This study quantitatively evaluated and modelled the performance of a sewage treatment plant (STP) using a Polynomial Neural Network (PNN) algorithm, an advanced machine learning approach, to predict effluent quality and optimize treatment efficiency in Ghaziabad, India. The PNN algorithm achieved prediction accuracies (R 2 > 0.95) for Biochemical Oxygen Demand (BOD), Chemical Oxygen Demand (COD), and Total Suspended Solids (TSS) removal efficiencies, demonstrating reductions of 87%, 90%, and 94%, respectively. A total of 6 datasets per parameter were collected and analyzed across three treatment stages (Inlet, Primary stage treatment (PST), Final stage treatment (FST)). Cross-validation and Mean Absolute Percentage Error (MAPE < 5%) confirmed model robustness. The model required refinement for parameters with low variability (pH, Oxygen Absorption (OA), Total alkalinity), highlighting chemical dose nonlinearity. This algorithmic insight supports sustainable wastewater management practices by ensuring compliance with BIS discharge standards.
Singh et al. (Thu,) studied this question.