Valve status detection using smart meter temperature and flow

Within recent years, smart meters have been installed at an increasing rate at the consumers’ households. The main drivers are improved collection of consumption data, and prospects for optimizing utility performance, e.g. reduced leakage. Some smart meters also measure the temperature of the consumed water, mainly serving as water quality indicator. However, we believe temperature data has a yet unexplored potential: depending on the time of the year, the drinking water is heated or cooled down on its way to the consumers and the temperature data might thus store valuable information about the water’s path throughout the water distribution network (WDN).
The manual process of identifying the correct valve status and location is a time consuming task with many resources wasted checking status of rarely used valves. Here, WDN models can be used as a tool to identify discrepancies between simulated and observed values and guide the utility to locations with doubtful valve settings. In Denmark, pipe diameters are usually larger than necessary due to false expectations of increasing future water demands. Consequently, low head losses are seen in the WDNs, making it difficult to detect leakages and incorrect valve settings based on hydraulics alone. Genetic algorithms (GAs) have been used to identify the correct valve statuses (e.g. [1], [2]), but during such and other WDN calibration processes, sufficient volumes of high quality data are often a restraining factor [3]. For example, usually pressure data inside smaller sub-networks is missing. We investigated how temperature data collected by smart meters can support an automated search for valve status identification and error correction.