Optimizing the chemical removal of phosphorus for wastewater treatment: Insights from interpretable machine learning modeling with binary classification of elasticity and productivity

Ensuring compliance with total phosphorus (TP) discharge standards is essential in wastewater sector to alleviate eutrophication. This study focused on optimizing chemical removal of TP from a typical wastewater plant (WWTP) where poly aluminum chloride (PAC) is used after anaerobic-anoxic-oxic technology. With PAC consumption and TP removal in one-year daily data combined as input-output system, binary classifications of decoupling and congestion patterns representing elasticity and productivity were conducted to mitigate irregular data mappings caused by inaccurate dosing. Through interpretable machine learning (IML) modeling, influent conditions were recognized as significant factors. Biochemical oxygen demand to TP ratio exceeding 36.07 and loading capacity rates departing 99.46 %∼106.64 % increased decoupled and congested probability. These findings highlighted the adjust on PAC dosage for redundancy prevention according to varied influent conditions. The evaluation and modeling workflow with IML emphasized the need for systematic optimization to achieve sustainable WWTP operations and low-carbon development in wastewater sector.