Neural Network Model Predictive Control Applied to a Fed-batch Sugar Crystallization

This paper is focused on a comprehensive study of neural network (NN) model based predictive control (MPC), as an operation strategy for a fed-batch sugar crystallizer. The process is divided into four subsequent control loops and for each of them an individual NN-based MPC is designed. The operation is tested for a number of scenarios and is compared with alternative (linear and batch nonlinear MPC) control solutions. The results demonstrate that the NN-MPC is a promising alternative of the traditionally applied linear controllers when the process is strongly nonlinear and input-output data is the only process information available.