Predicting unregulated disinfection by-products in water distribution networks using generalized regression neural networks

ABSTRACT

Disinfection by-products (DBPs) formation in water distribution networks (WDNs) is a common type of water quality failure. A reliable DBPs modeling can be a way to prevent a water quality failure. In this study, generalized regression neural network (GRNN)-based models were developed to predict the occurrence of three unregulated DBPs i.e. dichloroacetonitrile (DCAN), trichloropropanone (TCP), and trichloronitromethane (TCNM). Water sampling data of several WDNs were used to develop models. Water quality parameters and regulated DBPs were used as predictors to models. The results were validated and verified. Besides, key predictors were identified followed by the sensitivity analysis. The coefficient of determination (R²) of GRNN-based models was >75% for DCAN and TCP; whereas for TCNM, the R² < 45% was observed. The GRNN-based models exhibited better prediction accuracy compared with recently developed multiple linear regression models. The proposed framework can be used to develop models of other contaminants.

KEYWORDS:

• Unregulated disinfection by-products
• water quality
• water distribution networks
• artificial neural networks
• generalized regression neural network

Acknowledgements

The authors are thankful to the Natural Sciences and Engineering Research Council of Canada for supporting this research. The authors would also like to thank Industrial Research Chair Program in Drinking Water (CREPUL), at Laval University for collecting and providing the sampling data for this research.

Disclosure of potential conflicts of interest

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed https://doi.org/10.1080/1573062X.2021.1925707