Predictive Modelling of Electrocoagulation Efficiency Using Supervised Machine Learning for Turbidity Reduction in Municipal Wastewater Treatment

Résumé fourni par la source

Wastewater clarity is an important quality indicator for effluent management in both industrial and municipal treatment facilities. In recent years, electrocoagulation has emerged as an alternative approach for wastewater treatment whereby it uses electrical potential to induce suspended particle aggregation and removal. Although EC technology shows promising results for contaminant removal in synthetic wastewater, understanding the relationship between operational parameters and treatment outcomes remains challenging when applied to real wastewater with fluctuating quality. This study explores the potential of using supervised machine learning to predict turbidity reduction performance in real wastewater samples. Experimental data was collected at different levels of applied voltage of$10 \mathrm{V}, 15 \mathrm{V}, 20 \mathrm{V}$and 25 V, and the resultant turbidity was monitored over time. The experimental dataset underwent several variations of supervised machine learning modelling to identify the most suitable model for the process. The Random Forest model was found to outperform Linear Regression, XGBoost and Support Vector Machine (SVM), achieving the highest$\mathbf{R}^{\mathbf{2}}$value of$\mathbf{0. 9 0 4 9}$for turbidity reduction efficiency. The model effectively captured the complex non-linear relationships between electrocoagulation parameters and turbidity outcomes. Hence, this research establishes a predictive framework for EC implementation in wastewater treatment, with the potential for specific voltage recommendations based on required treatment duration.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.