Commenting the recent hike in the oil price, Total’s VP of E&P Christophe de Margerie said, ‘We all underestimated India and China. We were wrong!’ There has been too much focus on western stock levels and little focus on international growth. Worldwide, BOE/capita/year is increasing ahead of population growth. In 2000, consumption was 9 bn. tonnes, in 2030 it will be 15 bn. tonnes. But the reserves are there, Saudi Arabia still claims 260 bn. bbl. But forecasting future demand is tricky. If Asia consumes 0.7 units/head, the EU consumes 3.2 and N. America 8.3! Since Asian growth is racing ahead the big question is, ‘which consumption model is Asia going to align itself with?’
Schlumberger president Andrew Gould also believes in a durable price rise. For the first time, the problem is supplying a sustained demand growth. Spend is shifting away from new production and towards ‘decline rate’. ‘We know little about decline rate and have no clear pattern of what it is.’ There has been a remarkable ‘decline’ too in finding and developing costs, thanks to new technology like smart wells and intelligent completion. Geophysics used to be a language spoken only by explorationists, now 3D goes beyond exploration and is even used to optimize topside design. Knowledge management is an issue with the large scale retirement of workforce. But for Gould, ‘production decline’ is where it’s at, we should ‘manage decline through technology.’
For Pete Carragher (BP), tomorrow’s fields will be deeper, hotter and harder to image. Access to and timing of discoveries will be ‘unpredictable’. IHS Energy data shows steady decline in field size since 1970, with poor replacement for oil, although the situation is better for gas. Average depth to top pay is rising. Carragher cited a recent survey of geoscience skills in the oil industry which showed a growing need for non technical workers and IT. According to the survey, the oil industry does not need sedimentologists, structural geologists, core and many other traditional skills. Biostratigraphy, petrology and remote sensing are likewise ‘obsolete,’ although Carragher is not so sure.
On the exhibit floor, software vendors are partnering with big computer companies to offer more or less ‘shrink-wrapped’ solutions, integrating software, hardware, storage and networking. Schlumberger has partnered with HP on the port of its software to Linux and the companies plan to deliver a ‘plug & play’ upstream computing environment. Landmark has likewise partnered with IBM to offer a ‘Rapid Prospect Generation Engine’ and ‘on demand’ computing for infrastructure free deployment.
But the biggest buzz of the show was the amazing growth in the use of Linux clusters for high end visualization on common off-the-shelf (COTS) hardware. Fraunhofer Institute’s PV-4D offers rapid access to terabytes of stack or pre-stack data with high speed rendering and stereo display of multi-volume data. Cluster management has been developed in association with Linux Networx. Paradigm’s Reservoir Navigator leverages disk caching to make visualization possible on a laptop, with 1:20 or 1:40 memory to data ratios. This compares favorably with established techniques as used in Voxelgeo, and Magic Earth, which store all data in memory. HP’s Project Sepia, building on High Performance Computing work done for the US Department of Energy, uses COTS technology for simulation and seismic visualization. Sepia promises to be ‘10 times faster for 25% of the price of current solutions’. Schlumberger’s GigaViz also leverages clusters so that a 40GB volume can be interpreted from a laptop. All graphics processing is done on the cluster, only pixels that change are sent to the client. An oil company might install a big graphics processing cluster and give geoscientists low end machines. A demo with a 1400 km. sq. dataset showed that GigaViz does Magic Earth-like probes with fast pan and zoom. IBM was showing off its grid computing which promises a ‘self configurable’ environment. IBM’s Grid@petroleum offering includes guaranteed service levels, storage and computing on demand.
What do you do with all this compute power? Well you could try the IFP’s Cougar reservoir simulator driver. If you think reservoir simulation is resource hungry, try running multiple simulations for sensitivity analysis. Cougar output plots field oil prod total vs. time in days. Tornado plot shows sensitivity. Cougar selects Eclipse parameters using ‘experimental design,’ a statistical method for eliminating bias.
Tanks ’n tubes
Dave Hale’s (Landmark) work on seismic meshes has now extended into reservoir engineering. The seismic ‘image’ is segmented into a mesh of ‘tanks,’ polygonal reservoir blocks, and ‘tubes’ indicating transmissibility. Reservoir properties, such as tank pore volumes and transmissibility of tubes can be adjusted to test the impact of different seismic interpretations on fluid flow. Numerical experiments suggest that properties obtained from such coarse models can be used to constrain for more detailed models.
