Integrated kinematic time-lapse inversion workflow leveraging full-waveform inversion and machine learning

Maharramov, Musa; Willemsen, Bram; Routh, Partha S.; Peacock, Emily F.; Froneberger, Mark; Robinson, Alana; Bear, Glenn W.; Lazaratos, Spyros

doi:10.1190/tle38120943.1

Cited by 8 publications

(2 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There exists an extensive literature on wave-based time-lapse imaging aimed at producing images that show time-lapse differences in the image space by carrying out inversion rather than imaging [Qu and Verschuur, 2017, Yang et al, 2016, Queißer and Singh, 2013, Maharramov et al, 2019. A prominent example of such inversion method is formed by time-lapse monitoring via double differences Huang, 2013, Yang et al, 2015] where differences are obtained by inverting differences between the monitor and baseline residuals rather than between observed and synthetic monitoring data.…”

Section: Time-lapse Monitoring With the Joint Recovery Modelmentioning

confidence: 99%

Compressive time-lapse seismic monitoring of carbon storage and sequestration with the joint recovery model

Yin

Louboutin

Herrmann

2021

Preprint

View full text Add to dashboard Cite

Time-lapse seismic monitoring of carbon storage and sequestration is often challenging because the time-lapse signature of the growth of CO2 plumes is weak in amplitude and therefore difficult to detect seismically. This situation is compounded by the fact that the surveys are often coarsely sampled and not replicated to reduce costs. As a result, images obtained for different vintages (baseline and monitor surveys) often contain artifacts that may be attributed wrongly to time-lapse changes. To address these issues, we propose to invert the baseline and monitor surveys jointly. By using the joint recovery model, we exploit information shared between multiple time-lapse surveys. Contrary to other time-lapse methods, our approach does not rely on replicating the surveys to detect time-lapse changes. To illustrate this advantage, we present a numerical sensitivity study where CO2 is injected in a realistic synthetic model. This model is representative of the geology in the southeast of the North Sea, an area currently considered for carbon sequestration. Our example demonstrates that the joint recovery model improves the quality of time-lapse images allowing us to monitor the CO2 plume seismically.

show abstract

Section: Time-lapse Monitoring With the Joint Recovery Modelmentioning

confidence: 99%

Compressive time-lapse seismic monitoring of carbon storage and sequestration with the joint recovery model

Yin

Louboutin

Herrmann

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…In the geoscience community, there is also a boom of research interest in ML and the application of ML algorithms, specifically to seismic exploration. For example, NN are trained to perform facies analysis (Wrona et al, 2018;Zhong et al, 2019;Liu et al, 2018;Qi et al, 2020), seismic event or first arrival picking (Zhu and Beroza, 2018;Qu et al, 2019;Hu et al, 2019a), fault detection (Xiong et al, 2018;Qi et al, 2019;Wu et al, 2019b), salt body interpretation (Di et al, 2018;Di and AlRegib, 2019;Morris et al, 2019;Ye et al, 2019), de-noising (Dong et al, 2019;Sun et al, 2019;Wu et al, 2019a;Yu et al, 2019;Zu et al, 2019;Li, 2020), interpolation (Jia and Ma, 2017;Jia et al, 2018;Wang et al, 2019a,b), 4D monitoring (Maharramov et al, 2019;Liu and Grana, 2019;Yuan et al, 2019), acquisition optimization (Chamarczuk et al, 2019;Jiang et al, 2019;Nakayama et al, 2019) and geo-engineering applications (Gu et al, 2018;Terry et al, 2019).…”

Section: Introductionmentioning

confidence: 99%

ML-Misfit: Learning a Robust Misfit Function for Full-Waveform Inversion Using Machine Learning

Sun

Alkhalifah

2020

EAGE 2020 Annual Conference &Amp; Exhibition Online

View full text Add to dashboard Cite

Most of the available advanced misfit functions, including those that utilize a matching filter or the optimal transport measure, are all hand-crafted. Tuning and improving those methods requires profound mathematical and physical knowledge from FWI practitioners.Besides, the performance of those misfit functions is data-dependent, and it is challenging to find a solution when these approaches fail to invert a specific dataset. We propose to learn a robust misfit function for FWI, entitled ML-misfit, based on machine learning. Inspired by the recently introduced optimal transport of the matching filter (OTMF) objective, we design a specific neural network (NN) architecture for the misfit function in a form similar to comparing the mean and variance for two distributions in OTMF. To guarantee the resulting learned misfit is a metric, we accommodate the symmetry of the misfit with respect to its input and a Hinge loss regularization term in the meta-loss function to satisfy the "triangle 1 arXiv:2002.03163v2 [physics.geo-ph] 18 Mar 2020 inequality" rule. In the framework of meta-learning, we train the network by running FWI to invert for randomly generated velocity models and update the parameters of the NN by minimizing a meta-loss, which is defined as the accumulated difference between the true and inverted models. The learning and improvement of such an ML-misfit are automatic, and the resulting ML-misfit is data-adaptive. We first illustrate the basic principles behind the ML-misfit for learning a convex misfit function for travel-time shifted signals. Further, we train the NN on 2D horizontally layered models, and we demonstrate the effectiveness and robustness of the learned ML-misfit by applying it to the well-known Marmousi model.

show abstract

Time-lapse difference inversion based on the modified reflectivity method with differentiable hyper-Laplacian blocky constraint

Tang

Zhang

et al. 2021

GEOPHYSICS

View full text Add to dashboard Cite

Time-lapse (TL) seismic has great potential in monitoring and interpreting time-varying variations in reservoir fluid properties during hydrocarbon exploitation. Obtaining the disparities of reservoir elastic parameters by inversion is essential for TL reservoir monitoring. Conventional TL inversion is carried out by stages without coupling processing and leads to inaccurate results. We directly use the differences in seismic data responses from different vintages, namely, difference inversion, to improve the results’ credibility. It may reduce the deviations of the subtraction of base and monitor inversions in traditional methods. Moreover, most existing studies involving prestack inversion methods use the Zoeppritz equation or its approximants, which failed to consider the wave-propagation effects. We have developed a new TL difference inversion based on the modified reflectivity method (MRM), in which the internal multiples and transmission losses are taken into consideration to fine-tune the characterization of the seismic wave propagating underground. The new method is modified on the basis of the reflectivity method making it feasible in TL difference inversion, and it is derived from the Bayesian theorem. For further delineating the boundaries between layers, the differentiable hyper-Laplacian blocky constraint (DHLBC) is introduced into the prior information of a Bayesian framework, which heightens the sparseness in the vertical gradients of the inversion results and leads to sharp interlayer boundaries of the difference parameters. Synthetic and field data experiments demonstrate that our TL difference inversion method has clear advantages in accuracy and resolution compared with the Zoeppritz method and MRM without DHLBC.

show abstract

Integrated kinematic time-lapse inversion workflow leveraging full-waveform inversion and machine learning

Cited by 8 publications

References 24 publications

Compressive time-lapse seismic monitoring of carbon storage and sequestration with the joint recovery model

Compressive time-lapse seismic monitoring of carbon storage and sequestration with the joint recovery model

ML-Misfit: Learning a Robust Misfit Function for Full-Waveform Inversion Using Machine Learning

Time-lapse difference inversion based on the modified reflectivity method with differentiable hyper-Laplacian blocky constraint

Contact Info

Product

Resources

About