EAF Steelmaking: AI applications for estimating energy consumption|Ciclo EAF: applicazioni dell’IA per stimare i consumi energetici

Steel production through the electric cycle suffers from low diversification of energy supply sources and, with a view to efficiency and cost reduction, needs to be optimized on several aspects, among which production planning. Optimizing production costs means, for example, leveling out the electricity consumption along the day, whose prices may fluctuate in the short/medium term (variating on hour/daily basis) or if peak consumption exceeds a certain threshold. Production planning can, therefore, benefit from models predicting the energy consumed as a function of the quality of steel to be produced and the specific production recipe. These models can be effectively used by intelligent scheduling optimization systems. This paper presents a set of models based on neural networks that can predict with good accuracy the power consumption in the electric arc and ladle furnaces as a function of main production information. The models were trained and validated through real production and process data from a plant in Croatia with encouraging results.