Philippe Baldy told of the lessons learned during Total’s successive mergers. Geoscience data used to be stored across four different systems. Fortunately, Total, Fina and Elf all used Schlumberger’s LogDB with well reference data in Finder, although customizations did differ. Now, Total’s LogDB now holds 120,000 references, and nearly one terabyte of data. It is believed to be one of the world’s largest. Total’s physical data was gathered and indexed, totaling 676,000 reports and 2.4 million ‘other’ items. Migration took 11 man years of effort over a 14 month period. Baldy warns, ‘The best QC does not mean error-free! The final QC stamp will be given by users.’ Total now has an acquisition and divestment (A&D)-friendly approach to data management which avoids duplication of effort between HQ and subsidiaries. Total is now planning a fully-integrated system, based on Finder, and a new Geoscience Web Portal.
Piantanida’s presentation stressed ENI’s shared earth model which is ‘distributed’ across people, subsidiaries and the HQ. 3D Data sharing spans the application workspace—GeoFrame, OpenWorks, Tigress, FlowMap, Petrel and Eclipse. Remote access is supported over an ASP link using Citrix MetaFrame and MIT’s ThinAnywhere. ENI is experimenting with a new Grid computing tool from Softricity. Best practices are stored in the technical ‘Know-how’ Portal as workflows for reservoir model building. Project directory uses Lotus Domino, Accenture’s E&P OnLine and SAP Portal 5.0. Workflow management includes description of tasks, best practices, database access with drill down to specific tasks such as facies identification.
INT is branching out into end-user software with the commercial release of INTViewer for pre-stack and attribute seismic visualization. INTViewer offers a ‘simple workflow’ for pre-stack data manipulation, ‘replacing an entire suite of vendor tools.’ A new wellbore schematics tool has been developed with Chevron-Texaco leveraging WITSML.
IPRES’ IPResource is a new reserves database for company reporting including annual reports, SEC and key performance indicators. The tool is developed atop of Business Objects. IPRES software is used by ChevronTexaco, ConocoPhillips, Marathon, Norsk Hydro and Statoil.
Neuro Genetic Solutions (NGS) was showing its neural networks and ‘committee machines’. A (human) domain specialist performs initial log interpretation which ‘teaches’ the neural net. Committee machines link multiple neural networks together to enhance results. NGS comes from the same academic stable as recent Schlumberger acquisition, Decision Team.
Scandpower’s MEPO history matching is now commercial. MEPO runs as a front end to the reservoir simulator to test out multiple hypotheses by optimizing history matching prior to full-scale modeling. Mepo uses Baysean analysis to check the validity of a match. Scandpower is currently running pilot studies on a 22CPU HP cluster running Linux.
Weatherford’s ‘Clarion’ 4D permanent in-well seismic monitoring is being trialed in the Izaute gas storage field in southwestern France. Both time-lapse (4D) and VSP and microseismics are recorded by a permanent five-station, three-component array and used to map gas-water contact variations.
EpiSEM, Rainaud’s (IFP) ‘knowledge-driven shared earth models’ capture interpretation metadata in ‘geo-ontologies’ and ‘abstract descriptors.’ Annotated interpretations can be shared between applications and stored for re-use. Tchistiakov (TNO-NITG) described an internet database of fluid flow simulations from over 32,500 models of shallow marine reservoirs. The project quantifies the influence of sedimentology, structure and up-lift on reservoir quality. Taner (RSI) is using joint time-frequency analysis with unsupervised neural networks to produce seismic lithology maps. Automatic event recognition and classification simulates human hearing perception. Kayser (Schlumberger) showed how Inside Reality can visualize the internal structure of cores. Densely spaced microfocus computer tomography (ÁCT) images are digitized for display in the Cave. According to Naess, Statoil’s experience on Heidrun has underlined the importance of the single common database for subsurface data. ‘Firmly defined’ work processes underpin remote drilling operations from Statoil’s new Onshore Support Center in Stjørdal. Naylor described how Shell’s new operating model has E&P organized into a global business, with ‘unambiguous, single point accountability for performance, portfolio and resources.’ This model contrasts with the autonomous asset-based approach. Technology is the key, including 3D VR and Real Time Operations Centers ‘leveraging global expertise’ to enable rapid decision making in remote locations. Baker (SPE chair) believes that society meetings are important, ‘people are gregarious; they like to come together.’ The development of virtual meetings is likely to impact professional societies’ revenue but ‘if we are to stay relevant in these days of restricted travel, we have to become a player in the virtual or distance-meeting realm.’ Stinson (Data Modeling) demonstrated automatic velocity analysis computed at every CMP and time location. The Auto Imager has been tested with real and synthetic data. A 32 node Linux cluster ran the SEG/EAGE 3D Salt Model in 11 hours (60,000 velocity profiles). Results are ‘as good as or better than’ the human picked velocities.
This report is abstracted from an illustrated, 38 page report produced as part of The Data Room’s Technology Watch Reporting Service. For more information on this subscription-based service please email email@example.com.
© Oil IT Journal - all rights reserved.