Modeling and validation of a clinker production process model that captures the coupled internal dynamics

Cement clinker production is an energy-intensive process that accounts for a substantial share of global industrial energy consumption. Model predictive control (MPC) is commonly used to regulate this process, requiring a compact control-oriented model that describes the input–output relationships of the production system. A prevalent method for obtaining such a model is to identify direct input–output relationships. However, this approach often overlooks the coupled dynamics of internal process variables, which can limit prediction accuracy. This study introduces a methodology for mapping these dynamically coupled internal variables by modeling the production process as a network of interconnected chambers. A discrete-linear model that captures these couplings is then developed using data collected from an operational clinker plant. The proposed model is evaluated against widely-used linear modeling approaches, with a focus on multi-step-ahead prediction performance, a metric often neglected in the literature. Eight key variables were chosen as targets for prediction, and the proposed model consistently outperformed the alternatives in predicting their variations, particularly when the prediction horizon exceeded five sampling intervals. An MPC implementation of the proposed model is provided for illustrating its potential use.