Data-driven models for phosphorus forecasting in wastewater treatment plants: A tool to enhance operation

The discharge of effluents with high phosphorus concentrations into water bodies can lead to significant environmental problems. Addressing this issue is critical, particularly in developing countries, where independent wastewater treatment plants (WWTPs) are prevalent and often lack sensors for online monitoring, making their operation and control more difficult. This study presents a systematic methodology for developing data-driven models for time series phosphorus forecasting. Using historical data from an edible oil WWTP, various machine learning (ML) and deep learning (DL) techniques are evaluated, including Linear Regression (LR), Random Forest (RF), Support Vector Regression (SVR), and Long Short-Term Memory (LSTM) neural networks. Given the non-linear nature of wastewater treatment processes, several feature selection methods besides Pearson correlation are explored, such as Spearman correlation, RF feature importance ranking, and causal inference for time series. The models are evaluated across different phosphorus concentration ranges, as errors in predicting high concentrations have a greater impact on plant operations. Results show that LSTM networks with selected features outperform other models in forecasting next-day phosphorus concentration, though challenges remain in accurately predicting peak concentrations. The proposed methodology can be extended to develop data-driven models to predict other effluent quality parameters, showing their potential to enhance WWTP operation. The presented approach is particularly useful for plants with limited resources, providing innovative solutions for problem-solving and regulatory compliance.