Automatic history matching (AMH) allows the generation of multiple reservoir models and selects the ones that match reservoir production history. Including sedimentological parameters like facies geometry within the AHM process could result in the selection of reservoir models with unrealistic facies geometry (sand-body dimensions that have not seen in nature). The use of these models can cause problems in production forecasting and reserves estimation. In this work we present a way to ensure the facies geometry realism within the AHM process. The use of sedimentological prior information in reservoir models provides a way to control relations between facies geometry parameters. Sedimentological prior information can be obtained from outcrops, modern depositional environments and high-resolution geophysical data. Sedimentological prior models quantitatively describe the natural relations among geomorphic parameters (e.g. channel width, thickness, sinuosity, etc.). In this work we tackle the problem of preserving sedimentological realism in AHM by building robust prior models that describe the non-linear dependencies between sedimentological parameters of deltaic systems. We built multi-dimensional realistic priors using a machine learning technique called One-Class Support Vector Machine. OC-SVM captures hidden relations of deltaic parameters: Delta Plane Width, Length and Thickness; Distributary Channel Width and Thickness, Meander Amplitude, and Wavelength and Mouth bar dimensions. We sample from the realistic priors in order to assure facies realism. A Multiple Point Statistics (MPS) algorithm is used to model facies in a deltaic reservoir. Variability of the geometries is produced by changing the MPS geometrical parameter, different training images and regions. We developed a technique to link the MPS geometrical parameter with the observed geological characteristics described by the intelligent priors used in history matching. History-matched models produced under geological realistic constraints reduce uncertainty of the production prediction, ensures the realism of the selected reservoir and also helps in the identification of the reservoir geometry. The reservoir geometry of the selected models demonstrated to be closer to the “true case” reservoir geometry.
|Number of pages||1|
|Publication status||Published - 28 Sep 2014|
|Event||AAPG International Conference and Exhibition - Istanbul, Turkey|
Duration: 14 Sep 2014 → 17 Sep 2014
|Conference||AAPG International Conference and Exhibition|
|Period||14/09/14 → 17/09/14|
Rojas, T. S., Demyanov, V., Christie, M. A., & Arnold, D. (2014). Uncertainty reduction in modelling deltaic reservoirs using intelligent sedimentological prior information. 1-1. Abstract from AAPG International Conference and Exhibition, Istanbul, Turkey. http://www.searchanddiscovery.com/abstracts/html/2014/90194ice/abstracts/1947237.html