Neural Network-based modeling methodologies for energy transformation equipment in integrated steelworks processes

Abstract

Abstract The paper proposes a methodology for modeling of energy transformation equipment which are commonly found in integrated steelworks, mainly focusing on steam production in the Basic Oxygen Furnace and auxiliary boilers, the electric power production in off-gas expansion turbines and some relevant steam and electricity consumers. The modeling approach is based on standard neural networks and Echo State Networks (ESN) for forecasting the variables of interest. All the models are intended as processes predictors to be used in a hierarchical control strategy based on multi-period and multi-objective optimization techniques and model predictive control. The overall target is the optimization of the re-use of off-gas produced in integrated steelworks by minimizing costs and maximizing revenues. Training and validation of models have been carried out by exploiting real historical data provided by steelmaking companies and have been successful tested.