Application of Model-Predictive Control to Multi-Hearth Nickel Reduction Roasters

Abstract This paper describes the application of multivariable model-predictive techniques to the control and optimization of a set of nickel reduction roasters. Using empirical dynamic models derived from process response testing, the model-predictive controller calculates real-time control actions that accurately compensate for the time delays and strong interactions that characterize this type of unit. Compared with the previous operational mode, the deployment of this technology has resulted in a reduction of up to 30% in the variability of roaster conditions, an increase of 2% in energy efficiency and an increase of 1.2% in the recovery of nickel from the laterite ore feed stream.