A hybrid intelligent framework for energy consumption classification of mechanical equipment in the steel industry based on machine learning and metaheuristic optimization

Abstract This study presents a robust intelligent hybrid classification model to effectively classify energy consumption levels in the steel industry, thereby improving productivity and contributing to lower energy costs. By evaluating production variables, operational variables, and historical energy use data, the model produces a deep understanding of energy consumption patterns and a framework to make better decisions in achieving energy savings. The proposed approach highlights the combination of Quadratic Discriminant Analysis (QDA) and Decision Tree Classifier (DTC) classifiers with two advanced metaheuristic algorithms, Greylag Goose Optimization (GGO), and Manta Ray Foraging Optimization (MRFO) to optimise the parameters in respective classifiers and enhance the predictive ability of the classification system. The study employed a dataset of 10,000 samples containing nine inputs and one output, across 200 iterations. Experiments demonstrated that the hybrid DTC-GGO (DTGG) outperformed all the other finalist comparison systems with an overall training accuracy of 97.7 %. The lowest training accuracy with the QDA configuration was only 85.4 %. Compared to baseline models, the hybrid models yielded improvements of up to 12.3 % in classification accuracy and reduced misclassification rates. The research demonstrates several contributions including a novel intepretive hybrid framework, combining advanced machine learning classification with bio-inspired optimisation, demonstrates superior classification and computational efficiency, assesses the possibility of the ML model model in a cloud-based environment for smart grid integration, assesses the ability to train and develop deployments models in the future, and can be scaled to a broader industrial context. This integrated strategy offers a practical, high-performance solution to real-time energy management, supporting the transition toward more intelligent and sustainable industrial operations.