Genetic Algorithms for a Parameter Estimation of a Fermentation Process Model: A Comparison

In this paper the problem of a parameter estimation using genetic algorithms is examined. A case study considering the estimation of 6 parameters of a nonlinear dynamic model of E. coli fermentation is presented as a test problem. The parameter estimation problem is stated as a nonlinear programming...

Full description

Saved in:
Bibliographic Details
Published in:Bioautomation Vol. 3; no. 1; pp. 19 - 28
Main Author: Roeva, Olympia
Format: Journal Article
Language:English
Published: Sophia Bulgarska Akademiya na Naukite / Bulgarian Academy of Sciences 01.12.2005
Academic Publishing House
Subjects:
Fermentation
Fermentation processes
Genetic algorithms
Nonlinear programming
Optimization
Parameter estimation
Parameter identification
ISSN:1314-1902, 1313-261X, 1314-2321, 1312-451X
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In this paper the problem of a parameter estimation using genetic algorithms is examined. A case study considering the estimation of 6 parameters of a nonlinear dynamic model of E. coli fermentation is presented as a test problem. The parameter estimation problem is stated as a nonlinear programming problem subject to nonlinear differential-algebraic constraints. This problem is known to be frequently ill-conditioned and multimodal. Thus, traditional (gradient-based) local optimization methods fail to arrive satisfied solutions. To overcome their limitations, the use of different genetic algorithms as stochastic global optimization methods is explored. These algorithms are proved to be very suitable for the optimization of highly non-linear problems with many variables. Genetic algorithms can guarantee global optimality and robustness. These facts make them advantageous in use for parameter identification of fermentation models. A comparison between simple, modified and multi-population genetic algorithms is presented. The best result is obtained using the modified genetic algorithm. The considered algorithms converged very closely to the cost value but the modified algorithm is in times faster than other two.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1314-1902
1313-261X
1314-2321
1312-451X