When in pursuit of reproducible operation of batch processes, a lack of reliable models is often limiting. This contribution presents a methodology for rapid acquirement of discrete-time state space model representations of batch processes based on historical operation data. These state space models are parsimoniously parameterized as a set of local, interdependent models. The model identification in dependence of the grid resolution and the indepen-dent discretization variable are investigated. The present contribution furthermore presents how the asymptotic convergence of Iterative Learning Control is combined with the closed-loop performance of Model Predictive Control to form a robust and asymptotically stable optimal controller for reliable and reproducible operation of batch processes.

The modelling methodology produces both a Linear Time-Invariant state space model representation for inter-batch prediction and a Linear Time-Varying state space model representation for intra-batch prediction. With these two state space model representations, reproducible operation of batch processes can be pursued with model-based tools such as linear tracking controllers and linear observers. The modeling methodology approximates the non-stationary and nonlinear behavior of batch processes with a set of local but interdependent linear regression models, and applies Tikhonov Regularization to estimate the parameters this model set. The local models are parameterized as AutoRegressive Moving Average models with eXogenous inputs (ARMAX). The present contribution furthermore presents the control methodology learning Model Predictive Control for control of repeated operation of stochastic Linear Time-Varying systems with finite time horizons. For this methodology tuning requirements for guaranteed convergence and hence closed-loop stability are presented.

The above stated methodologies have been implemented as Matlab toolboxes and in the present contribution their applicability is demonstrated on a simulated fed-batch protein fermenter.