Guide Objective Assisted Particle Swarm Optimization and Its Application to History Matching

Alan P. Reynolds¹, Asaad Abdollahzadeh², David W. Corne¹, Mike Christie², Brian Davies³, and Glyn Williams³

¹School of Mathematical and Computer Sciences, Heriot-Watt University, Edinburgh, Scotland EH14 4AS

²Institute of Petroleum Engineering, Heriot-Watt University, Scotland

³BP, UK

Abstract. As is typical of metaheuristic optimization algorithms, particle swarm optimization is guided solely by the objective function. However, experience with separable and roughly separable problems suggests that, for subsets of the decision variables, the use of alternative ‘guide objectives’ may result in improved performance. This paper describes how, through the use of such guide objectives, simple problem domain knowledge may be incorporated into particle swarm optimization and illustrates how such an approach can be applied to both academic optimization problems and a real-world optimization problem from the domain of petroleum engineering.

LNCS 7492, p. 195 ff.

Full article in PDF | BibTeX

lncs@springer.com
© Springer-Verlag Berlin Heidelberg 2012