Speaking at the 2014 PNEC Conference in Houston this month, Jane McConnell presented the results of a joint development project with Statoil to test the Teradata data warehouse appliance in an oil and gas context. Teradata’s first appearance at PNEC goes back to 2006 with Nancy Stewart’s seminal presentation of the use of the technology chez Wal-Mart. In the Statoil case, Teradata is used to store georeferenced seismic data from successive surveys alongside other data sets such as pressures and fluid content. Interpretation and simulator results are also loaded to the ‘reservoir data warehouse.’
While conventional interpretation technology can be used to demonstrate, for instance, how the time shift between successive surveys can be used as a proxy for reservoir pressure, the data warehouse approach allows for more in-depth analysis. Standard business intelligence/big data queries can be used to investigate a multi-dimensional data space and provide a ranking of different correlations. The approach is claimed to speed processing of new time lapse data which is now acquired every six months, or in some cases continuously, allowing for insights obtained from the data to be turned into actionable information for production operations in a timely fashion.
The test involved a Teradata EDW 6690 data warehouse appliance and tiered storage combining ‘hot’ solid-state drives and ‘cold’ hard disks with automated data management. An online analytical processing database optimized for fast access is claimed to provide more flexibility and performance than traditional applications such as Oracle database-driven environments. Data is stored in a ‘fit-for-purpose’ monolithic database that includes seismic traces and other domain specific data as user defined binary data type similar to SEG-Y. This ensures the seismic data fits in an acceptable amount of space. All data types are exposed via an extended SQL vocabulary that includes spatial query.
In a reservoir monitoring proof of concept, a large number of queries were run automatically against the data to derive a meaningful subset of correlations. Further root investigation of possible causes for seismic time shifts between surveys can be carried out with business intelligence applications such as Spotfire. The approach is claimed to provide insights into the data that would otherwise require much interpreter ‘grunt-work’ using conventional workflows. Current research is expanding the project’s scope to pre-stack seismic and history matching. More from PNEC and Teradata.
© Oil IT Journal - all rights reserved.