Adjoint-Based Probabilistic Method for Source Identification in Water Distribution Systems.

摘要

The events of September 11, 2001 have increased the focus on protecting utilities and infrastructure from acts of terrorism. For water utilities, this increased focus has led to researching more efficient and effective methods for finding the source of contamination in the event of contaminant intrusion. Better source identification can significantly reduce both the population affected by water contamination (with subsequent loss of service) and the resources required to mitigate the spread of contamination. Water contamination and/or loss of service have a clear impact on public welfare (both physical and psychological), whether it is due to terrorist activity or accidental contamination.;Source identification can be accomplished using system observations (i.e. the location, time, and magnitude of contamination in the system) and modeling software, such as EPANET. We develop an adjoint-based probabilistic method which uses the system observations as the input information and propagates the information in a backward simulation to determine all potential contamination node and release time scenarios for a system observation. By using multiple system observations and conditioning the results using the system uncertainty and the potential range of source masses, we probabilistically determine the true source node and contamination time.;We develop and test the adjoint-based probabilistic method for source identification in water distribution systems with pipes, nodes, tanks, and pumps; steady and transient flows; perfect and imperfect sensors; and complete and incomplete mixing at the nodes.