Optimization of hydrochar production from almond shells using response surface methodology, artificial neural network, support vector machine and XGBoost

Abstract

Herein, we investigate how to improve the hydrochar yield via hydrothermal carbonization process, focusing on the use of almond shells as a carbon resource. We have employed various modeling techniques, including response surface methodology, artificial neural networks, support vector machines and XGBoost, to determine the optimal conditions for maximizing hydrochar yield. The results highlight the excellent performance of the ANN and XGBoost models, with high R² values and lower RMSE values, while the SVM model offers insights into the interactions between the factors. The choice of modeling method depends on the specific nature of the problem and accuracy requirements. In summary, this study contributes to a better understanding of HTC processes and opens the way to further optimization of hydrochar performance, with significant practical implications in the energy and environmental fields. It also highlights the importance of further research to explore more modeling techniques.