Abstract:Seismic inversion is the process of converting seismic reflectivity data into physical subsurface properties. The most common inversion methods use physics-based forward modeling, but these require time-consuming steps such as initial model building and wavelet extraction. Coherent noise in the seismic volume may also lead to suboptimal results. Advances in deep learning enable the development of new geophysical workflows that may help overcome these challenges. One example is the temporal convolutional network (TCN), a deep neural network that learns from sequential data such as seismic traces. Previous research using the TCN architecture showed promising inversion results on synthetic data. Applying the method to real data has several additional challenges that need to be considered, including complex noise and limited well availability. A noisy poststack field seismic data set and seven wells were used to evaluate using a TCN for acoustic impedance inversion under these conditions. We found a TCN trained using only field data acquired at the well locations produced better results than conventional inversion, with the neural network learning to ignore the coherent noise. This method also makes semi-automated seismic inversion a possibility. Obtaining acceptable results may only be possible in relatively simple geological scenarios and ideally needs many wells for training. To overcome the issue of training using small data sets, we developed a workflow for generating realistic synthetic data to provide more samples and variation for the model to learn from. A TCN model trained using synthetic data ultimately produced the best impedance estimates, but care is needed to ensure the synthetic traces contain realistic noise. Overall, we show that a TCN can successfully invert seismic data contaminated with coherent noise, producing superior results compared to model-based inversion in this example.

Seismic Impedance Inversion Using Fully Convolutional Residual Network and Transfer Learning

Transformer and Convolutional Hybrid Neural Network for Seismic Impedance Inversion

Semi-Supervised Seismic Impedance Inversion With Convolutional Neural Network and Lightweight Transformer

Residual Learning of Deep Convolutional Neural Network for Seismic Random Noise Attenuation

High-resolution Seismic Impedance Inversion Integrating the Closed-Loop Convolutional Neural Network and Geostatistics: an Application to the Thin Interbedded Reservoir

High-resolution closed-loop seismic inversion network in time-frequency phase mixed domain

Deep learning for high-resolution multichannel seismic impedance inversion

ROBUST DEEP LEARNING-BASED SEISMIC INVERSION WORKFLOW USING TEMPORAL CONVOLUTIONAL NETWORKS

Acoustic Impedance Prediction Using an Attention-Based Dual-Branch Double-Inversion Network

Acoustic Impedance Inversion from Seismic Imaging Profiles Using Self Attention U-Net

Low-frequency constrained seismic impedance inversion combining large kernel attention and long short-term memory

Deep Learning-Based Low-Frequency Extrapolation and Impedance Inversion of Seismic Data

Semi-supervised Impedance Inversion by Bayesian Neural Network Based on 2-d CNN Pre-training

InversionNet: A Real-Time and Accurate Full Waveform Inversion with CNNs and continuous CRFs

Physics-Constrained Seismic Impedance Inversion Based on Deep Learning

Seismic impedance inversion based on cycle-consistent generative adversarial network

Research on Elastic Parameter Inversion Method Based on Seismic Facies-Controlled Deep Learning Network

Seismic velocity inversion based on CNN-LSTM fusion deep neural network

Multiscale Data-driven Seismic Full-waveform Inversion with Field Data Study

Multitask Seismic Inversion Based on Deformable Convolution and Generative Adversarial Network

Deep Learning Inversion of Electrical Resistivity Data