Abstract
This study addresses the critical challenge of optimizing energy consumption in direct reduction iron (DRI) units, a vital component of the steel industry. By utilizing operational data from a DRI unit, this research identifies and analyzes the key factors influencing energy consumption through three advanced modeling approaches: RSM, MLP, and RBF neural networks. The RSM model demonstrated strong predictive capability, achieving a coefficient of determination (R(2)) of 0.9879. However, the ANN models surpassed the RSM model in terms of accuracy. Among the ANN models, the MLP model exhibited the highest performance, with an R(2) of 0.99601 and a MSE of 0.00037, while the RBF model achieved an R(2) of 0.99336 and an MSE of 0.00062. Leveraging the optimized MLP model, this study identifies optimal operational conditions that minimize energy consumption. The findings indicate that strategic adjustments to parameters such as cooling gas flow and main burner flow can lead to substantial energy savings. Specifically, the model predicts daily savings of 60,000 kWh and annual savings of approximately 21,900,000 kWh, reflecting a 10.34% improvement in energy efficiency. These results underscore the potential of machine learning to enhance operational efficiency and sustainability in energy intensive industries, providing a robust framework for data-driven energy management strategies.