## Tesi etd-04232014-093704 |

Tipo di tesi

Tesi di dottorato di ricerca

Autore

ALEARDI, MATTIA

URN

etd-04232014-093704

Titolo

Seismic inversion methods for overpressure prediction and elastic seabed characterization

Settore scientifico disciplinare

GEO/11

Corso di studi

SCIENZE DI BASE

Commissione

**tutor**Prof. Mazzotti, Alfredo

Parole chiave

- seabed characterization
- overpressure prediction
- full-waveform inversion
- amplitude versus angle
- stochastic methods

Data inizio appello

08/05/2014;

Disponibilità

parziale

Riassunto analitico

Two key steps in planning a well position are the so-called shallow hazard assessment and well site analysis. These operations aim to delineate the seafloor and the shallow subsurface geologic characteristics that may influence drilling operations. Such characteristics include both seafloor geological hazards (i.e., fault scarps, gas vents, unstable slopes and reefs) and shallow subsurface geological hazards (i.e., faults, gas charged sediments, buried channels and abnormal pressure zones).

On one hand, a knowledge of the seabed characteristics is essential to identify safe zones for installing underwater structures, such as drilling platforms. On the other hand, abnormal pore pressure identification is a key issue in planning and drilling an exploration well. In fact, during drilling operations, it is important that the pore fluid pressure be controlled; otherwise, kicks and, potentially, blowouts may occur when the drill bit penetrates overpressured formations. It is also necessary to ensure that the mud pressure is not too high because drilling fluids can flow into permeable formations, causing severe damage to the formation permeability.

This thesis assesses the limits and prediction capabilities of the Amplitude Versus Angle (AVA) and the Full-Waveform Inversion (FWI) methods in evaluating seabed properties and predicting overpressured zones.

The AVA method exploits the variation in seismic reflection amplitude with increasing distance between shot points and receivers. These variations indicate differences in seismic velocities and densities at the reflecting interface, which can then be transformed into petrophysical properties, such as saturation, porosity, and so on.

The FWI technique aims to produce high-resolution quantitative models of the subsurface by minimizing the difference between observed and modeled seismic waveforms. This approach goes beyond refraction and reflection tomography techniques, which use only the travel-time kinematics of the seismic data, by exploiting the additional information provided by the amplitude and phase of the seismic waveform.

Throughout this thesis, rock physics theories will be extensively used. On one hand, these theories allow us to build realistic geological models (in terms of the elastic properties) from petrophysical characteristics. On the other hand, rock physics theories serve to predict how fluid saturation or pressure variations will affect the seismic velocities and rock density. Therefore, from our viewpoint, these rock physics theories provide connections between seismic responses and the intrinsic properties of rocks, such as their mineralogy, porosity, pore shapes, pore fluids, pore pressures, permeability, viscosity and stress conditions.

For the seabed characterization, I first analyze the classical linear AVA inversion, which is based on linear simplifications of the Zoeppritz equations and a limited range of incidence angles. We will see that these linearized approximations are not useful for estimating the shear velocity contrast occurring at a fluid-solid interface (such as a water-seabed interface). Nevertheless, this shear wave velocity contrast is essential for obtaining a useful and complete description of seabed characteristics. To this end, I consider the linearized AVA inversion in which the full Zoeppritz equation is iteratively linearized around an initial guess of the true model, making use of the Gauss-Newton method. The advantages and drawbacks of this method will be demonstrated using both synthetic and field data. We will see that independently of the method being used, the short-offset reflections do not provide enough information about the shear wave velocity of the seabed. Moreover, for linear-, short-offset methods, we will see that the prediction capability and stability are strongly influenced by the background Vp/Vs ratio. I find that for these methods, the estimation of the shear wave velocity is a hopelessly non-unique problem in the case of high Vp/Vs values (as in the seabed case).

For a more reliable prediction of the elastic seabed characteristics, I assess the feasibility of exploiting the long-offset information in an AVA and/or PVA (Phase Versus Angle) inversion. For this purpose, instead of the classical plane-wave Zoeppritz formula, the spherical wave equation will be considered. In fact, theoretically, this equation is best suited for long-offset inversion because it accounts for the spherical wave effects that become significant near and beyond the critical angle. The limits of the long-offset AVA and PVA inversions as well as the opportunities in considering long-offset reflections for a reliable Vp/Vs estimation lead us to consider the FWI method. In this thesis, I consider an elastic 1D FWI in which the optimization process is driven by a genetic algorithm stochastic method. This technique is then broadly tested on both synthetic and field seismic data.

For the pore pressure prediction, I focus on an overpressured “step”, namely, a confined overpressured layer that produces a subtle change in pressure conditions. We first study the effects of pore pressure on the petrophysical rock properties and seismic velocities, using theoretical rock physics theories and both experimental and empirical equations. For the AVA method, I introduce a new AVA attribute that may identify the velocity anomalies that are generally related to an overpressured layer. For a more quantitative pressure evaluation, I analyze the limits and the accuracy of a non-linear petrophysical AVA inversion aimed at evaluating the pore pressure value. Once again, the limits of AVA inversion lead us to introduce FWI methods for detecting the abnormally high Vp/Vs values that are usually related to overpressured layers.

