Predictive Modeling and Machine Learning for Optimal Wastewater Treatment Performance

This paper proposes new wastewater treatment performances predictive model based on a novel Outlier-robust extreme learning machine (ORELM). The new proposed ORELM model is proposed to resolve some limitations of the standard extreme learning machine (ELM) model and to enhance their performances. The ORELM is than developed for predicting influent five days biochemical oxygen demand (BOD5) measured at the inlet of the wastewater treatment plant (WWTP). The ORELM is developed using several influent wastewater quality variables namely, wastewater temperature (Tw), pH, specific conductance (SC), suspended solids (MES), and wastewater chemical oxygen demand (COD). The performances of the ORELM were compared with those of the regularized extreme learning machine (RELM), weighted regularized extreme learning machine (WRELM), and the standard ELM models. The models were evaluated and compared using Pearson correlation coefficient (R), Nash–Sutcliffe efficiency (NSE), root-mean-square error (RMSE), and mean absolute error (MAE), and the historical wastewater dataset are used for training and testing the proposed models using a ratio of 70% and 30%, respectively. The comparison demonstrates the superior performance of the proposed ORELM model compared to the ELM, RELM, and WRELM, exhibiting an R, NSE, RMSE and MAE values of 0.807, 0.649, 63.83, and 45.21. Moreover, obtained results in the present study have demonstrated the potential benefits of the proposed models.