p-Nitrophenol Degradation Prediction in Microbial Fuel Cells Using RF and GBRT

p-Nitrophenol Degradation Prediction in Microbial Fuel Cells Using RF and GBRT

Arezu Chajaei1 Farzaneh Vahabzadeh2

1) 1. M.Sc. Student, Department of Chemical Engineering, Amirkabir University of Technology (Tehran Polytechnic), Tehran, Iran,
2) Professor, Department of Chemical Engineering, Amirkabir University of Technology (Tehran Polytechnic), Tehran, Iran,

IOI Code : ABEB-ABHKHG
Publication : 4th International Conference on Recent Advances in Engineering, Innovation and Technology - Belgium(eitconf.com)
Abstract :
    Microbial fuel cells (MFCs) have attracted considerable attention as a sustainable bioelectrochemical technology for the degradation of organic pollutants. However, the complex nonlinear interactions between operating conditions and degradation performance make accurate prediction of pollutant removal efficiency challenging. This study presents a machine learning-based approach for predicting p-nitrophenol (PNP) degradation efficiency in a microbial fuel cell system. A dataset extracted from a previously reported experimental study, including the effects of initial PNP concentration, temperature, and pH, was employed for model development. Due to the limited size of the original dataset, a controlled data augmentation strategy based on Gaussian perturbation was applied to enhance model robustness while maintaining the original data characteristics, resulting in an expanded dataset containing 68 samples. Random Forest Regression (RFR) and Gradient Boosting Regression Tree (GBRT) algorithms were developed using MATLAB R2025b. Prior to model training, input variables were normalized, and the predictive performance of the models was evaluated using the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE). Both models demonstrated high prediction accuracy; however, GBRT showed superior performance compared with RFR. The GBRT model achieved R² values of 0.9997 and 0.9897 for training and testing datasets, respectively, with RMSE and MAE values of 1.5829 and 1.2096 for the testing dataset. In comparison, RFR achieved an R² value of 0.9383 with RMSE and MAE values of 3.8792 and 3.4181 in the testing phase. The results demonstrate that ensemble machine learning models, particularly GBRT, can be effectively applied for accurate prediction of p-nitrophenol degradation efficiency in microbial fuel cell systems.

Keywords : microbial fuel cell p nitrophenol degradation machine learning random forest regression gradient boosting regression