On-site sensoring & monitoring

Determining nutrient concentrations with electronic finger prints and an algorithm based on an artificial neural network.


Within TKI Water Technology’s Smart Water Systems theme, a method has been developed which processes data using an algorithm based on an artificial neural network. Data on water acidity, oxygen content and temperature are combined with data from an electro-chemical finger-print sensor (Liqum, Finland), which registers seven electro-chemical potentials that are sensitive to ions in the water. The combination of these data produces, thanks to the algorithmic method, important information about the concentrations of nutrients like ammonium, nitrate and phosphorus.


Determining nutrient concentrations in surface water and wastewater is important for water-quality monitoring, wastewater treatment process management and safeguarding quality guidelines for surface water. The method developed, with a smart combination of sensors and models, contributes to maximising the efficiency and effectiveness in the application of water treatment technology: the objective of TKI’s Smart Water System theme.


The method was developed in partnership by KWR Water Research Institute, Interline Systems B.V., Waterboard Hollands Noorderkwartier and Waterlaboratorium Noord. They selected and trained the algorithms to extract the concentrations of ammonium, nitrate and phosphorus nutrients from the data, and to compare the observed concentrations with reference measurements from Waterboard Hollands Noorderkwartier and the de Punt water intake area of the Groningen water company. A big advantage of this new method with sensors and an algorithm is that it supplies a continuous stream of current data on the nutrient concentrations: this is not feasible with the traditional, labour-intensive wet chemical method. The research delivered a proof-of-concept: it is possible to precisely measure dosed nutrients, as long as the water quality does not vary greatly.

Alarm system

In the case of wide variations in water quality, the method does not provide any precise forecasts, because the data found differ significantly from the data with which the model is trained. However, with due modifications, the method works very effectively in giving an alarm signal for sudden alterations in water quality, environmental factors or sensor behaviour. The intention is to further develop this application in a TKI follow-up project.

On-site sensoring & monitoring