Integrated enviro-economic optimization of solar-powered electrocoagulation for sustainable nitrate removal from...

This study investigates the use of machine learning (ML) models-Linear Regression (LR), Support Vector Regression (SVR), Gradient Boosting (GB), K-Nearest Neighbour (KNN), Random Forest (RF), Artificial Neural Network (ANN), Multilayer Perceptron Regressor (MLPR), and Decision Tree (DT)-to optimize and predict energy consumption in the electrocoagulation (EC) process for nitrate (NO₃^-) removal from groundwater. Alongside these data-driven approaches, Response Surface Methodology (RSM) with a Box-Behnken design (BBD) was applied to statistically evaluate the operational parameters. Among the tested models, the GB model showed the best performance with R² = 0.9924, Mean Squared Error (MSE) = 0.0135, Root Mean Squared Error (RMSE) = 0.1164 and Mean Absolute Percentage Error (MAPE) = 8.7418. Optimal operating conditions were identified to achieve NO₃^- removal below permissible limits. The specific energy consumption under these conditions corresponds to operational costs of 0.46, 0.55, and 0.25 $/m³ for Al, Fe, and Al/Fe combination electrodes, respectively. These results indicate that EC powered by photovoltaic energy (PV) can serve as a sustainable and decentralized solution for groundwater treatment in rural areas, offering both high removal efficiency and economically favourable operation.