Optimal placement of contaminant detection sensors in water distribution systems considering conflicts of stakeholders

Water Distribution Systems (WDSs) are important and critical structures which transfer safe drinking water to consumers through network pipes. Today, intentional and accidental contaminant events can be considered as threat for public health of society. In this regard, water utility managers increasingly feel the need to detect the contamination. Contaminants can be injected at any times and locations. Therefore, continuous monitoring and investigating of water quality parameters cause increasing inherent safety of WDS against internal threat and deliberate attack. It should be noted that although monitoring all of the nodes is ideal in the network, it is not an economic method. So, optimal sensors placement is investigated as a cost-effective framework for reducing further damage of contamination events. Also, optimal designing of detection sensors reduces the network vulnerability through maximizing the probability of contamination detection and minimizing affected population. In the last decade, optimal placement of contamination detection sensors in water distribution systems was investigated in a large number of studies. Based on these strategic problems, confliction of interests and priorities among involved stakeholders can cause challenges. In this study, a multi-objective simulation-optimization model is developed based on the EPANET water quality simulation model and NSGA-II optimization. In this paper, deliberate contamination injection scenarios are generated using Monte-Carlo simulation model and simulated by EPANET water quality simulation model. It’s necessary to consider uncertainties in deliberate contamination injection to investigate WDS in critical conditions, including: mass of contamination, duration of injection, time of injection at a day and location of injection. The output data of EPANET was stored and used as input for multi-objective optimization model to obtain trade-off curve among utility functions of stakeholders. Storing all water quality matrix for single injection, considering affected population and detection time in database and then using them in optimization model provide a way to deal with the high volume of simulation data and reduce running time of the model. The aforementioned database was imported in the optimization model to obtain the optimum placement of the sensors among all injection scenarios with respect to utility functions. The main object of this paper is presenting an appropriate process to obtain optimal layout and number of detection sensors in WDS which should maximize the satisfaction of involved stakeholders. In this study, Water and Wastewater Company, National Disaster Management organization and Ministry of Health and Medical Education are the three main stakeholders in the optimization of number and location of sensors in WDS. For this propose, three utility functions, which are linear combinations of four objective functions, are assigned to the three stakeholders. The mentioned utility functions are equal to sum of weighted normalized objective functions which represent utility of stakeholders. These objective functions are the number of sensors, detection time, affected population and probability of undetected events. In the proposed methodology, layout and number of sensors not only integrate for obtaining water security goals, but also for satisfying stakeholders based on their utility function. Solving the NSGA-II optimization model results a trade-off curve among utility function of the involved stakeholders. Two approaches are utilized to determine a compromise solution on trade-off curve, including social choice theory and fallback bargaining method. According to the desirability of the objective function from stakeholders’ point of view, different weights allocated to objective functions. Then, Social Choice Theory and Fallback Bargaining Methods are applied to choose a point within the interaction curve which minimizes the conflict among stakeholders. Finally, based on the best compromise solution among all methods, 6 sensors are selected which scattered in water distribution system, properly. To evaluate the efficiency and applicability of the proposed methodology, it is applied to real-world WDS, Lamerd city, by arsenic injection as deliberate chemical contaminant injection. Based on the 6 selected sensors in Lamerd WDS, the values of affected population, detection time and probability of undetected events are equal to 4735 persons, 33 minutes and 13.6 percent, respectively. In this simulation-optimization model, injection scenarios are assumed to occur at only one node of WDS at a time. Therefore, future research should be considered in contamination injection from more nodes at one time or different times. It’s necessary to consider uncertainties in these deliberate contamination injections to investigate WDS in early and critical conditions, including mass of contamination, duration and time of injection at a day and location of injection.