Author(s): Juhwan Kim; Jinwoo Song; Seungmin Lee; Soojin Kim

Linked Author(s):

Keywords: Machine learning; Residual chlorine concentration; Water treatment plant

Abstract: Machine learning models are developed to stabilize the residual chlorine concentration of the outflow section from the sedimentation basin for the purpose of real-time monitoring of water quantity and quality data and intelligent control of the chlorination process in water treatment plant (WTP). Multiple regression model, artificial neural network, and random forest among artificial intelligence algorithms were used to predict the residual chlorine concentration in the outflow section of the sedimentation pond of the B Water purification plant, and the results were compared and analyzed. Residual chlorine concentration in the sedimentation basin, water temperature, turbidity, pH, electrical conductivity, inflow volume, and alkalinity data are collected and used as input variables, and residual chlorine concentration in settling basin as output variable. These results showed that the random forest model made the most accurate predictions for the B WTP. The mathematical model, multiple regression, performed the worst in terms of goodness of fit. The results can be shown due to the difference in scale and dimensionality of water quantity and quality data and the variability of chlorine input due to seasonal water quality changes. As results, it is concluded that a decision tree-based model such as a random forest is suitable for the application of artificial intelligence algorithms in the B WTP. Based on the results from this study, it is expected that the residual chlorine concentration in the outflow section of the sedimentation basin can be maintained consistently by adjusting the chlorine injection in real time in the B WTP.

DOI: https://doi.org/10.3850/iahr-hic2483430201-28

Year: 2024