Soft Sensor for Determining Municipal Wastewater BOD and COD from Excitation-Emission Matrix Coupled with Partial Least Square Regression

Abstract

Biochemical Oxygen Demand (BOD) and Chemical Oxygen Demand (COD) are critical indicators for assessing the organic pollution load in municipal wastewater. However, conventional BOD (BOD₅) and COD analytical methods are slow, hazardous, or require extensive laboratory preparation, limiting their use for real-time monitoring and rapid decision-making in wastewater treatment plants. This study proposes a soft-sensor approach based on fluorescence Excitation–Emission Matrix (EEM) spectroscopy coupled with multivariate analysis to rapidly estimate BOD and COD in municipal wastewater. Wastewater samples were collected from a treatment plant and pre-processed through filtration, pH adjustment, dilution, and scatter removal. The resulting three-dimensional EEM datasets were analyzed using Parallel Factor Analysis (PARAFAC) to identify dissolved organic matter (DOM) fluorophore components, and Partial Least Squares Regression (PLSR) was applied to develop predictive models for BOD and COD estimation. PARAFAC identified two dominant fluorophore groups: a tryptophan-like, protein-like component (Ex/Em ≈ 275/340 nm) and a humic-like component (Ex/Em ≈ 350/437 nm), consistent with known wastewater DOM profiles. While PARAFAC successfully characterized DOM, it showed limited predictive capability for BOD estimation when used alone. In contrast, PLSR demonstrated strong predictive performance, effectively reducing data dimensionality and correlating EEM spectral variations with measured BOD and COD values. The integration of EEM fluorescence spectroscopy with PLS regression shows high potential for developing rapid, reagent-free, and real-time soft sensors for wastewater monitoring, enabling timely operational decisions in wastewater treatment plants.