Prediction of power consumption from real process data of an industrial wood chip refining plant

Improving the efficiency of production processes is fundamental to minimize their environmental impact and energy consumption. The pulp and paper industry is a highly energy-intensive one that urgently needs to become more efficient, especially in the refining phase. In this framework, the model identification of a wood chips refining process operating in closed loop, pertaining to the production of Medium Density Fiberboard (MDF), is presented here, aimed to provide a long-term prediction of power consumption. We perform the identification via multi-batch Simulation Error Minimization (SEM), employing real process data collected on a large-scale MDF production plant during operation, without using sophisticated models or adhoc experimental sessions. The derived model obtains extremely high accuracy on a validation dataset while being simple enough to be used efficiently for production planning optimization. Moreover, it allows us to derive further models to predict the wear of the refiner disc, to be accounted for in a plant optimization procedure as well.