Abstract : Discovery of potential hydrocarbon zones is largely dependent on visual perception of sub surface features in seismic images. However, noise and other image artefacts introduced by acquisition constraints, instrument constraints and acoustic attenuation of earth degrade the quality of seismic images , thereby increasing the uncertainty in exploration decisions. The aforementioned quality problem is further exacerbated by the growing industry trend towards deeper and thinner exploration targets. Hence, enhancement of seismic images is a primary endeavour in the exploration workflow. This study considers seismic image enhancement as a problem of improving seismic resolution , upscaling seismic images and attenuating seismic noise. Approaching the enhancement problem as a prediction problem, this study explores the use of deep learning enabled intelligent de-noising and Super Resolution methods to enhance seismic images. This study aims to find efficient deep learning networks to achieve automatic and generalised seismic image enhancement through resolution improvement and de-noising which result in proven improvement in seismic interpretation workflows. Building on the recent research in SRGANs and conditional GANs, this study measures the seismic enhancement performance of deep learning at various types and degrees of noise contamination. It further assesses the learning efficacy of using seismic texture conditional priors in place of lithology labels. It attempts to determine the efficacy of deep learning networks in emulating denser seismic acquisition by adaptation of super resolution methods at 2x upscaling. Moreover, this study attempts to bring consistency to the performance measurement of seismic enhancement through uniform evaluation of different methods using metrics like PSNR, SSIM, MS-SSIM, LS for supervised methods. It also evaluates signal preservation capacity of deep learning methods used to automatically remove incoherent noise , through local similarity metric.
Complete Research Proposal can be found here.
Figure 1 : Solution Approach
Sonic logs are required for well-seismic ties, an essential step in the exploration workflow. However, due to various constraints, sonic logs might not be recorded in several wells within a field. This is especially true for older wells. It is also seen that basic logs like GR, RT, RHOB, NPHI are very likely to be recorded in all wells.
So far, geoscientists approximate missing sonic logs from sonic logs of nearby wells using empirical techniques. Such techniques might not be a robust representation of the the sub surface environment. Further, these techniques are manual and inconsistent. Moreover, such missing logs are approximated as required by a certain study and may not be shared for general consumption by all projects.
I propose a new data driven approach to predict missing sonic logs on a field scale. I hope to predict sonic logs for all the wells in a field and make them easily available.
Complete proposal can be found here
Updated June 14, 2021 9:43 AM (GMT+5:30)
Prepared in collaboration with I/c Special Projects Team at GEOPIC. This project aims to automatic well seismic ties using machine learning.
All the details can be found here.
