Chlorine decay modelling and contamination identification strategies for water distribution systems

Increased diligence in identifying pathogen-related outbreaks is demonstrated, such that the number of infections arising from water supplies is much larger than previously acknowledged. For protection against risks of microbiological impacts, the chlorine residual within the distribution system is important. Understanding chlorine decay mechanisms within water distribution systems greatly assists efforts to maintain chlorine residual, while simultaneously minimizing disinfectant byproducts. Both chlorine bulk decay and wall decay mechanisms in water distribution systems are studied herein. A corrected and extended two-component second-order chlorine bulk decay model is developed and applied to wall decay and overall decay studies. A novel procedure based on Bayesian statistics, Monte Carlo Markov Chain and Gibbs Sampling is developed for parameter estimation. This new approach is shown to be efficient and effective in data analysis, and can be easily applied to other water distribution systems.; Monitoring is another very important way to secure a water supply against the effects of chemical and biological contaminants, which may be accidentally or intentionally added to a water distribution system. Because only a limited number of sensors can be installed in a water distribution system, it is necessary to have an algorithm to help in the selection of optimal locations for monitors to provide the best capability for detecting contaminants. This research proposes a novel GA (Genetic Algorithm) approach in conjunction with data mining to solve the multi-objective sensor placement problem. The proposed algorithm was successfully tested on three real-world water distribution systems. The results are then used in the next step: contamination source identification. Once contamination occurs, it is essential to have a methodology to identify the source of contamination. This research developed a novel data mining approach in conjunction with maximum likelihood approach to identify the intrusion location with limited sensor data. Uncertainties, including water demand uncertainty, measurement uncertainty, and modelling uncertainty, are all considered in the proposed method. Applicability of the proposed algorithm on single intrusion events and multiple-intrusion events are evaluated. The effectiveness of the proposed method is demonstrated by application to a network model for a real municipal water supply system.