combining Pipe Sizing and Isolation Valve Placement

Abstract

This paper presents a novel methodology for the coupled optimization of water distribution network (WDN) pipe sizing and isolation valve placement. The methodology is based on the effective application of biobjective genetic algorithm optimization to simultaneously minimize installation cost and average demand shortfall due to segment isolations, followed by a postprocessing financial analysis including assessment of valve maintenance costs. For speeding up convergence, a constraint is implemented in the optimization to enforce the telescopic property of pipe diameters, which are expected to shrink from source(s) to external parts of the WDN. The methodology is applied to two case studies with different levels of complexity, enabling the results to be compared with the traditional approach based on decoupled least-cost optimization of pipe sizing and placement of isolation valves at all (N_valve rule) or all but one (N-1_valve rule) pipes connected to the generic demand node. The least-cost pipe design equipped with the N_valve rule valve placement is always dominated. The adoption of the N-1_valve rule yields very close results to the optimization, as long as the best among the very numerous possible N-1_valve layouts is suitably identified. However, more beneficial solutions featuring lower valve maintenance costs can be selected from the optimal solutions obtained from the novel methodology. The use of the weighted average demand shortfall as objective function, instead of the simple average demand shortfall, tends to produce a larger variety of solutions.