Document Type
Presentation
Publication Date
8-1-2014
Abstract
In order to bring remarkable benefits to operation and management of water networks, the analysis of sensed data can be used to locate water leaks using of a model-based methodology. However, the number of sensors installed is usually limited because of budget constraints and hence a strategy for optimizing their number and placement is required. This optimization is tightly coupled to the performance of the real-time model-based leakage diagnosis operation and hence the former should consider the requirements of the latter: (1) high distinguishability among all potential leaks to be detected; and (2) strong robustness in front of model-reality mismatches and other uncertainties. This paper describes a model-based pressure sensor placement optimization technique that focuses on the previous aspects and addresses practicality issues that arise in a real deployment. The technique uses an optimization method based on Genetic Algorithms that, unlike most common approaches in literature, avoids using a binary reasoning process. This increases the information granularity resulting in an improvement of both the leak distinguishability and the method robustness. Moreover, the technique also addresses the practical concerns by deriving an enhanced cost function. Finally, the method is validated in a District Metered Area of the Barcelona water distribution network. Results indicate that a good enough detection accuracy can be achieved with a low number of optimally placed sensors.
Comments
Session R53, Water Distribution Networks: Operations and Sensor Placement