--> Paleokarst Reservoir Seismic Facies Classification Using Frequency Spectral Decomposition, Machine Learning and Electrofacies Constrained Approach
[First Hit]

2019 AAPG Annual Convention and Exhibition:

Datapages, Inc.Print this page

Paleokarst Reservoir Seismic Previous HitFaciesNext Hit Previous HitClassificationNext Hit Previous HitUsingNext Hit Frequency Spectral Decomposition, Previous HitMachineNext Hit Previous HitLearningNext Hit and Electrofacies Constrained Approach

Fei Tian1, Qingyun Di1, Zhongxing Wang1, Shiyin Li2, Yao Lv1, Xiaocai Shan1, Chunguang Shen2, Baozhu Guan2, Xin Liu2, Xiaoxue Wang2


1Institute of Geology and Geophysics, Chinese Academy of Sciences;
2Research Institute of Exploration & Development, Tarim Oilfield Company PetroChina


May 19-22 2019 – 2019 AAPG Annual Convention and Exhibition, San Antonio, Texas

Posted: June 30, 2019

Abstract

More than 1.5 billion tons of hydrocarbons have been found in the Central Tarim basin, which are mainly stored in caves, vugs and fractures. Seismic characterization of such carbonate reservoirs is real challenging for geophysicists because of the deeply buried depth, interior structures and fine scales. Conventional qualitative descriptions of the internal structure and geometry of seismic reflections can’t meet the requirements from geological modeling and reservoir simulations. Based on frequency spectral decomposition, electrofacies and Previous HitmachineNext Hit Previous HitlearningNext Hit techniques, a new quantitative seismic Previous HitfaciesNext Hit analysis approach was developed, whose procedure is sequential according to the different seismic reflections, and has successfully applied for reservoir characterization.

Based on geological understanding and related seismic reflections, the seismic Previous HitfaciesNext Hit analysis was divided into four steps: 1) First, we employ continuous short-time Fourier transform to divide high, middle, and low frequency components of seismic data to quantitatively predict reservoirs with different scales; 2) The amplitude, acoustic impedance, RMS, chaos, variance, and sweetness etc., were used as input seismic datasets. Supervised Previous HitmachineNext Hit Previous HitlearningNext Hit method was used for the strong amplitude analysis, in which we used cores and electrofacies interpretations as prior knowledge; 3). Unsupervised Previous HitmachineNext Hit Previous HitlearningNext Hit method was used for weak amplitude seismic Previous HitfaciesNext Hit Previous HitclassificationNext Hit and the resulted seismic Previous HitfaciesNext Hit were interpreted by electrofacies near boreholes. 4). 6 seismic Previous HitfaciesNext Hit were classified, and their spatial geometry were delineated and analyzed with faults and unconformity distributions.

Our integrated approach honored seismic frequency, multi attributes and multi scale combinations among geology-well logging-seismic data, classified seismic Previous HitfaciesTop quantitatively, provided an accurate overall solution for paleokarst reservoir characterization, and can be applied to similar sedimentary basins that accommodate deeply buried carbonate reservoirs affected by karstifications and faults.

AAPG Datapages/Search and Discovery Article #90350 © 2019 AAPG Annual Convention and Exhibition, San Antonio, Texas, May 19-22, 2019