Contamination source identification based on sequential Bayesian approach for water distribution network with stochastic demands

Efficient identification of the source of contamination in a water distribution network is crucial to the safe operation of the system. In this article, we propose a real-time sequential Bayesian approach to deal with this problem. Simulations are conducted to simulate hydraulic information and the propagation of contamination in the network. Sensor alarms are recorded in multiple simulations to establish the observation probability distribution function. Then this information is used to compute the posterior probability of each possible source for the observed alarm pattern in real time. Finally, the contamination source is identified based on a ranking of the posterior probability. The key contribution of this work is that the probability distributions for all possible observations are organized into a concise hierarchical tree structure and the challenge of combinatorial explosion is avoided. Furthermore, a variation analysis of the posterior probability is conducted to give significance probability to the obtained identification result. The effectiveness of this method is verified by a case study with a realistic water distribution network.