Regression-based analytical models for dissolved oxygen in wastewater

Abstract

The limited freshwater resources and increasing demand for clean water require minimizing organic contamination in wastewater. High levels of biochemical oxygen demand (BOD) in water reduce available oxygen, harm ecosystem biodiversity, and degrade water quality. Here, regression-based analytical models are suggested to minimize organic contamination by estimating desired dissolved oxygen (DO) and dilution factors (df) correlated to the organic decomposition. Training datasets of defined independent inputs (i) ultimate biochemical oxygen demand (UBOD), (ii) minimum BOD_T (BODM), (iii) average BOD_T (BODA), (iv) COD, (v) O₂ consumption (X), and (vi) time (T) were collected and/or calculated based on literature. Results showed that there should be specified oxygen dosing amounts dependent upon BOD₅ levels, noting that BOD₅ and DO₅ are inversely proportional (proportionality might differ based on the microbial concentration). An increase in df is predominated by BOD₅, with df≈9.2 for storm (STM), df≈12 × 10³ for industrial (IND), and df≈18.5-28.5 for domestic (DOM) wastewaters. Mixing/matching between the input features used in training regressors including medium trees (MT) and ensembles boosted trees (EBT) showed high accuracy > 94% for predictor combinations: (i) MT-[UBOD-X], MT-[UBOD-X-T-COD], and EBT-[UBOD-X-T-COD] for DO₅ predictions, and (ii) EBT-[BODM-BODA] and EBT-[BODM-BODA-UBOD-X-T-COD] for df predictions, knowing the general term XX-[a-b-c-d-e-f] has XX = regressor and a,b,c,d,e,f = predictors for the training parameters used as inputs. The models are capable of predicting changes in DO₅ against BOD with deviations 5-10%, whereas a suggested correction factor [Formula: see text] further reduced this deviation to < 5%, where i = 0, 1, 2…6 refers to the BODM datapoint and its corresponding UBOD with the constant α = f(i). The optimized collective models (cubic equations derived for df and DO₅ from BODM that is an exponent function in UBOD) would enable effluent quality evaluation to manage organic contamination, bridging the gap between science and industry best practices.

Keywords: Dissolved oxygen; Organic contamination; Quality control; Regression; Wastewater.