Data mining for water - Executive Summary - Data mining can help optimise water treatment
An exploratory research project was carried out within the Water Treatment theme on the possibilities of big data in drinking water treatment. This resulted in a literature overview of different application areas, such as the characterisation of water-quality parameters (soft sensors), the optimisation of coagulant dosage rates and the prediction of membrane fouling. One process was elaborated: operational data from a coagulation process followed by sand filtration were used to train a model for the purpose of predicting the pressure build-up in the sand filters, based on water-quality parameters and operational parameters. The model permits the optimisation of filter run time, backwashing events and iron dosage rates.
Interest: increased volume of data and more chances to use them As a result of the increasing digitisation of society the volume of stored data is growing sharply. In many sectors the stored data are already being applied to optimise processes. Data on drinking water treatment are also being increasingly collected. This raises the question about the extent to which big data can be applied to water treatment to improve treatment processes.
Approach: literature scan and use of treatment data to train model A short literature scan was done on the application possibilities of big data in (drinking) water treatment. Data mining techniques were also applied to data from drinking water production; these practice data were provided by PWN and concerned a pre-treatment process, consisting of a coagulation step followed by a sand filter. The data included water-quality data (pH, turbidity, temperature) and operational data (flow, Fe dosage rates), from several parallel filters over a period of 2 years. These data were used to train a data model.
Report: This report is described in the report Datamining voor de zuivering (BTO 2019.001).