Enhancing Sustainable Energy Practices through Innovation in Steel Industries: A Machine Learning Approach to Consumption Forecasting

Abstract

The steel industry is one of the most energy-intensive manufacturing sectors, and small-scale producers often struggle to manage energy consumption efficiently while meeting sustainability goals. This study investigates the use of machine learning, specifically the Extreme Gradient Boosting (XGBoost) algorithm, to forecast energy usage in a small-scale steel manufacturing plant. A dataset comprising 35,040 instances and nine operational, temporal, and environmental features was pre-processed and used to develop a regression-based forecasting model. The XGBoost model demonstrated excellent predictive performance, achieving R² values of 0.999 on the training set and 0.997 on the test set, with low RMSE and MAE. Feature importance analysis revealed that reactive power factors and CO₂ emissions were the most influential variables affecting energy consumption patterns. The findings confirm that machine learning-driven forecasting can significantly enhance energy management by providing accurate predictions that support operational optimization, cost reduction, and environmental sustainability. This study underscores the potential of integrating intelligent predictive systems into small-scale steel manufacturing to promote more efficient, sustainable energy practices and to provide a foundation for future research on adaptive, real-time forecasting models.