--> ABSTRACT: A Hybrid Deterministic-Stochastic Approach to Modeling Reservoir Facies Distribution in the Deepwater Environment: The Example of Landana 1A Reservoir, Block 14, Angola

Datapages, Inc.Print this page

A Hybrid Deterministic-Stochastic Approach to Modeling Reservoir Facies Distribution in the Deepwater Environment: The Example of Landana 1A Reservoir, Block 14, Angola

Ingles, Antonio M.1; Clark, Julian 2; Jenkins, Steve 3; Grimes, David 2; Levy, Marjorie 2
(1) Applied Reservoir Management, Chevron Africa\Latin America EP, Luanda, Angola. (2) ETC, Chevron North America EP, San Ramon, CA. (3) TENGIZCHEVROIL, Atyrau, Kazakhstan.

The Landana reservoirs are composed of high quality sands deposited in a deepwater slope valley environment on the northwestern flank of the Congo River Fan. Target reservoir intervals in Landana 1A are Lower Miocene in age, consisting of thin to moderately thick successions of sandy turbidities.

Historically the use of seismic amplitude and inversion imaging techniques to pick well locations has been met with mixed success. Typically well predictions solely based on seismic products turn out to be very optimistic when measured up to actual well performance. Thus in a bid to improve reservoir quality prediction in deep water environments we hereby propose a fully integrated approach which combines seismic inversion, deterministically mapped channels, and deepwater facies interpretation into a robust 3D Geologic Model.

Initially deterministic channels are mapped using data from a coherence cube extracted along horizons parallel to an overlying regionally extensive condensed section. Centerlines are then digitized through the mapped channels in order to provide information on individual channel areal location, length and sinuosity which is required for deepwater training image generation.

A set of centerline-based deterministic deepwater training images are generated for each mapped channel or assembly of channels. These deterministic training images are generated using the digitized centerlines and are filled with axial and margin facies whose proportions are inferred from well log facies interpretation and analogue data.

Facies simulation conditioned to the deterministic training images (i.e. deterministically mapped channels) is then carried out using MPS (Multipoint Statistics Simulation) whereby the deterministic training images are treated as geobodies (and therefore hard data). In addition, a stochastic training image is used in conjunction with a seismic inversion-derived facies probability cube to predict reservoir quality outside the deterministically mapped geobodies and map out the degree of vertical connectivity between channels.

A thorough discussion of the methodology and lessons learned constitute the subject matter of the proposed presentation.

 

AAPG Search and Discovery Article #90135©2011 AAPG International Conference and Exhibition, Milan, Italy, 23-26 October 2011.