Nosiheptide, a sulfur-containing
peptide antibiotic obtained through
fermentation, is a perfect feed additive, but its yield in industry
is not high. Process optimization is a good way to increase nosiheptide
yield, maintaining the optimum operating conditions of the fermentation
process, while optimization of the process requires a sufficiently
accurate and robust process model. In this paper, the mechanism model
for nosiheptide fed-batch fermentation is first established. Then,
in order to improve performance of the mechanism model, a hybrid model
is built using least-squares support vector machines to compensate
the errors between the mechanism model and the process. The hybrid
model not only overcomes pure black-box model’s shortcoming
that it often has poor generalization ability but improves the mechanism
model’s accuracy. A yield optimization model of nosiheptide
fed-batch fermentation process is then established based on the hybrid
model. An improved particle swarm optimization algorithm is used to
solve the optimization model, greatly improving the end nosiheptide
production, which also proves the validity of the proposed particle
swarm optimization algorithm.