Regression modeling for rapid prediction of wastewater bod5

Abstract

It is known that the biochemical oxygen demand (BOD5 ) is not directly achieved because it takes a standard procedure of 5 d, and since simple variations in the biological processes in wastewater treatment plants will drastically change the required BOD5 output; the main target in this work is to find measurable parameters at the effluent stream to find the effluent BOD5 easily. Multi-linear regression is proposed here as a simple mathematic entity to predict the effluent BOD5 . In the first step, the data of the Irbid wastewater treatment plant was collected for 10 y. From the available measured data, it is aimed to find the effluent parameters that are correlated with the effluent BOD5 . The elected effluent quality parameters are dissolved oxygen (DO), pH, temperature, flow rate (Q), total suspended solids (TSS), and chemical oxygen demand (COD). These parameters are examined to check their correlation with the BOD5 . The worked data is 114 sets for each quality parameter; of which, 96 sets are used for training, and 18 sets are used for validation. By using the Pearson correlation, it is found that the BOD5 is correlated with the following parameters: Q, TSS, DO, and COD. By using a series of multiple linear regression, it is found that the only significant correlated parameters are the COD, and the TSS. The evolved BOD5 prediction model is a function of COD and TSS, and has the value of Pearson correlation R of (0.97), coefficient of determination R2 of (0.94), P-values of (<0.05), and the significance level of (6.95E-59). It can be concluded that the obtained model can be applied as an automated system that predicts the effluent BOD5 and tunes the parameters together with the required treatment efficiency.