Decision Making under Information Constraints

The purposes of placing sensors in water distribution systems vary from complying with water quality regulations, monitoring accidental contamination events, and detecting intentional contamination events. The information available to water utilities varies: some only have information on the topological structure of their systems; some have calibrated hydraulics models for their systems; and some even have water quality models for their systems. Different sensor location models have been proposed for the different purposes, targeting the objectives of maximizing the coverage, minimizing the contaminant travel time prior to detection, minimizing the contaminated water consumed prior to detection, and/or minimizing the population at risk prior to detection. These different sensor location models require different computational efforts and the availability of information on water distribution systems to different extents. The current sensor location models proposed mainly include graph-theory-based model (network analysis), deterministic optimization models, stochastic optimization models and robust optimization models. In this paper, we develop a protocol to help water utilities to select sensor location models based on the specific purposes of sensor placement, the information available to the water utilities, and the computational requirements of the models.