Integrating an Agent-Based Model with a Multi-Objective Evolutionary Algorithm to Warn Consumers in a Water Contamination Event Using Emergency Vehicles

Abstract

Listen

Translate article

Many protective response actions are available to protect consumers affected by a water distribution contamination event. Routing a fleet of emergency vehicles to disseminate water warnings is considered in this study. Identifying optimal trips to effectively notice consumers is difficult as a water contamination event is a complex and dynamic system and vehicle routing problems are categorized as NP-hard domain. A novel framework is developed by coupling an agent-based model with an optimization algorithm to search for optimality in a warning dissemination. Solutions representing an optimal front are explored to demonstrate the relationship between the number of exposed consumers and the maximum travel time as conflicting objectives. NSGA-II is used with a tree-based representation. The developed framework is applied to a virtual case study, Mesopolis. INTRODUCTION Public health is threaten by toxins that are accidentally and intentionally introduced to a water network. Public trust in utility mangers can be eroded if actions to protect consumers are taken inappropriately. To avoid the loss of social trust in utility managers, decisions about protecting consumers should be made accordingly to mitigate the consequence of these events. The decision making process gives priority to a warning message dissemination process to inform consumers about the hazard associated with activities related to using tap water, such as drinking water. Utility managers have different choices to deliver a warning: broadcast a warning via media, route a fleet of emergency vehicles to use a siren as a warning sign, and deliver a leaflet at houses to warn and give instructions for short-term alternative water supply. To protect consumers effectively, the EPA (USEPA 2003) suggests locating the contaminant in the network before conducting response actions. When routing emergency vehicles, projecting a trip for each emergency vehicle is difficult because the characterization of a contaminant changes dynamically as more consumers become notified about the event. Notified consumers suspend the use of tap water for waterrelated activities that are perceived as a threat to their health. Also, utility managers may use containment strategies to remove the contaminant. The containment strategies directly change the hydraulics of a water network. Therefore, water distribution contamination events are defined as dynamic and complex systems. Using models with static demands to find the concentration of contamination at nodes may neglect important interactions. A sociotechnical model was proposed based on agent-based modeling (ABM) (Holland 1995) to evaluate the public health consequences of a water contamination event (Zechman 2011). The ABM approach models elements (agents) of a system and simulates interactions among the agents using a set of predefined rules to predict the emergent behavior of the system. The ABM model was extended 2550 World Environmental and Water Resources Congress 2013: Showcasing the Future © ASCE 2013