Predicting Regional Wastewater Treatment Plant Discharges Using Machine Learning and Population Migration Big Data

Quantifying the temporal variation of wastewater treatment
plant
(WWTP) discharges is essential for water pollution control and environment
protection in metropolitan areas. This study develops an ensemble
machine learning (ML) model to predict discharges from WWTPs and to
quantify the contribution of extraneous water (mixed precipitation
and infiltrated groundwater) by leveraging the power of ML and population
migration big data. The approach is applied to predict the discharges
at 265 WWTPs in the Guangdong–Hong Kong–Macao Greater
Bay Area (GBA) in China. The major conclusions are as follows. First,
the ensemble ML model provides an efficient and reliable way to predict
WWTP discharges using data easily accessible to the public. The predicted
treated sewage amount increased from 20.4 × 106 m3/day in 2015 to 24.5 × 106 m3/day
in 2020. Second, the predictors, including daily precipitation, average
precipitation of past proceeding days, daily temperature, and population
migration, play different roles in predicting different city’s
discharges. Finally, mixed precipitation and infiltrated groundwater
account for, on average, 1.6 and 10.3% of total discharges from WWTPs
in the GBA. This study represents the first attempt to bring population
migration big data into data-driven environmental engineering modeling
and can be easily extended to predict other environmental variables
of concern.