The use of model-based control approaches can improve the operation of many complex
processes significantly but necessitates the availability of accurate dynamic process models,
and the development of such models is time-consuming which is a major bottleneck for the
application of advanced control. In this paper, a hybrid modeling approach is proposed,
where a simple mechanistic model is augmented by a machine-learning model to
compensate the plant-model mismatch based on observed data. As the data-based models …