Machine Learning for Pattern Recognization of Stacked Sands

All around the world, significant oil reserves are contained in mature fields consisting of thin inter-bedded sand reservoirs. Due to the high level of geological heterogeneity in such reservoirs, the conventional reservoir recovery strategy often results in a high risk of bypassed pay. One key challenge to recover the remaining reserves is being able to accurately map sand geo-bodies versus shale. A second, more difficult challenge is in determining the porosity variation, clay and fluid content of the sand facies. An effective field redevelopment plan often requires to map 3-5 meter thick sands and be able to identify residual oil distribution within these sands.

Challenges

As seismic data is band limited most of the thin sand reservoirs could not be individually resolved and were seen as packages of thin sand and shales in the seismic inversion volumes. These predictions were limited to the resolution of the seismic data. Multiple realizations of the geostatistical seismic inversion directly in the geomodel grid can be used to understand the uncertainty associated with fine sub seismic resolution and better characterize the thin sand distribution which can be used to identify bypassed oil for future wells planning. However, seismic inversion technologies are often only performed on seismic amplitude data, therefore do not utilize the subtle information embedded in seismic attributes.  

Benefits

  • Identify thin beds
  • See detailed stratigraphy below seismic resolution
  • Identify stacking pattern of channelized units
  • Reducing risk in drilling marginal or dry holes
scenario testing

Seismic expression of varied stacking pattern of thin sand beds

All around the world, significant oil reserves are contained in mature fields consisting of thin inter-bedded sand reservoirs. Due to the high level of geological heterogeneity in such reservoirs, the conventional reservoir recovery strategy often results in a high risk of bypassed pay. One key challenge to recover the remaining reserves is being able to accurately map sand geo-bodies versus shale. A second, more difficult challenge is in determining the porosity variation, clay and fluid content of the sand facies. An effective field redevelopment plan often requires to map 3-5 meter thick sands and be able to identify residual oil distribution within these sands.

Challenges

As seismic data is band limited most of the thin sand reservoirs could not be individually resolved and were seen as packages of thin sand and shales in the seismic inversion volumes. These predictions were limited to the resolution of the seismic data. Multiple realizations of the geostatistical seismic inversion directly in the geomodel grid can be used to understand the uncertainty associated with fine sub seismic resolution and better characterize the thin sand distribution which can be used to identify bypassed oil for future wells planning. However, seismic inversion technologies are often only performed on seismic amplitude data, therefore do not utilize the subtle information embedded in seismic attributes.  

Benefits

  • Identify thin beds
  • See detailed stratigraphy below seismic resolution
  • Identify stacking pattern of channelized units
  • Reducing risk in drilling marginal or dry holes
scenario testing

Seismic expression of varied stacking pattern of thin sand beds

Solution

Attribute Generator

Closed-loop™ allows extraction of a broad range of seismic attributes used for multi-attribute analysis. To facilitate the task of selecting a specific attribute, Closed-loop™  categorizes the attributes into 7 main types. 

Family of Seismic Attribute
  • Bright & dim spot
  • Oil and gas position abnormality
  • Formation discontinuity
  • Structure discontinuity
  • Pinch-out
  • Unconformity traps
  • Block upwarping 
  • In common use
attribute generator

Solution

Attribute Generator

Closed-loop™ allows extraction of a broad range of seismic attributes used for multi-attribute analysis. To facilitate the task of selecting a specific attribute, Closed-loop™  categorizes the attributes into 7 main types. 

Family of Seismic Attribute
  • Bright & dim spot
  • Oil and gas position abnormality
  • Formation discontinuity
  • Structure discontinuity
  • Pinch-out
  • Unconformity traps
  • Block upwarping 
  • In common use
attribute generator

Cluster

iRes-Geo’s proprietary cluster method is based on advanced multi-attributes statistics. It groups similar objects and separates non-similar objects. For patterns without class information, the cluster method uses a certain standard to label the members of a class and the extent to which a pattern belongs to the classes. Cluster is often used in quantitative seismic facies analysis or pattern identification. We extract a variety of characteristic parameters, which can be divided into four categories: frequency-related parameters, energy-related parameters, waveform-related parameters, and self-regression coefficients, as the input to our multi-attribute quantitative forecast workflow. 

attribute plot
cluster
seismicdriving

Cluster

iRes-Geo’s proprietary cluster method is based on advanced multi-attributes statistics. It groups similar objects and separates non-similar objects. For patterns without class information, the cluster method uses a certain standard to label the members of a class and the extent to which a pattern belongs to the classes. Cluster is often used in quantitative seismic facies analysis or pattern identification. We extract a variety of characteristic parameters, which can be divided into four categories: frequency-related parameters, energy-related parameters, waveform-related parameters, and self-regression coefficients, as the input to our multi-attribute quantitative forecast workflow. 

attribute plot
cluster
seismicdriving