Probabilistic Analysis and Optimization to Characterize Critical Water Distribution System Contamination Scenarios

• Published in: Journal of water resources planning and management Vol. 139; no. 2; pp. 191 - 199
• Main Authors: Rasekh, Amin, Brumbelow, Kelly
• Format: Journal Article
• Language: English
• Published: American Society of Civil Engineers 01-03-2013
• Subjects: Computer simulation; Contamination; Monte Carlo methods; Multiple objective analysis; Optimization; Risk; Systems analysis; Technical Papers; Water distribution
• ISSN: 0733-9496; 1943-5452
• DOI: 10.1061/(ASCE)WR.1943-5452.0000242

AbstractCharacterization of critical water distribution system (WDS) contamination scenarios—defined by a set of attributes, a probability of occurrence, and a specific level of consequences—is a prerequisite for preparation of reliable and cost-effective mitigation, preparedness, and emergency response plans. This study develops Monte Carlo and risk-based optimization schemes to evaluate contamination risk of WDSs for generation of this important class of scenarios, which are representative of the most vulnerable aspects of the system. Defining attributes of contamination scenarios are identified as contaminant type and amount, contamination location, start time, duration, and time of year scenario occurs. Well-documented waterborne outbreaks reported in developed nations are analyzed to empirically estimate statistical characteristics of defining attributes in accidental events. Monte Carlo simulation is conducted to determine the probability distribution of public-health consequences, aggregate conditional risk, and significance of different scenario attributes. A multiobjective optimization methodology is proposed to capture the attributes of critical accidental contamination scenarios. The principal risk components of likelihood and health consequences are treated as optimization objectives and are maximized simultaneously to identify an ensemble of nondominated critical scenarios. The multiobjective approach provides insight into system risk and potential mitigation options not available under maximum-risk or maximum-consequences analyses. Performance and applicability of developed models is demonstrated on the WDS of a virtual midsize city that possesses characteristics of complex real-world distribution networks.