Uncertainty of Contamination Source Characterization within Water Distribution Systems

Current research on contamination source identification within water distribution systems assumes a perfect hydraulic model and accurate contaminant concentration measurement. In reality, hydraulic models always contain certain level of uncertainty, and it may be unrealistic to have contaminant specific sensors deployed in water systems. More likely, some contamination detection algorithms based on different water quality measurements will be used to determine whether water is contaminated or not before reaching the monitoring station. Such contamination detection algorithms will provide positive (existence of contamination) and negative (non existence of contamination) signals at some sampling frequency. In this paper, the impact of both hydraulic model and contamination detection uncertainty on contaminant source identification is studied. In hydraulic model, the demands at junctions are assumed to be uncertain, and the contamination detection algorithms uncertainty is characterized by false positive and false negative rates. The contamination source identification approach applied in this paper uses backtrack algorithm to find the hydraulic connectivity between possible sources and positive/negative measurements at monitoring stations and characterizes all the possible sources with some heuristics based on the number of positive and negative measurements they are connected to. If a location during one time step is only connected to negative measurements, the chance that it is one of the contamination sources is very low. On the other extreme, if a location during one time step is connected to all the positive measurements but no negative measurement, it is very likely that it is one of the true source or even the only source. The impact of hydraulic model and contamination detection uncertainty on such heuristics associated with all the possible sources are analyzed through simulation study.