Reliable optimal control of a batch polymerisation reactor based on neural network model with model prediction confidence bounds

Abstract Bootstrap aggregated neural networks are used to model a batch polymerisation reactor from limited batches of process operational data. The model can predict the number average molecular weight, weight average molecular weight, and monomer conversion from the batch recipe and reactor temperature profile. A further advantage of bootstrap aggregated neural network models is that model prediction confidence bounds can be obtained. One of the most important issues of empirical model based batch process optimal control is that the calculated optimal control profile can degrade very significantly when applied to the actual process due to model plant mismatches. To address this issue, this paper presents a new optimal control method where the optimisation objective function includes an additional term to penalise wide model prediction confidence bound at the end-point of a batch. By such a means, the calculated optimal control profile is more reliable and, when being applied to the actual process, the degradation in control performance can be limited.