Methodology to optimally place pressure sensors for leak detection in water distribution systems using value of information and entropy

The presence of leaks in water distribution networks (WDN) is a very common problem that impacts these systems in a variety of ways, including the loss of drinking water, the deterioration of the quality of supply water due to the intrusion of substances into the system, the structural damage caused by infiltrations, and the uneven distribution of pressures and flows in the network, among others. For this reason, the use of sensors to identify possible where leaks could be located, and to operate the network is imperative. However, placing the sensors in the network is not a trivial problem, due to the complexities of the networks, the uncertainties around them, and the limited resources. Although many researchers have proposed different methodologies for locating sensors, methods accounting for the costs of the decision-making situation and the nature of the decision-maker are very limited. Moreover, the existing methods largely rely on demand-driven analysis approaches, which limit their applicability. This thesis presents a methodology to determine the optimal location of pressure sensors for identifying the existence of leaks in water distribution systems. The methodology is based on two unrelated concepts, Value of Information (VOI) and Entropy. The former is common in areas such as medicine and economics but with very limited application in water distribution systems. The latter is a concept from Information Theory, which have been used in the past for locating sensors in the water environment. The proposed methodology, which has a simplified framework with practical interpretation, involves the judgment of decision-makers and switches from the assumption of a demand-driven analysis to a more realistic approach in which the demands are a function of the pressure's system (pressure-driven analysis). Additionally, the methodology considers, to a certain extent, the possible sources of uncertainty and parameter variation along time, towards robust solutions. The methodology is applied in the case study of the city of Braila, in Romania, which has the goal of reducing water losses caused by leaks by a value of 50 L/h/km. Results show that the sensor locations obtained agree with previous methods in simpler networks, and that well distributed sensors are obtained, although the performance of the method is depending on the selection of the threshold in pressure that is considered critical. A by-product of this research is the Sensor Detection Value (SDV) index, which is based on VOI concepts and that facilitate the application of the methodology and the interpretation of the results