Fast and Accurate Reservoir Modeling: Genetic Algorithm versus DIRECT Method

Authors: Mohsen Ebrahimi, Milad M. Rabieh

Abstract:

In this paper, two very different optimization algorithms, Genetic and DIRECT algorithms, are used to history match a bottomhole pressure response for a reservoir with wellbore storage and skin with the best possible analytical model. No initial guesses are available for reservoir parameters. The results show that the matching process is much faster and more accurate for DIRECT method in comparison with Genetic algorithm. It is furthermore concluded that the DIRECT algorithm does not need any initial guesses, whereas Genetic algorithm needs to be tuned according to initial guesses.

Keywords: DIRECT algorithm, Genetic algorithm, Analytical modeling, History match

Digital Object Identifier (DOI): doi.org/10.5281/zenodo.1057525

References:

[1] Horne, R. N. "Modern well test analysis: acomputer aided approach". Palo Alto :Stanford University, 1990.
[2] Hobgood, D. A. "Computerized derivative type curve matching based on derivative curve height". M.S. thesis. Texas Tech University, 1989.
[3] Jones, D. R., Perttunen, C. D., and Stuckman, B. E. "Lipschitzian optimization without the lipschitz constant". Journal of Optimization Theory and Applications. Vol 79. 157-181, 1993.
[4] Chiter L., and Benseghir A. "Another face of DIRECT". Informatica Journal. Vol. 54. 68-74, 2009.