A Machine Learning-based Algorithm for Water Network Contamination Source Localization

Gated graph neural networks for identifying contamination sources in water distribution systems

Contamination events in water distribution networks (WDN) pose significant threats to water supply and public health. Rapid and accurate contamination source identification (CSI) can facilitate the development of remedial measures to reduce impacts. Though many machine learning (ML) methods have been proposed for fast detection, there is a critical need for approaches capturing complex spatial dynamics in WDNs to enhance prediction accuracy. This study proposes a gated graph neural network (GGNN) for CSI in the WDN, incorporating both spatiotemporal water quality data and flow directionality between network nodes. Evaluated across various contamination scenarios, the GGNN demonstrates high prediction accuracy even with limited sensor coverage. Notably, directional connections significantly enhance the GGNN CSI accuracy, underscoring the importance of network topology and flow dynamics in ML-based WDN CSI approaches. Specifically, the method achieves a 92.27% accuracy in narrowing the contamination source to 5 points using just 2 h of sensor data. The GGNN showcases resilience under model and measurement uncertainties, reaffirming its potential for real-time implementation in practice. Moreover, our findings highlight the impact of sensor sampling frequency and measurement accuracy on CSI accuracy, offering practical insights for ML methods in water network applications.

Machine Learning and Simulation-Optimization Coupling for Water Distribution Network Contamination Source Detection

This paper presents and explores a novel methodology for solving the problem of a water distribution network contamination event, which includes determining the exact source of contamination, the contamination start and end times and the injected contaminant concentration. The methodology is based on coupling a machine learning algorithm for predicting the most probable contamination sources in a water distribution network with an optimization algorithm for determining the values of contamination start time, end time and injected contaminant concentration for each predicted node separately. Two slightly different algorithmic frameworks were constructed which are based on the mentioned methodology. Both algorithmic frameworks utilize the Random Forest algorithm for classification of top source contamination node candidates, with one of the frameworks directly using the stochastic fireworks optimization algorithm to determine the contamination start time, end time and injected contaminant concentration for each predicted node separately. The second framework uses the Random Forest algorithm for an additional regression prediction of each top node’s start time, end time and contaminant concentration and is then coupled with the deterministic global search optimization algorithm MADS. Both a small sized (92 potential sources) network with perfect sensor measurements and a medium sized (865 potential sources) benchmark network with fuzzy sensor measurements were used to explore the proposed frameworks. Both algorithmic frameworks perform well and show robustness in determining the true source node, start and end times and contaminant concentration, with the second framework being extremely efficient on the fuzzy sensor measurement benchmark network.

Machine-Learning Classification of a Number of Contaminant Sources in an Urban Water Network

In the case of a contamination event in water distribution networks, several studies have considered different methods to determine contamination scenario information. It would be greatly beneficial to know the exact number of contaminant injection locations since some methods can only be applied in the case of a single injection location and others have greater efficiency. In this work, the Neural Network and Random Forest classifying algorithms are used to predict the number of contaminant injection locations. The prediction model is trained with data obtained from simulated contamination event scenarios with random injection starting time, duration, concentration value, and the number of injection locations which varies from 1 to 4. Classification is made to determine if single or multiple injection locations occurred, and to predict the exact number of injection locations. Data was obtained for two different benchmark networks, medium-sized network Net3 and large-sized Richmond network. Additionally, an investigation of sensor layouts, demand uncertainty, and fuzzy sensors on model accuracy is conducted. The proposed approach shows excellent accuracy in predicting if single or multiple contaminant injections in a water supply network occurred and good accuracy for the exact number of injection locations.