Iop Publishing Journal of Geophysics and Engineering

78th EAGE Conference and Exhibition 2016, 2016

Seismic inversion of under-sampled reservoirs at early exploration stages is still a research challenge, given the lack, or total absence, of data for the inversion: acoustic impedance distributions, wavelet etc. In this study the available information is a 2D seismic line, the final section of interval velocity obtained from processing and seismic horizons resulting from the interpretation. No well information is available in the area of study.

1999

Information about reservoir properties usually comes from two sources: seismic data and well logs. The former provide an indirect, low resolution image of rock velocity and density. The latter provide direct, high resolution (but laterally sparse) sampling of these and other rock parameters. An important problem in reservoir characterization is how best to combine these data sets, allowing the well information to constrain the seismic inversion and, conversely, using the seismic data to spatially interpolate and extrapolate the well logs. We have developed a seismic/well log inversion method that combines geostatistical methods for well log interpolation (i.e., kriging) with a Monte Carlo search technique for seismic inversion. Our method follows the approach used by Haas and Dubrule (1994) in their sequential inversion algorithm. Kriging is applied to the well data to obtain velocity estimates and their variances for use as a priori constraints in the seismic inversion. Further, inversion of a complete 2-D seismic section is performed one trace at a time. The velocity profiles derived from previous seismic traces are incorporated as "pseudo well logs" in subsequent applications of kriging. Our version of this algorithm employs a more efficient Monte Carlo search algorithm in the seismic inversion step, and moves progressively away from the wells so as to minimize the kriging variance at each step. Numerical experiments with synthetic data demonstrate the viability of our seismic/well data inversion scheme.

All Days, 2005

Integrating different types of data having different scales is the major challenge in reservoir characterization studies. Seismic data is among those different types of data, which is usually used by geoscientists for structural mapping of the subsurface and making interpretations of the reservoir's facies distribution. Yet, it has been a common aim of geoscientists to incorporate seismic data in high-resolution reservoir description through a process called seismic inversion. Using geostatistical models in this kind of studies becomes insufficient in dealing with the uncertainty and the non-linearity, because of the stationarity assumption of variogram models. As an alternate, soft computing has been widely used in reservoir characterization, as a method which is tolerant of uncertainty, imprecision, and partial truth. In this study, a new intelligent seismic inversion methodology is presented to achieve a desirable correlation between relatively low-frequency seismic signals, and the much higher frequency wireline-log data. Vertical seismic profile (VSP) is used as an intermediate step between the well logs and the surface seismic. A synthetic seismic model is developed by using real data and seismic interpretation. This model represents the Atoka and Morrow formations, and the overlying Pennsylvanian sequence of the Buffalo Valley Field in New Mexico. Generalized regression neural network (GRNN) is used to build two independent correlation models between; 1) Surface seismic and VSP, 2) VSP and well logs. After generating virtual VSP's from the surface seismic, well logs are predicted by using the correlation between VSP and well logs. The values of the density log, which is a surrogate for reservoir porosity, are predicted for each seismic trace through the seismic line with a classification approach, having a correlation coefficient of 0.81. The same methodology is then applied to real data taken from the Buffalo Valley Field, to predict interwell gamma ray logs and neutron porosity logs through the seismic line of interest. The same procedure can be applied to a complete 3D seismic block to obtain 3D distributions of reservoir properties with less uncertainty than the geostatistical estimation methods, which would hopefully help to increase the success of drilling new wells during field development. To my mother, Nilgün; my father, Bilsel; and my sister, Selin. iii First of all, I would like to express my sincere gratitude and appreciation to my advisor, Prof. Shahab Mohaghegh, for his endless support and guidance during my studies. He has not been only a great advisor, but also a mentor and a friend to me, and wherever I go through out my future career, I will carry on the fundamentals and professionalism that he taught me. I would like to extend my appreciation to Prof. Jaime Toro, Prof. Tom Wilson, and Alejandro Sanchez, for their cooperation to make this multidisciplinary work achieve its goals. I am also so grateful to my committee members; Prof. Sam Ameri, Dr. Razi Gaskari, and Dr. Grant Bromhal, for their support and valuable suggestions. I would also like to thank Prof. Ilkin Bilgesu, for his friendship and support during my stay in Morgantown. Mrs. Ruth Long, Prof. Erdogan Gunel, and Mrs. Janis Gunel were like a second family to me, and I am thankful for that. I feel very lucky that I have such wonderful friends around me. I have shared many great moments that will always be remembered with my colleagues and office mates; Jalal Jalali,