In my PhD studies, I have widely compared and applied stochastic search algorithms (partial results of these tests are discussed in the first appendix). Moreover, I have attempted to combine the rapid convergence of the genetic algorithm approach with the accuracy of a Markov Chain Monte Carlo algorithm, in order to cast the genetic algorithm method in a Bayesian framework. This approach allows us to obtain an unbiased estimation of the uncertainties affecting the final solution resulting from a stochastic inversion.

On one hand, a knowledge of the seabed characteristics is essential to identify safe zones for installing underwater structures, such as drilling platforms. On the other hand, abnormal pore pressure identification is a key issue in planning and drilling an exploration well. In fact, during drilling operations, it is important that the pore fluid pressure be controlled; otherwise, kicks and, potentially, blowouts may occur when the drill bit penetrates overpressured formations. It is also necessary to ensure that the mud pressure is not too high because drilling fluids can flow into permeable formations, causing severe damage to the formation permeability.

This thesis assesses the limits and prediction capabilities of the Amplitude Versus Angle (AVA) and the Full-Waveform Inversion (FWI) methods in evaluating seabed properties and predicting overpressured zones.

The AVA method exploits the variation in seismic reflection amplitude with increasing distance between shot points and receivers. These variations indicate differences in seismic velocities and densities at the reflecting interface, which can then be transformed into petrophysical properties, such as saturation, porosity, and so on.

The FWI technique aims to produce high-resolution quantitative models of the subsurface by minimizing the difference between observed and modeled seismic waveforms. This approach goes beyond refraction and reflection tomography techniques, which use only the travel-time kinematics of the seismic data, by exploiting the additional information provided by the amplitude and phase of the seismic waveform.

Throughout this thesis, rock physics theories will be extensively used. On one hand, these theories allow us to build realistic geological models (in terms of the elastic properties) from petrophysical characteristics. On the other hand, rock physics theories serve to predict how fluid saturation or pressure variations will affect the seismic velocities and rock density. Therefore, from our viewpoint, these rock physics theories provide connections between seismic responses and the intrinsic properties of rocks, such as their mineralogy, porosity, pore shapes, pore fluids, pore pressures, permeability, viscosity and stress conditions.

For the seabed characterization, I first analyze the classical linear AVA inversion, which is based on linear simplifications of the Zoeppritz equations and a limited range of incidence angles. We will see that these linearized approximations are not useful for estimating the shear velocity contrast occurring at a fluid-solid interface (such as a water-seabed interface). Nevertheless, this shear wave velocity contrast is essential for obtaining a useful and complete description of seabed characteristics. To this end, I consider the linearized AVA inversion in which the full Zoeppritz equation is iteratively linearized around an initial guess of the true model, making use of the Gauss-Newton method. The advantages and drawbacks of this method will be demonstrated using both synthetic and field data. We will see that independently of the method being used, the short-offset reflections do not provide enough information about the shear wave velocity of the seabed. Moreover, for linear-, short-offset methods, we will see that the prediction capability and stability are strongly influenced by the background Vp/Vs ratio. I find that for these methods, the estimation of the shear wave velocity is a hopelessly non-unique problem in the case of high Vp/Vs values (as in the seabed case).

For a more reliable prediction of the elastic seabed characteristics, I assess the feasibility of exploiting the long-offset information in an AVA and/or PVA (Phase Versus Angle) inversion. For this purpose, instead of the classical plane-wave Zoeppritz formula, the spherical wave equation will be considered. In fact, theoretically, this equation is best suited for long-offset inversion because it accounts for the spherical wave effects that become significant near and beyond the critical angle. The limits of the long-offset AVA and PVA inversions as well as the opportunities in considering long-offset reflections for a reliable Vp/Vs estimation lead us to consider the FWI method. In this thesis, I consider an elastic 1D FWI in which the optimization process is driven by a genetic algorithm stochastic method. This technique is then broadly tested on both synthetic and field seismic data.

For the pore pressure prediction, I focus on an overpressured “step”, namely, a confined overpressured layer that produces a subtle change in pressure conditions. We first study the effects of pore pressure on the petrophysical rock properties and seismic velocities, using theoretical rock physics theories and both experimental and empirical equations. For the AVA method, I introduce a new AVA attribute that may identify the velocity anomalies that are generally related to an overpressured layer. For a more quantitative pressure evaluation, I analyze the limits and the accuracy of a non-linear petrophysical AVA inversion aimed at evaluating the pore pressure value. Once again, the limits of AVA inversion lead us to introduce FWI methods for detecting the abnormally high Vp/Vs values that are usually related to overpressured layers.

In my PhD studies, I have widely compared and applied stochastic search algorithms (partial results of these tests are discussed in the first appendix). Moreover, I have attempted to combine the rapid convergence of the genetic algorithm approach with the accuracy of a Markov Chain Monte Carlo algorithm, in order to cast the genetic algorithm method in a Bayesian framework. This approach allows us to obtain an unbiased estimation of the uncertainties affecting the final solution resulting from a stochastic inversion.

File

Nome file | Dimensione |
---|---|

Ci sono 1 file riservati su richiesta dell'autore. |