Model-based control of a pilot scale fermentation process using a predictive control strategy

Industrial fermentation systems are often operated in fed-batch mode [1], however this mode of operation can make planning and scheduling challenging due to batch-to-batch variation. There is natural variability in the length of the batch phase due to biological variations, in addition to batch-to-batch variations in initial process conditions from raw materials [2]. These affect the reproducibility of the batch, and also affect the final batch fill due to the feed addition. This therefore leads to variation in the total product achieved from each batch. It is desirable to always achieve full vessel capacity [3] to maximize the productive volume of each tank in every fermentation run, within a given process time.