Post-Disaster Restoration of Drinking Water Distribution Networks
A multi-objective optimization approach for isolation valve placement and restoration task scheduling
More Info
expand_more
Abstract
Drinking Water Distribution Network (DWDN) vulnerability to disasters yields ever threatening consequences to society. Earthquakes are shown to have devastating effects on these networks, causing multiple pipe breaks and leaks to occur simultaneously. This results in a lack of pressure and unavailability of water for extended time periods. Increasing the resilience of a DWDN by efficient restoration is therefore desired. To restore broken pipes, isolation is required, where isolation valves are installed in the network to allow for this. Closing the valves and replacing the pipes are physical tasks to be executed by workforces. Moreover, as not every pipe has isolation valves on both sides, isolating a single pipe often results in the isolation of a large segment of the network. The locations of these valves and the scheduling of the restoration tasks are therefore crucial for the restoration process. This thesis aims to enhance DWDN resilience by considering the combined optimization of Restoration Task Scheduling (RTS) and Isolation Valve Placement (IVP). Four hydraulic objectives are proposed to quantify resilience. The objective values corresponding to restoration schedules are obtained by simulation of the restoration process using Pressure-Driven Modeling (PDM). From three candidate methods, the priority rule-based scheduling heuristic is identified to be most suited for combined optimization. A novel Genetic Algorithm (GA) application is proposed to optimize IVP while including the priority rule-based heuristic for RTS. The genetic information of the individual used in the GA represents a valve layout. The valve layout is applied to a DWDN model, which in turn is subdued to five disaster scenarios. The heuristic provides restoration schedules for all scenarios, where the resulting objective values are averaged to compute the fitness of the individual. The method is shown to be able to generate valve layouts for any number of valves. Comparing these layouts to known strategic IVP methods leads to significant improvement in resilience, or can alternatively be used to decrease the number of valves while maintaining performance. Moreover, the method allows for analysis of critical valves coupled to the network properties, pipe failure probability, and pipe diameters. Lastly, by applying single-objective optimization, it elucidates the trade-offs between the objectives.