Next: About this document ... Up: Reproducible Documents

SEP177 -- TABLE OF CONTENTS

DAS observations of perforation-induced guided waves in a shale reservoir [

SRC]
Ariel Lellouch, Steve Horne, Mark Meadows, Stuart Farris, Tamas Nemeth, and Biondo Biondi
Perforation shots can be recorded by downhole DAS arrays. In this study, we demonstrate that guided waves induced by perforation shots propagate in a low-impedance shale reservoir layer. Such guided waves have an extremely high frequency content of up to 700 Hz and are dispersive, with lower frequencies propagating faster than higher frequencies. They can propagate as both P- and S-waves, and their group velocity is higher than their phase velocity. The high temporal and spatial resolution of the DAS array allows for their unaliased recording despite their short wavelengths. The guided waves disappear from the records when the well exits the shale formation. Synthetic modeling predicts their existence for both the acoustic and elastic cases in simple velocity models. We also show that perforation shots from an offset well at a distance of 260 m can be recorded by the DAS array. Induced guided waves undergo significant disturbances when propagating through previously stimulated zones. These disturbances manifest as kinematic and dynamic changes of the recorded wavefield, and as scattered events. The nature of the stimulation-induced changes remains unresolved and their behavior is interpreted as a combination of unknown spatial and temporal effects. Guided waves hold tremendous potential for high-resolution reservoir imaging and should be used in conjunction with conventional DAS arrays and state-of-the-art DAS interrogators.

Can we image guided waves by elastic full waveform inversion? [

SRC]
Biondo Biondi, Ariel Lellouch, Ettore Biondi and Stuart Farris
We investigate the feasibility, and potential advantages, of applying elastic full-waveform inversion to guided waves generated by perforation shots and recorded by a distributed acoustic sensing fiber cemented into a horizontal well. Since a correct source model is essential for waveform inversion, we first show that the amplitudes and phases observed in the field data are better modeled using a dipole source than a monopole source. We then illustrate the high-resolution potential of waveform inversion by comparing data modeled assuming a horizontally layered medium with and without a thin high-velocity layer in the reservoir. Significant phase differences are observed with a two-meter thick high-velocity layer, and smaller, but still visible, phase differences are observed with a one-meter thick intrusion. Finally we analyze the data residuals of a first iteration of a hypothetical waveform inversion process when the starting model is a vertical average of the true model. We show that data-residuals phases at either short offsets ($\leq$150 m), or at low frequencies ($\leq$80 Hz), are sufficiently close to the corresponding phases of data modeled with the starting model. This result indicate that a waveform inversion process carefully bootstrapped from near offsets and low frequencies should not suffer from cycle skipping.

Full waveform inversion by model extension [

SRC]
Guillaume Barnier and Ettore Biondi
We show that full waveform inversion by model extension (FWIME) combined with a model-space multi-scale approach has the potential to become a robust velocity model-building algorithm that mitigates the cycle-skipping issues inherent to conventional FWI. Its consistent and concise mathematical formulation coupled with an automatized implementation makes it simple to apply and thus more accessible to a broad range of non-expert users. We first apply FWIME to recover a very accurate and high-resolution Marmousi model by starting with a naive initial guess and without the use of low-frequency energy. We dramatically improve the results obtained in our previous report. In a second numerical example, we show that FWIME has the ability to correctly update large areas with substantial kinematic errors and mispositioned sharp interfaces. These conclusive numerical tests improved our understanding of the technique and showed promise as we move towards applying FWIME to field data which requires handling elastic effects and complex overburdens.

Anisotropic full waveform inversion with pore pressure constraints: A field data application [

SRC]
Huy Le, Biondo Biondi, Robert G. Clapp, and Stewart A. Levin
We presents an application of anisotropic full waveform inversion with pore pressure constraints on a 3D field data. The frequency band chosen for FWI is 3-15 Hz and 3-30 Hz for RTM. The source wavelet is inverted directly from the data. The objective function is normalized in amplitude in a trace by trace manner. The constraints are derived for vertical velocity only, based on the rock physics workflow described our previous report. An upper bound on velocity is constructed from hydrostatic pore pressure while an lower bound is constructed from mud weight data. After the inversions, the objective function is reduced by 25\%. When the constraints are enforced, vertical velocity changes by 5\% while anisotropy changes by 10\%. On the other hand, after unconstrained inversion, velocity changes by 10\% and anisotropy changed by 5\%. The accuracy of the velocity model is improved, demonstrated by enhanced reflectors' focusing and continuity, especially when the constraints are incorporated.

Regularization strategies for time-lapse full-waveform inversion [

SRC]
Yinbin Ma
In this chapter, we study regularization strategies for time-lapse full-waveform inversion in VTI media. We show that total-variation regularization on the model differences improves the inversion results by removing fine-scale fluctuation. For integrated reservoir monitoring where geomechanical information is available, we can construct a geomechanics constrained regularization on the model differences, based on the third order elastic theory. With extra information on the reservoir compaction, we can impose a regularization on vertical velocity changes based on the RTM image alignment.

Automatic detection of VZ noise and an observation of how such noise is induced [

SRC]
Joseph Jennings and Shuki Ronen
We developed a scalar attribute whose strength indicates how much VZ noise is present on a particular node. Using a 3D OBN dataset acquired in the Gulf of Mexico, we are able to show that our attribute successfully detects VZ noise. As a result of this attribute calculation, we found that nodes deployed over shallow salt had VZ noise while nodes far away from the node did not.

Joint Inversion of Reflectivity and Background subsurface components (JIRB) [

SRC]
Alejandro Cabrales-Vargas
I perform a joint inversion for the reflectivity and background components of the subsurface. I pose this method as an optimization problem that is linear with respect to the reflectivity component, and non-linear with respect to the background component, therefore becoming overall nonlinear. Feasible solutions are confined to background models that drive the migration image to maximum focusing. The latter is achieved by maximizing the energy of the migration image in a second functional. The numerical results demonstrate that the method can recover a corrected reflectivity value.

Non-Stationary Low Frequency De-noising using Prediction Error Filters[

SRC]
Milad Bader and Robert Clapp
Following the low-frequency de-noising method introduced previously, where a stationary prediction-error filter is estimated from the high-frequency data, we extend the process to the non-stationary case. We account for the non-stationarity by estimating a bank of prediction-error filters defined on a regular grid. Two sets of filters are built, one for the high-frequency signal and one for the low-frequency data. The first is expanded in time and space, then used with the second in a regularized inverse problem to estimate the low-frequency signal. The method is tested on a synthetic shot gather.

About this document ...

Next: About this document ... Up: Reproducible Documents

2019-09-15