NONLINEAR PREDICTIVE CONTROL OF AN ELECTRIC ARC FURNACE

Electric arc furnaces are widely used in the steel industry for melting scrap. The highly energy intensive nature of these operations, coupled with their complexity and typically low level of automation, make them prime candidates for optimization. However, application of open-loop optimal control policies may result in suboptimal or infeasible operation in the presence of model uncertainty and unmeasured disturbances. This paper describes a strategy that combines a comprehensive dynamic model, rigorous optimization, and feedback within a nonlinear model predictive control framework. Key features are the direct use of an economic performance objective, a shrinking prediction horizon, and use of a nonlinear model. The components of the algorithm are described, and its performance illustrated through application to a case study.