Reservoir Characterization Using Fuzzy Kriging and Deep Learning Neural Networks
- M. M. Korjani (University of Southern California) | A. S. Popa (Chevron Corporation) | E. Grijalva (Chevron Corporation) | S. Cassidy (Chevron Corporation) | I. Ershaghi (University of Southern California)
- Document ID
- Society of Petroleum Engineers
- SPE Annual Technical Conference and Exhibition, 26-28 September, Dubai, UAE
- Publication Date
- Document Type
- Conference Paper
- 2016. Society of Petroleum Engineers
- 5.1 Reservoir Characterisation, 6.1.5 Human Resources, Competence and Training, 7.6.6 Artificial Intelligence, 5.1.5 Geologic Modeling, 6 Health, Safety, Security, Environment and Social Responsibility, 5 Reservoir Desciption & Dynamics, 6.1 HSSE & Social Responsibility Management
- reservoir characterization, fuzzy kriging, deep learning neural neural networks
- 50 in the last 30 days
- 747 since 2007
- Show more detail
- View rights & permissions
|SPE Member Price:||USD 9.50|
|SPE Non-Member Price:||USD 28.00|
The sandy-clayed heavy oil hydrocarbon reservoirs of the Miocene formations in San Joaquin Valley, California, present relatively high complex geologic settings. These reservoirs consist in general of sandstone, silts and shale bodies, and present anywhere from six to eight degrees of inclination. Geostatistical tools were used to characterize the reservoir and develop a full earth model for the field. The data used was mainly based on the suite of resistivity logs (shallow, medium and deep), which seems to best represent the formations, while a classic kriging technique was used for modeling.
In the present research, we introduce a new approach for characterization of these heavy-oil reservoirs by using fuzzy kriging and deep learning neural networks. The new model better captures the uncertainty associated with the characteristics of the reservoirs. Once successfully trained, the system was applied to generate a 3D earth model and estimate reservoir properties at any point in the field. The new earth model successfully improves the resolution and realization of the subsurface model, as compared to the existing one developed using classic kriging. Integrating fuzzy kriging solves the limitation of smoothness of kriging estimation and reproduction of extreme values. Additional capability includes generation of synthetic logs for wells anywhere in the reservoir, with a significant application for infill drilling.
The methodology presented in this research proved to be very successful as developed for modeling alluvial deposits such as the heavy oil reservoirs in San Joaquin Valley, California. Further research should be considered to address other geologies where laterally formations vary in characteristics or where pinch outs or faulting occur.
|File Size||4 MB||Number of Pages||15|
A. Aggarwal, and S. Agarwal, "ANN Powered Virtual Well Testing," Offshore Technology Conference, doi:10.2118/174871-MS, 2014.
S. P. Ketineni, T. Ertekin, K. Anbarci, and T. Sneed, "Structuring an Integrative Approach for Field Development Planning Using Artificial Intelligence and its Application to an Offshore Oilfield." Society of Petroleum Engineers. doi:10.2118/174871-MS, September 28, 2015.
M. Korjani, A. Popa, E. Grijalva, S. Cassidy, and I. Ershaghi, "A New Approach to Reservoir Characterization Using Deep Learning Neural Networks," Society of Petroleum Engineers, doi:10.2118/180359-MS, 2016.