Abstract:The development of shale gas can optimize the energy structure and alleviate the sharp contradiction between energy supplies and demands. Stochastic simulation of shale reservoirs is a common and practical method for predicting and evaluating shale gas distribution and extent. Shale has complex pore structure, strong heterogeneity and low permeability with multi-scale natural fractures, which makes it difficult to characterize or simulate using common numerical simulation methods. Generative adversarial network (GAN), as the mainstream generative model of recently flourishing deep learning methods, can theoretically learn any feature distribution of training images and generate fake images that are similar to real ones, which has been successfully used in geological stochastic simulation. The original GAN is unstable due to the defect of the loss function and its structure, possibly leading to significant differences between the reconstructed results and the real samples. Moreover, the ability of extracting features from the original GAN is restricted by the demand of a large number of training datasets, which makes a difficult prerequisite for shale reconstruction. The multi-scale structure can capture the characteristics of shale at different scales, suitable for multi-scale internal structure of shale. Attention mechanisms can focus on important information and suppress unwanted information. Therefore, a multi-scale GAN combining with attention mechanisms for 3D shale reconstruction is proposed based on only one shale image. The original GAN's loss function is replaced by the hinge-GAN loss function. By comparing with other methods using various measurement metrics, it is proved that our method is effective in shale reconstruction and has good consistency in physical properties with the real data.

Super-resolution of Shale Image Using Double Deep Convolutional Neural Networks in Pixel-Gradient Domain

Super-resolution Reconstruction Algorithms Based on Fusion of Deep Learning Mechanism and Wavelet

A Super-Resolution Reconstruction Method for Shale Based on Generative Adversarial Network

Depth Image Super-resolution via Two-Branch Network

Single Image Super-Resolution Based On Deep Learning And Gradient Transformation

Research on image super-resolution algorithm based on mixed deep convolutional networks

Image Super-Resolution Algorithm Based on Dual-Channel Convolutional Neural Networks

A Real-Time Super-Resolution Method Based on Convolutional Neural Networks.

Reconstruction of Shale Image Based on Wasserstein Generative Adversarial Networks with Gradient Penalty

Deep Deconvolution Neural Network for Image Super-Resolution

Shale 3D Reconstruction Method Based on Context-Aware Generative Adversarial Networks and High-Resolution Optical Flow Estimation

Stochastic reconstruction of shale combining multi-scale generators and discriminators with attention mechanisms

Deep Learning for Subsurface Penetrating Super-Resolution Imaging

Prediction of Fractal Dimension in Shale CT and Its Robustness to Interference Based on Convolutional Neural Networks

Single Image Super-resolution Reconstruction Based on Enhanced Residual Network

Single Image Super Resolution Reconstruction Algorithm Based on Deep Learning

Learning Depth Super-Resolution by Using Multi-Scale Convolutional Neural Network.

SHBGAN: Hybrid Bilateral Attention GAN for Seismic Image Super-Resolution Reconstruction

Towards Underwater Image Enhancement Using Super-Resolution Convolutional Neural Networks.

From Deep Image Decomposition to Single Depth Image Super-Resolution

Super-resoloution Image Reconstruction Based on Wavelet Packet Transform and Artifcial Neural Networks