An integrated fracture parameter prediction and characterization method  in deeply-buried carbonate reservoirs based on deep neural network

Yasin, Qamar; Ding, Yan; Baklouti, Syrine; Boateng, Cyril D.; Du, Qizhen; Golsanami, Naser

An integrated fracture parameter prediction and characterization method in deeply-buried carbonate reservoirs based on deep neural network

Files

6.pdf(29.97 MB)

Date

2022

Authors

Publisher

Journal of Petroleum Science and Engineering

Abstract

However, fracture prediction in deeply buried carbonate reservoirs has always been challenging. Fracture prediction in the deep-buried carbonate structure of North China is problematic because of multiphase tectonic movements, variable sediment lithology, and complex diagenesis. Because of deep burial depth and complex heterogeneity, the resolution of seismic reflection data beneath the buried-structure is poor, making it challenging to identify the fault reflection characteristics. This paper proposes a novel idea to identify natural fractures in carbonate reservoirs using conventional logs with seismic reflection data. The proposed model can also predict the fracture aperture and fracture density, a distinctive feature. Another novel hybrid model based on deep-learning neural network (DNN) and cluster analysis is proposed to predict further the spatial variations of lithology, porosity, and fracture parameters from seismic inversion. The proposed models provide valuable insights that help determine fracture parameters in the Paleozoic strata and associated reservoirs through quantitative analysis using petrophysics, rock physics, seismic inversion, and seismic attributes. The overlapping of seismic interpreted fault networks and spatial variations of the inverted fracture parameters indicate a high correlation of fracture development zones. The methodology proposed in this study presents a valuable template valid for the characterization of fractured reservoirs in deeply-buried carbonate reservoirs throughout the world.