Optimal Sensor Placement In Water Distribution Systems

It is very important to place a limited number of water quality sensors at crucial nodes in a water distribution system for controlling the water quality at the customer' s taps. This thesis aims to develop the methodology and solution procedure for the water quality sensor placement in water distribution systems in virtue of location theory and modern heuristic optimization algorithm such that it can provide some basic and significant suggestions for the set-up of early warning system.The deterioration of water quality in a distribution systems has two main situations: (1) self decay or growth of water quality constituents that takes place during the transport process makes the water quality deteriorative(traditional situation); (2) an accidental or intentional external pollutant intrusion makes the water quality deteriorative(nontraditional situation).A notion of partial coverage is introduced for the optimal placement of water quality monitoring stations in water distribution systems under traditional situation. A single coverage criterion is extended to a coverage criterion interval. It assumes that the water quality at a particular upstream node can be partially inferred by the water sampled at some downstream nodes if it delivers the proportion of water to the sampled nodes that are between the minimum coverage criteria and maximum coverage criteria. Based on the notion of partial coverage, an optimal mathematical model is proposed to locate the water quality monitoring stations in a water distribution system, and a discrete binary particle swarm optimization algorithm is proposed to solve the model.In succession, this thesis gives emphasis to the research of the sensor placement in water distribution systems under nontraditional situation. Firstly.the coverage problems of sensor placement are proposed, including set covering problem and maximal covering location problem. In order to solve the maximal covering location problem, a hybrid particle swarm optimization (HYPSO) algorithm is proposed, which incorporates the basic particle swarm optimization algorithm with the crossover, mutation operators of Genetic Algorithm. HYPSO algorithm can quickly find the optimal solutions.Secondly, to solve the difficulty of evaluating the attack probability of one node, a specific contaminant is injected at the node, the health impact on the human being resulting from that can be computed according to the placement objective at the end of the simulation. The larger the placement objective value is, the bigger the probability of its node being selected is. A normalized index of average coordinates of sensor nodes is introduced as a simple measure of the placement results, then the cumulative distribution functions of normalized average coordinates of sensor nodes under different placement objective and contaminative scenarios are obtained with the help of a case.Lastly, a multi-objective optimization model called SPM-R is proposed to the sensor placement in water distribution systems, which considers the reliability of sensor placement. SPM-R model tries to tradeoff between maximizing detect likelihood of accidental or deliberate contamination events and maximizing the reliability of sensor placement. That is. if one sensor is malfunctioned, other sensors can continue to detect the contamination event and the resulting increment of detect time is within allowed bound. The solution procedure is detailed illustrated based on Non-Dominated Sorted Genetic Algorithm-II (NSGA-II) and the sensitive analysis of parameter is also implemented. After that, SPM-R model is extended to SPM-R-G model which can consider various sensor placement objectives.