Online estimation and adaptive control of penicillin fermentation

The paper describes an investigation into the application of state estimation and adaptive control to fed-batch fermentation for penicillin production. The work forms part of an industrial collaborative project, the aim of which is the optimising control of large fed-batch fermenters. Estimates of biomass are made using an extended Kalman filter from on-line measurements of carbon dioxide production rate and fermentation volume. The estimated biomass is controlled to a reference trajectory by an adaptive generalised predictive controller manipulating the sugar feed rate. A comparison of performance is made with control by conventional proportional plus integral (PI) techniques. The results presented are obtained from simulation studies while validation studies are carried out on both a 30 1 pilot plant fermenter and the industrial plant.