A Stochastic Optimization Algorithm for Automatic History Matching
Guohua Gao,Gaoming Li,A.C. Reynolds
DOI: https://doi.org/10.2523/90065-ms
2004-01-01
Abstract:A Stochastic Optimization Algorithm for Automatic History Matching Guohua Gao; Guohua Gao University of Tulsa Search for other works by this author on: This Site Google Scholar Gaoming Li; Gaoming Li University of Tulsa Search for other works by this author on: This Site Google Scholar A.C. Reynolds A.C. Reynolds University of Tulsa Search for other works by this author on: This Site Google Scholar Paper presented at the SPE Annual Technical Conference and Exhibition, Houston, Texas, September 2004. Paper Number: SPE-90065-MS https://doi.org/10.2118/90065-MS Published: September 26 2004 Cite View This Citation Add to Citation Manager Share Icon Share Twitter LinkedIn Get Permissions Search Site Citation Gao, Guohua, Li, Gaoming, and A.C. Reynolds. "A Stochastic Optimization Algorithm for Automatic History Matching." Paper presented at the SPE Annual Technical Conference and Exhibition, Houston, Texas, September 2004. doi: https://doi.org/10.2118/90065-MS Download citation file: Ris (Zotero) Reference Manager EasyBib Bookends Mendeley Papers EndNote RefWorks BibTex Search Dropdown Menu nav search search input Search input auto suggest search filter All ContentAll ProceedingsSociety of Petroleum Engineers (SPE)SPE Annual Technical Conference and Exhibition Search Advanced Search AbstractFor large-scale history matching problems, optimization algorithms which require only the gradient of the objective function and avoid explicit computation of the Hessian appear to be the best approach. Unfortunately, such algorithms have not been extensively used in practice because computation of the gradient of the objective function by the adjoint method requires explicit knowledge of the simulator numerics and expertise in simulation development. Here we apply the simultaneous perturbation stochastic approximation (SPSA) method to history match multiphase flow production data. SPSA, which has recently attracted considerable international attention in a variety of disciplines, can be easily combined with any reservoir simulator to do automatic history matching. The SPSA method uses stochastic simultaneous perturbation of all parameters to generate a down hill search direction at each iteration. The theoretical basis for this probabilistic perturbation is that the expectation of the search direction generated is the steepest descent direction.We present modifications for improvement in the convergence behavior of the SPSA algorithm for history matching problems and compare its performance to the steepest descent, gradual deformation and LBFGS algorithm for history matching examples. Although the convergence properties of the SPSA algorithm are not nearly as good as our most recent implementation of a quasi- Newton method (LBFGS), the SPSA algorithm is not simulator specific and it requires only a few hours of work to combine SPSA with any commercial reservoir simulator to do automatic history matching.To the best of our knowledge, this is the first introduction of SPSA into the history matching literature. Thus, we make considerable effort to put it in a proper context.IntroductionA resurgence of interest in automatic history matching of production data, especially in the context of data integration and uncertainty analysis, has occurred during the last decade, see Bissell1, Oliver2, Chu et al.3, Omre et al.4, Oliver et al.5, He et al.6, Landa and Horne7, Wu et al.8 Reynolds et al.9, Glimm and Sharp10, Vasco et al.11, Barker et al.12, Hegstad and Omre13, Gosselin et al.14, Zhang et al.15, Aanonsen et al.16, Aanonsen et al.17, Li et al.18, Zhang et al.19 This work is a natural outgrowth of much earlier work20,21,22, but has been facilitated by the rapid increase in computer power, the widespread application of geostatistics and interest in data integration.The history matching problems of interest here is equivalent to the minimization of an objective function which includes data mismatch terms weighted by an inverse covariance matrix for measurement errors. Virtually all of the papers cited above use a gradient based algorithm to match data, but because of the unavailability of adjoint software for computing gradients or the impracticality of computing all sensitivity coefficients, most works limit the number of model parameters estimated or simulated. A major exception to this is the work of Zhang and Reynolds23 who use an adjoint method based implementation of Li et al.24 to compute the gradient of the objective function and implement a limited memory Broyden- Goldfarb-Shanno (LBFGS) algorithm to minimize the objective function. Recently Gao and Reynolds25 made some further improvements in the LBFGS to further enhance its robustness and computational efficiency. Keywords: upstream oil & gas, machine learning, bayesian inference, objective function, gradient, spsa, optimization problem, nonlinear example, iteration, spsa algorithm Subjects: Reservoir Simulation, Evaluation of uncertainties, History matching This content is only available via PDF. 2004. Society of Petroleum Engineers You can access this article if you purchase or spend a download.