Improved Reservoir Characterization through Evolutionary Computation

Abstract

Improved Reservoir Characterization through Evolutionary Computation C.E. Romero; C.E. Romero Imperial College Search for other works by this author on: This Site Google Scholar J.N. Carter; J.N. Carter Imperial College Search for other works by this author on: This Site Google Scholar R.W. Zimmerman; R.W. Zimmerman Imperial College Search for other works by this author on: This Site Google Scholar A.C. Gringarten A.C. Gringarten Imperial College Search for other works by this author on: This Site Google Scholar Paper presented at the SPE Annual Technical Conference and Exhibition, Dallas, Texas, October 2000. Paper Number: SPE-62942-MS https://doi.org/10.2118/62942-MS Published: October 01 2000 Cite View This Citation Add to Citation Manager Share Icon Share Twitter LinkedIn Get Permissions Search Site Citation Romero, C.E., Carter, J.N., Zimmerman, R.W., and A.C. Gringarten. "Improved Reservoir Characterization through Evolutionary Computation." Paper presented at the SPE Annual Technical Conference and Exhibition, Dallas, Texas, October 2000. doi: https://doi.org/10.2118/62942-MS Download citation file: Ris (Zotero) Reference Manager EasyBib Bookends Mendeley Papers EndNote RefWorks BibTex Search 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 AbstractComputer-aided history matching techniques are increasingly playing a role in the reservoir characterisation process. In general, these techniques provide a single solution that matches the observed data. It has been widely recognised that a solution found to a history matching problem is not unique, but alternative solutions are usually not sought, due to time and cost restrictions.This paper describes how evolutionary computation can be used to obtain multiple possible solutions without repeating the history matching process. Once a geological and structural model has been defined, Genetic Algorithms are used to carry out hydrocarbon reservoir characterisation by conditioning the reservoir simulation model to production data (history matching). The proposed approach uses ideas from Darwinian evolutionary theory, combining genetic algorithms and geostatistical modelling to generate multiple reservoir descriptions.The algorithms return a suite of solutions (corresponding to different realisations of the reservoir model) from which the reservoir team can select a representative group for further analysis. The solutions give history matches with a quality comparable to what might be achieved by hand. The method is relatively easy to implement, robust with respect to lost or corrupted solutions, cheap in terms of computational costs, and can be easily parallelised.IntroductionThe use of computer-aided techniques to assist in the reservoir characterisation process is becoming standard in the petroleum industry. As computers become faster and more computing power is made affordable, companies dedicate more resources to data analysis. The techniques aim at the incorporation into the reservoir models of all data available so that more realistic models can be generated for improved prediction capabilities.The process makes use of measurements made on the field to restrict the range of values that the parameters might take. The measurements used are wide-ranging and include seismic data, data from geological analogues, core and log data from wells, well test data, and production data. The result of the process is a set of flow simulation models that, to a greater or lesser extent, agree with the measurements made upon the reservoir. There are very many papers in the literature that describe this process, including Tan and Kalogerakis,1 Oliver et al.,2 Wu et al.,3 and Floris et al.4To properly characterise the flow simulation model requires that every part of the reservoir be adequately described. Direct inversion of the complete model is not a practical proposition, so it is necessary to introduce a coarse grid of pilot points (or control points) and some interpolation method. There are numerous papers that discuss the choice of pilot points and interpolation methods.5,6 Having selected these, some optimisation technique is needed to match the numerical results to the measurements. Most attempts to automate this process have used gradient-type methods.7Evolutionary systems have been studied for possible applications to optimisation problems. Evolutionary computation techniques such as Genetic Algorithms (GAs), Evolutionary Strategies (ESs), Classifier Systems (CSs), and Evolutionary Programming (EP), draw their inspiration from nature. Particularly, Genetic Algorithms (GAs) were invented by John Holland8 as an abstraction of biological evolution, drawing on ideas from natural evolution and genetics for the design and implementation of robust adaptive systems. Over the last 20 years, GAs have received much attention because of their potential as optimisation techniques for complex functions.9Genetic Algorithms have been shown to be effective over a wide range of problems. Examples include nuclear reactor management,10 gas pipeline operation,11 process control,12 and aircraft design.13 GAs have also been applied to study inverse problems in fields such as airborne pollution,14 geophysics,15 crop yield estimates,16 groundwater flow,17 and electromagnetics.18 Keywords: mutation, zimmerman, application, operator, history, machine learning, carter, objective function, reservoir simulation, reservoir characterization Subjects: Reservoir Characterization, Reservoir Simulation, Information Management and Systems, Geologic modeling, Artificial intelligence This content is only available via PDF. 2000. Society of Petroleum Engineers You can access this article if you purchase or spend a download.