Optimisation of a fermentation process for butanol production by particle swarm optimisation (PSO)

Abstract

BACKGROUND: The performance of three particle swarm optimisation (PSO) algorithms was assessed in relation to their capability to optimise an alternative fermentation process for the production of biobutanol. The process consists of three interconnected units: fermentor, cell retention system and vacuum flash vessel (responsible for the continuous recovery of butanol from the broth). The dynamic behaviour of the process was described by a non-linear mathematical model. Four constrained optimisation problems were formulated concerning the operation and design of flash fermentation: (1) maximisation of butanol productivity; (2) maximisation of substrate conversion; (3) and (4) adjustment of operating conditions in the face of problems of fluctuations in the quality of the agricultural raw material and changes in the kinetics of the microorganisms.RESULTS: The design and operation of the flash fermentation process based on the optimisation of productivity, instead of substrate conversion, resulted in a smaller fermentor and provided satisfactory values of operating conditions able to overcome problems of variations in the glucose concentration in the raw material and changes in kinetics.CONCLUSIONS: The differences among the PSO algorithms, i.e. the velocity equation and parameters values, had significant effects on the optimisation, the best results being obtained with the original velocity equation with the inertia weight decreasing linearly with each iteration. The PSO algorithms obtained solutions that obeyed constraints, demonstrating that a constraint handling method originally developed for genetic algorithms can be applied successfully to PSO algorithms. (C) 2010 Society of Chemical Industry