Prediction of total phosphorus removal in hybrid constructed wetlands: a machine learning approach for rice mill wastewater treatment.
Journal:
Water environment research : a research publication of the Water Environment Federation
Published Date:
May 1, 2025
Abstract
Efficient prediction of pollutant concentrations in constructed wetlands is critical for optimizing treatment performance, yet existing methodologies often fail to account for the influence of meteorological conditions and flow rate variations in real-world scenarios. This study addresses this gap by developing predictive models for Total Phosphorus (TP) removal in hybrid constructed wetlands (HCWs)-combining vertical subsurface flow and free water surface flow-treating rice mill wastewater. Four modeling techniques: Multiple Linear Regression (MLR), Artificial Neural Networks (ANN), Support Vector Machines (SVM), and Random Forest (RF), were used to predict the best-fit model for the removal of the TP based on specific conditions. The study evaluates the impact of key parameters such as temperature (TEMP), hydraulic loading rate (HLR), initial concentrations of chemical oxygen demand (CODin), total nitrogen (TNin), total phosphorus (TPin), and turbidity (TBin) on TP removal efficiency. The results revealed that the SVM_rbf model achieved the highest predictive accuracy with an R value of 0.735763, followed by ANN (R: 0.73), RF (R: 0.721298), and MLR (R: 0.689199). This research highlights the potential of machine learning in enhancing the quantification and prediction of TP reduction in HCWs, offering a robust framework for improving wastewater treatment performance under varying environmental and operational conditions. PRACTITIONER POINTS: Machine learning models (SVM, ANN, RF, and MLR) were applied to predict total phosphorus removal in hybrid-constructed wetlands treating rice mill wastewater. The SVM_rbf model demonstrated the highest predictive accuracy (R = 0.7358), outperforming ANN, RF, and MLR in TP removal prediction. The study analyzed the impact of temperature, hydraulic loading rate, CODin, TNin, TPin, and TBin on phosphorus removal efficiency. The integration of ML with wetland engineering principles enhances HCW design and operational efficiency for sustainable wastewater treatment. This study fills a critical gap by providing a standardized experimental setup for ML-based TP removal prediction, improving model reliability and applicability.