Technology Focus: History Matching and Forecasting (April 2014)
- Alexandre Emerick (Petrobras Research Center)
- Document ID
- Society of Petroleum Engineers
- Journal of Petroleum Technology
- Publication Date
- April 2014
- Document Type
- Journal Paper
- 118 - 118
- 2014. Copyright is retained by the author. This document is distributed by SPE with the permission of the author. Contact the author for permission to use material from this document.
- 1 in the last 30 days
- 225 since 2007
- Show more detail
- View rights & permissions
|SPE Member Price:||Free|
|SPE Non-Member Price:||USD 2.00|
In the past decade, there has been a great deal of research and progress in the development of computational methods to assist reservoir engineers in the arduous task of history matching their models. Developments in computer hardware and software and the use of geostatistics, optimization, and Monte Carlo methods are among the reasons for such progress. However, the requirements for a good history match also have been increasing. If, in the past, we were satisfied with a good data fit, this is definitely not the case anymore. Modern history matching is a much more comprehensive discipline. It entails geological modeling, geostatistics, reservoir simulation, scale issues, data analysis, deep understanding of reservoir mechanisms, interdisciplinary approaches, optimization methods, statistics, and inverse-problems theory.
Nevertheless, we have to recognize that history matching in itself is not the goal. The goal is to generate models for production forecasting aiding the decision-making process involved in the development and management of petroleum reservoirs. One can definitely argue that a good history match does not ensure a reliable forecast. On the other hand, forecasting from models without a reasonable history match is most certainly temerarious (and probably meaningless). In fact, forecasting is an inherently uncertain process. In this sense, the goal of history matching is to mitigate this uncertainty. However, we have to keep in mind that uncertainty will never be removed completely. First, models are only representations of the actual physical phenomena; there will always be approximation errors. Second, data are insufficient and have measurement errors; hence, it is impossible to fully determine all unknown model parameters. Finally, uncertainty quantification has a high degree of subjectivity. Hence, bad forecasting often will be related to wrong assumptions and misinterpretations.
Among current research trends in history matching and forecasting, it is worth mentioning uncertainty quantification, methods based on ensembles, assimilation of time-lapse seismic data, and history matching of models with complex physics (e.g., fractured reservoirs and thermal recovery processes). However, one particular topic has caught the attention of many researchers: the development of history matching workflows that fully integrate the geological modeling in the loop. This is often called the “big-loop approach.” Despite the developments in methods, software, and computers, human intervention and good engineering judgment will continue to be vital. Therefore, it is important to keep up to date in the literature. The papers summarized in this feature and those indicated in the additional-reading list are good examples of recent developments in, and field applications of, history matching and forecasting. JPT
Recommended additional reading at OnePetro: www.onepetro.org.
SPE 163638 Estimation of Mutual Information and Conditional Entropy for Surveillance Optimization by Duc H. Le, The University of Tulsa, et al.
SPE 167150 Prior Information Enhances Uncertainty Quantification in Shale- Gas Decline-Curve Forecasts by Raul Gonzalez, Texas A&M University, et al.
SPE 163604 Preventing Ensemble Collapse and Honoring Multipoint Geostatistics With the Subspace EnKF/ EnS and Kernel PCA Parameterization by Pallav Sarma, Chevron ETC, et al.
|File Size||77 KB||Number of Pages||1|