Author(s): Laura Enriquez
Linked Author(s):
Keywords: Chlorine decay; Water supply network; Symbolic machine learning
Abstract: This work presents the application of a symbolic machine-learning metamodel as an alternative to differential equations approaches for the computation of chlorine decay. The water supply network of Bogota is proposed as case study. The training dataset was obtained from advanced hydraulic and water quality analysis with different chlorine concentrations and decay coefficients. The results using data of all the nodes with water quality variables as inputs provided a general satisfactory performance. However, some of the nodes had a low performance. Hence, individual models for each node including hydraulic variables as inputs were developed, which significantly improved the prediction capacity. The potential of the approach was demonstrated, and further research is required for better generalization.
Year: 2024