Dynamic on‐line optimization of a bioreactor

An algorithm was developed which uses recursive least squares to identify a dynamic, discrete time model of a poorly defined system and uses both the dynamic and static portions of the identified model for on‐line optimization. To test this new algorithm, a model of an continuous biochemical reactor was used as the “process.” The objective, here, was to maximize ethanol production from the reactor by manipulating the feed rate to the reactor. The new algorithm, which uses dynamic information, was found to be superior to previously published algorithms which use only the steady‐state portion of the identified model.