Multi-Point Statistics Inversion: The Test and Evaluation of a New Approach

Abstract

Multipoint statistics, along with seismic data, has long been used as a tool in addressing subsurfaces uncertainties. Multi-point statistics inversion, in particular, is an area of focus in recent years. Despite this, multi-point statistics inversion is still in its early stage, and significant work yet needs to be done to push the technique forward.

The current study introduces, test and evaluates a new approach for multi-point statistics inversion. The approach can be divided into 5 steps, including: 1) establish training images, and derive probabilistic distribution functions (PDFs) of seismic attributes for different facies using near-well seismic trace analysis or forward modeling; 2) facies modeling with multipoint statistics (MPS); 3) based on facies modeling results for a particularly cell, randomly select seismic attributes/impedance using the PDFs derived in step 1; 4) generate seismic traces through convolution, best-match to real seismic traces and compare with previously generated traces; 5) repeat steps 2-4 until all the cells are modeled to get the final inversion results.

The approach is tested and evaluated using a 2-D synthetic profile, and results shows that reproduced facies models and seismic matches the synthetics very well. The approach, thus, potentially could be utilized in real subsurface data to address reservoir geological uncertainties for well positioning and production optimization purposes.

This approach is financed by the National Nature Science Foundation(No: 41572081) ,the National Science Important Project(No:2016ZX05031002-001) and Natural Science Innovative group project of Hubei Province(No: 2016CFA024)

key words: multi-point statistics, inversion, subsurface uncertainties, seismic attributes, probabilistic distribution functions

AAPG Datapages/Search and Discovery Article #90323 ©2018 AAPG Annual Convention and Exhibition, Salt Lake City, Utah, May 20-23, 2018