A comparative study of differential evolution algorithms for parameter fitting procedures

Carlos E. Torres-Cerna, Alma Y. Alanis, Ignacio Poblete-Castro, Marta Bermejo-Jambrina, Esteban A. Hernandez-Vargas

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

7 Citations (Scopus)

Abstract

Parameter fitting consists on the estimation of model parameters using experimental data from the studied process, which can be considered as a nonlinear optimization problem. In this sense, evolutionary computation has shown its great capability to solve multimodal nonlinear optimization problems. This paper compares different variants of the Differential Evolution (DE) algorithm to minimize the residual sum of squares between the outcome of the mathematical model and experimental data. To compare the different variants of the DE algorithm, a biopolymer production model is considered. Simulations results suggest a trend for the best fit using the DE/best/ variants. However, the DE/rand/ variant provides more stable results respect to the average and standard deviation of different trials. Finally, the biopolymer production problem is discussed.

Original languageEnglish
Title of host publication2016 IEEE Congress on Evolutionary Computation, CEC 2016
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages4662-4666
Number of pages5
ISBN (Electronic)9781509006229
DOIs
Publication statusPublished - 14 Nov 2016
Event2016 IEEE Congress on Evolutionary Computation, CEC 2016 - Vancouver, Canada
Duration: 24 Jul 201629 Jul 2016

Other

Other2016 IEEE Congress on Evolutionary Computation, CEC 2016
Country/TerritoryCanada
CityVancouver
Period24/07/1629/07/16

Keywords

  • Bioprocesses
  • Differential Evolution Algorithm
  • Parameter Estimation

ASJC Scopus subject areas

  • Artificial Intelligence
  • Modelling and Simulation
  • Computer Science Applications
  • Control and Optimization

Fingerprint

Dive into the research topics of 'A comparative study of differential evolution algorithms for parameter fitting procedures'. Together they form a unique fingerprint.

Cite this