Quantitative evaluation of fracture porosity from dual laterlog based on deep learning method

• DOI: 10.1016/j.engeos.2021.08.006
• Publication Date: 2021-09-02T02:49:13Z
• Authors: Song Hu; Xiaochang Wang; Jin Wang; Lei Wang
• Keywords: Bedrock

Fracture porosity is one of the key parameters for characterizing fractured reservoirs. However, fracture calculation difficult with conventional logging data due to severe anisotropy To deal problem, equivalent macroscopic anisotropic formation model based on dual laterolog (DLL) adopted cyclically assign such as bedrock resistivity (RB), fluid in fractures (RFL), dip angle (FDA) and thickness well spacing, produce massive modeling. A large number training obtained through three dimensional finite element forward modeling functional relationship between DLL responses that are trained summarized by deep neural network, combined establish a new fast calculating formations. inversion reservoirs gradient optimization algorithm multi-initial strategy then proposed. While running model, divided into eight intervals according from 0° 90° every 0.5° improve operation speed efficiency. The results numerical verification show when greater than 1000 Ω m, mean absolute error (MAE) 0.001658% horizontal fractures, 0.00413% intermediate 0.0027% quasi-vertical fractures. When 100 m MAE 0.003% 0.0034% 0.00348% determined actual good agreement micro imaging logging.