Online data from the KNMI and CBS in self-learning EDWARD algorithm

Oasen forecasts peaks in water demand with data-driven algorithm from KWR

This spring the Oasen water utility began using a new, data-driven algorithm which, on the basis of population statistics and climate scenarios, forecasts the future development of water demand. Oasen uses this information to shape its future drinking water infrastructure. The EDWARD algorithm was developed by KWR within the framework of the Joint Research Programme with the water utilities, and is continuously fed with customised data: from Statistics Netherlands (CBS) on holiday practices, and from the Royal Netherlands Meteorological Institute (KNMI) on climate change.

The algorithm, which is the product of a unique collaboration between Oasen, KWR and CBS, enables the calculation of possible future scenarios in greater detail than was previously possible. It has been dubbed EDWARD, an acronym which reflects its function: simulating Extremes in Drinking WAteR Demand. Since EDWARD is continuously fed with the most recent data from the KNMI and CBS, any new developments are immediately incorporated into the model forecasts. Thanks to these forecasts, Oasen can properly prepare its drinking water infrastructure for the future. A key success factor for the implementation of EDWARD at Oasen is the availability of detailed external data from CBS and the KNMI, as well as the utility’s internal daily production figures.

Peak water demand determines capacity

Over the last few years, water utilities have put a lot of effort into acquiring a clearer picture of the possible effects of climate change on water demand in the Netherlands. For example, KWR has carried out research for the Dutch water utilities on the influence of climate change on peaks in water demand, because the peaks determine the drinking water infrastructure’s required capacity. This research demonstrated that the influence of climate change is very area-specific, and that demand peaks in 2050 could exceed those currently experienced by up to 22%. This means that in the years to come the water utilities will invest, among others, in the extension of their supply capacity. A crucial element in this regard are the holiday practices in the Netherlands: peak water demand is typically much higher if many people stay at home on hot days or during drought periods. Further information on this research, including a short video, is available on the project page.

Trained with historical data

In order to take account of all relevant meteorological and social factors in forecasting peaks in water demand, KWR trained the self-learning, machine-learning model EDWARD with large amounts of historical data from Oasen.

KWR trained the self-learning, machine-learning model EDWARD with large amounts of historical data from Oasen.

Reliable drinking water provision

The Oasen water utility provides drinking water to the eastern part of the Province of South-Holland. The utility wants to be well prepared for the future.  Disposing of reliable information about the future, with regard to growing water demand and higher demand peaks during hot periods, is essential for Oasen. ‘To maintain today’s high level of supply reliability, we need insight into future peak demands so that we can organise our infrastructure accordingly, and don’t have to ask our customers to restrict their consumption,’ says Bas Bouwman of Oasen. ‘EDWARD gives us current new insights into peak demands, based on new data on climate- and holiday-staggering scenarios and water demand. This will also allow us in the future to ensure that the water keeps flowing from the tap at the right pressure.’