IWA WaterLoss2022 Conference

Water utilities worldwide are under constant stress to reduce water loss due to urbanization, population growth, and climate change. Globally, Water Distribution Networks (WDNs) lose about 30% of the treated water on an average during supply. In addition to the amount of water lost, leaky WDNs consume additional energy and increase the risk of contamination. Deteriorating pipes and pipe network elements such as valves and joints, as well as improper pressure management are the main contributing factors for water loss in WDNs. Due to the increasing concern about water loss, leakage detection and localization have been widely researched in recent decades, both in continuously pumped and intermittently pumped systems.The techniques used for leakage detection and repair range from conventional methods with direct inspection on-site to model-based optimization methods. In the present era of low-cost sensors and the availability of high computing power, the transformation of WDNs into smart water systems is higher than ever. This has led to the research and development of data-driven and hybrid methods for solving leakage detection and localization methods. Irrespective of the class of methods used, their ultimate goal can be distilled primarily into two questions - a) How quickly and reliably can the presence of leak(s) be detected, and b) How accurate and precise can the location and size of the leak(s) be estimated?Answers to these questions include uncertainties inherent to the methods and models used, their underlying assumptions and necessary abstractions. Although much research has been done for many years to reduce uncertainties in leakage detection and localization, a comprehensive study using a consistent terminology of their types, sources, and effects on the outcome are missing. The main contribution of this work is to discuss (i) why there are uncertainties in the formulation of leakage detection and localization problem, (ii) identify the sources and types of uncertainties for different classes of modeling approaches (i.e., data-driven vs. model-based), and (iii) provide a brief review of their influence concerning error bounds from existing literature.

Sensors used for wastewater flow measurements need to be robust and are, consequently, expensive pieces of hardware that must be maintained regularly to function correctly in the hazardous environment of sewers. Remote sensing can remedy these issues, as the lack of direct contact between sensor and sewage reduces the hardware demands and need for maintenance. This paper utilizes off-the-shelf cameras and machine learning algorithms to estimate the discharge in open sewer channels. We use convolutional neural networks to extract the water level and surface velocity from camera images directly, without the need for artificial markers in the sewage stream. Under optimal conditions, our method estimates the water level with an accuracy of ±2.48% and the surface velocity with an accuracy of ±2.08% in a laboratory setting—a performance comparable to other state-of-the-art solutions (e.g., in situ measurements).

https://www.mdpi.com/2073-4441/14/3/424