Forging the future: enhanced assessment of power consumption in steel manufacturing

Abstract

The rapid growth of the heavy industry backbone in China’s Steel Industry (SI) began in the mid-1990s. In 2023, crude steel production amounted to 1.019 billion tons, and the energy consumed by the sector totaled 561 million tons of coal equivalent. This study predicts power consumption in China’s SI using Decision Tree Classifier (DTC), Light Gradient Boosting Classifier (LGBC), and Extreme Gradient Boosting Classifier (XGBC) with data mining techniques, enhanced by the Honey Badger Algorithm (HBA) for improved classification accuracy. This offers three Hybrid Models (HMs) that merge the benefits of advanced machine learning algorithms with optimization techniques. Different performance metrics are applied to evaluate these models, which improves their predictability. Among the various models tested, the LGHB model achieved the highest performance model optimized by the HBA, which was outstanding at the final iterations. Specifically, LGHB obtained 0.9809 from the training set evaluation to 0.9493 from the testing phase, showing its potential for accurate power consumption forecast within the SI.