Integrating Genetic Programming and Agent-based Modeling to Identify Sensor-based Rules for Flushing Contaminated Water from a Pipe Network

摘要

A utility manager may become aware of a threat of contamination to a water distribution network through water quality sensor information, which may indicate that a biological pathogen or chemical contaminant was introduced to the network. In response, a utility manager can select a set of hydrants to flush contaminant from the network. As an event unfolds, a decision-maker may not be able to ascertain source characteristics, creating additional difficulties in determining the set of hydrants that should be opened. The research presented here develops a Genetic Programming (GP)-based approach to identify a set of response actions that are based on sensor information, instead of source characteristics, for guiding selection of hydrants. GP is a method within the class of evolutionary computation, and a solution is represented as a combination of values and symbols to represent a computer program for executing computations, such as a mathematical equation. GP is developed in this research to program a list of rules for opening and closing hydrants that will effectively protect public health for an ensemble of contamination events. An ensemble of contamination events is developed based on a set of similar activated sensors. As the public health effects of a contamination event are influenced by a set of complex interactions among consumers, utility operators, and the pipe network, an agent-based modeling framework is used to predict the dynamic location of a contaminant plume during a contamination event and the number of exposed consumers. To identify optimal hydrant strategies to flush a contaminant while considering the complexity of interactions in the system, a simulation-optimization model couples agent-based modeling with GP. Multiple contamination scenarios are modeled to evaluate potential solutions, and the simulation-optimization framework is demonstrated for a virtual mid-sized municipality, Mesopolis.