Semi-realtime optimization and control of a fed-batch fermentation system

This article proposes a method to semi-realtime optimize and control a fed-batch fermentation system using hybrid neural networks (HNN) and genetic algorithms (GAs). The fermentation system is to cultivate Bacillus thuringiensis (Bt) for thuringiensin production. Thuringiensin, which is a bioinsecticide, is one of the major exotoxins of Bacillus thuringiensis. The cultivation system is modeled into a hybrid neural network model, which serves as the search domain of the genetic algorithm to determining the optimal feeding rate. Semi- realtime optimization is carried out using the HNN model and the measured state variables to re-optimize the system every 1 h. The results show a great increase in production of thuringiensin.