Refreshing drinking water thanks to water temperature mapping

Dirk Vries PhD MSc

  • Start date
    01 Jan 2015
  • End date
    31 Dec 2017
  • collaborating partners
    Brabant Water, Neelen & Schuurmans

A new type of water meter makes it possible to monitor water temperature. It provides a minimum temperature value once day, on account of the battery capacity. In the first instance this capacity might seem limited, but a large-scale application makes it possible to draw up a daily ‘heat’ map of a particular area. By combining this map with other measurements taken in the same area, one can monitor the temperature of the pipe water and identify the ‘hot’ and ‘cool’ spots. This allows for the more accurate determination of (possible) risks, and the formulation of more targeted control measures.


Up until now, the monitoring of water temperature in mains networks has only been done on an ad hoc basis, which has made the accurate identification of ‘hot’ and ‘cool’ spots difficult. The idea of the project is to make use of data-driven techniques to distil information and to use it for a dynamic ‘hot/cool-spots’ map.


Relatively limited water temperature data are measured and collected in the mains network. The water temperature at a connection is only measured once a day, on account of the battery capacity of the water meter. And while the region contains thousands of connections, only a few dozen smart water meters are available for this pilot. Moreover, not all the factors that contribute to the local heat transfer – for example, an aquifer thermal storage system – are known and defined. Uncertainties also remain about the flow paths taken by the water between the production company and the client, because there is no reliable way of determining for all the (water) valves whether they are open or shut.


This project applies big data techniques to provide the missing information required to gain a clear picture of water temperature in the mains network. To this end, we make use of information from smart water meters as shown in Figure 1, as well as data on the network itself and on its surroundings. With regard to the latter category, we consider a large number of factors that might influence the temperature at the tap, including air temperature, sunshine hours, population, urban heat-island effect, land-use type, building height, soil type, pipe material and installation year. It is not entirely clear whether only the connection locations are important, or whether the area surrounding the connection also plays a role. To study this, the factors are averaged out, either for a circle surrounding the measurement location or over the most likely flow path taken by the water to the tap.  We employ machine-learning algorithms to extract the most relevant information from the data, and to find the closest possible connection between these data and the temperatures measured by the smart water meters. The connections are tested and used to determine the key factors.

Currently, there are not enough smart meters installed to produce a temperature map covering an entire mains network. The approach taken is however suitable for upscaling and for processing the data from a multitude of smart meters. This offers a new perspective of achieving a more accurate picture of the water temperature and the determining factors. It provides water companies with new possibilities to underpin their management measures and to limit temperature-related health risks to a minimum.