Initiating a Digital Oilfield Operating System on Brage field
- Vedad Hadziavdic (Wintershall Norge) | Luc Sandjivy (Seisquare) | Arben Shtuka (Seisquare)
- Document ID
- Society of Exploration Geophysicists
- SEG/AAPG/EAGE/SPE Research and Development Petroleum Conference and Exhibition, 9-10 May, Abu Dhabi, UAE
- Publication Date
- Document Type
- Conference Paper
- 114 - 117
- 2018. Society of Exploration Geophysicists
- workflow, drilling, horizontal wells
- 1 in the last 30 days
- 30 since 2007
- Show more detail
This document is an expanded abstract.
Uncertainty quantification has become an increasingly important request in the decision-making process for targeting infill drilling. Serious weaknesses in current reservoir models can be directly attributed to the lack of proper uncertainty handling. In a mature field as the Brage oil field in the northern part of the North Sea, east of the Oseberg field, where there are many long horizontal wells, uncertainty in well markers and well paths have a tremendous impact on the structural uncertainty (+/-10m). These variations on their own can either create or kill infill drilling targets. A stochastic workflow for dealing with uncertainty on horizontal well trajectories has been derived for a consistent updating of velocity and depth models. Such an automated stochastic workflow is an example of a Digital oilfield operation system (DOOS).
The coming age of Digital Oil Field operations in oil companies implies the conversion of today knowledge-based data processing and modelling workflows into automated intelligent ones, representing digital operating systems or DOOS.
DOOS are designed for optimizing performance and turnaround time of geophysical processing and modelling workflows, thus enabling real time and safer E&P decision making. They make use of machine learning algorithms specific to Earth Data, that are regionalized (they have coordinates) and uncertain. As statistics are key to Artificial Intelligence of Big Data, geo statistics (and more precisely probability models) are the corner stone of machine learning algorithms for processing Big Geo Data.
Geostatistics for optimizing reservoir operations
The main characteristic of an oil reservoir is that it has not been manufactured and is an a priori unknown environment to human activity, generating natural uncertainty and risk when operating it. Oil reservoir is a natural resource that is out of direct reach and there is no such thing as an exact representation of the subsurface that would allow for making 100% confidence or 0 risk operational decisions.
Handling this natural “uncertainty” and making best possible operation decisions have been the driver of the research and developments of Geostatistics as described by Georges Matheron in the “Theory of Regionalized Variables” (1), and “Estimating and choosing” (2).
|File Size||1 MB||Number of Pages||4|