Optimizing sensor placement for enhanced observability in water distribution networks

Water is a valuable, but limited resource. At least half of the world’s population lives under high water scarcity. The water distribution networks (WDNs) are essential for ensuring the supply of potable water to communities and urban areas, and leaks represent a major problem for proper water resource management. Many methods, based on the system observability, have been developed to detect these leaks with the aim of minimizing their damage. This property strongly depends on the structure of the sensor network installed, and is calculated using the singular values of the system observability Gramian. The aim of this work is to investigate and demonstrate a strategy for sensors location with observability purposes in WDNs. A tolerable observability degree degradation factor is defined, and the tradeoff between cost and observability is addressed by formulating binary optimization problems. This approach is appropriate when it is desired to maximize the system observability degree and the instrumentation budget is limited. The resulting optimization problems are nonlinear with binary decision variables and solved using a technique based on genetic algorithms. A linearized state-space model of the Hanoi hydraulic network, taken from the literature, is used to show the proposed instrumentation design methodology.