Using machine learning for anomaly detection in drinking water from distribution network

To guarantee public health, it is crucial to control the quality water in the distribution network, since anomalies can occur and lead to health and safety problems for consumers. Collecting and analyzing data are necessary to quickly detect water quality anomalies from drinking water network. Smart sensors allow to collect reliable data on water quality at different points of the network. Models built from machine learning algorithms would allow to evaluate drinking water quality. This work proposes an efficient solution for detecting anomalies related to water quality through physico-chemical parameters data with sensors and machine learning algorithms. An architecture for collecting, storing and analysing water quality parameters data is proposed. To evaluate drinking water quality, machine learning algorithms are studied. Some of these algorithms are implemented to propose an accurate model for detecting water quality anomalies on drinking water distribution network. Results of our experiments and comparison with existing propositions show the effectiveness of our selected model for anomalies detection on water potability evaluation.