Biochemical Oxygen Demand Prediction Based on Three-Dimensional Fluorescence Spectroscopy and Machine Learning

Biochemical oxygen demand (BOD) is an important indicator of the degree of organic pollution in water bodies. Traditional methods for BOD₅ determination, although widely used, are complicated and dependent on accurate chemical measurements of dissolved oxygen. The aim of this study was to propose a facile method for predicting biochemical oxygen demand by fluorescence signals using three-dimensional fluorescence spectroscopy and parallel factor analysis in combination with a machine learning algorithm. The water samples were incubated for five days using the national standard method, during which the dissolved oxygen contents and three-dimensional fluorescence spectroscopy data were measured at eight-hour intervals. The maximum fluorescence intensity of three fluorescence components was decomposed and extracted by parallel factor analysis. The relationship between the maximum fluorescence of the three fluorescence components and the BOD₅ values was established using a random forest model. The results showed that there was a good correlation between the fluorescence components and BOD values. The BOD₅ values were effectively predicted by the random forest model with a high goodness of fit (R² = 0.878) and low mean square error (MSE = 0.28). Although this method did not shorten the incubation time, successful BOD₅ prediction was realized by the non-contact measurement of fluorescence signals. This avoids the complicated operation of DO determination, improves detection efficiency, and provides a convenient solution for analyzing large quantities of water samples and monitoring facile water quality.