Waveform inversion methods can be used to obtain high-resolution images of the elastic and acoustic property changes of petroleum reservoirs under production, but remains computationally challenging. Efficient approximations in modelling the wavefield based on scattering approaches are desirable for solving time-lapse inversion problems and to test the settings in which they give accurate predictions. In this thesis, we are concerned with acoustic waveform modelling and inversion in frequency domain with emphasis on first-order scattering methods. Key themes of review and discussion include the derivation of scattering problems using Green's function techniques; Born approximation, distorted-Born approximation, Distorted Born iterative T-matrix method (DBIT); and the concepts of seismic waveform inversion and its application to different approaches in time-lapse seismic imaging. We employ the Born approximation and distorted-Born to simulate time-lapse synthetic seismograms and test the settings in which these methods are valid by benchmarking them against the exact T-matrix approach. The new distorted-Born approximation presented considers a general heterogeneous reference medium and provides a framework for imaging of regions of time-lapse variation using the baseline survey as a reference and the monitor survey as perturbed to directly estimate the perturbation. This poses a linear inverse scattering problem for which suitable linear and non-linear inversion methods are applicable. Synthetic testing based on different 2D models demonstrate that the new distorted-Born approximation provides accurate predictions of the difference data seismograms, at least in the settings considered while the Born approximation is limited to applications involving small volume and velocity contrasts. The inversion results show that Born inversion (in its limits) and DBIT method sufficiently retrieves the time-lapse velocity changes even in the cases of relatively low signal to noise ratio. Inversion of the difference data (differential approach) not only gives improved results but also proves to be useful and efficient since a single inversion procedure is performed for a pair of seismic experiments and hence computationally less expensive. The DBIT method which considers a dynamic background media and a variational T-matrix approach may be very useful in seismic characterisation of petroleum reservoirs under production and may be more efficient in monitoring of CO2 sequestration.

2000

Information about reservoir properties usually comes from two sources: seismic data and well logs. The former provide an indirect, low resolution image of rock velocity and density. The latter provide direct, high resolution (but laterally sparse) sampling of these and other rock parameters. An important problem in reservoir characterization is how best to combine these data sets, allowing the well information to constrain the seismic inversion and, conversely, using the seismic data to spatially interpolate and extrapolate the well logs. We develop a seismic/well log inversion method that combines geostatistical techniques for well log interpolation (Le., kriging) with a Monte Carlo search method for seismic inversion. We cast our inversion procedure in the form of a Bayesian maximum a posteriori (MAP) estimation in which the prior is iteratively modified so that the algorithm converges to the model that maximizes the likelihood function. We follow the approach used by Haas and Dubrule (1994) in their sequential inversion algorithm. Kriging is applied to the well data to obtain velocity estimates and their covariances for use as a priori constraints in the seismic inversion. Inversion of a complete 3-D seismic section is performed one trace at a time. The velocity profiles derived from previous seismic traces are incorporated as "pseudo well logs" in subsequent applications of kriging. Our version of this algorithm employs a more efficient Monte Carlo search method in the seismic inversion, and moves sequentially away from the wells so as to minimize the kriging variance at each step away from the inverted wells. Numerical experiments with synthetic data demonstrate the viability of Our seismic/well data inversion scheme. Inversion is then performed on a real 3-D data set provided by Texaco.

SEG Technical Program Expanded Abstracts 2009, 2009

This research aims at making optimal updates of geological models by jointly inverting flow and seismic data while honoring the geologic spatial continuity. Numerical models for reservoir characterization are increasing in complexity, due in part to the greater need to model the complex spatial heterogeneity and fluid flow in the subsurface. These models, once properly calibrated, can make better forecasts. This calibration process requires in essence the solving of an inverse problem. The inversion problem is formulated as minimizing the mismatch function between observations and the output of the numerical models. The optimal search is carried out by adjusting model parameters, typically one or more for each grid-point of the reservoir. The optimization problem is large-scale in nature, with a nonlinear and nonconvex objective function, that often involves time-expensive simulations. Additionally, this problem is generally ill-conditioned, because the number of degrees of freedom usually is larger than the number of observations. We present a robust and fairly efficient methodology to deal with these difficulties in the framework of oil reservoir characterization. The illconditioned character of the optimal search can be attenuated in two ways. By Principal Component Analysis (PCA) the search space can be projected to a subspace of much smaller dimension, while keeping consistency with prior spatial geological features already known for the reservoir. The number of optimal solutions can be reduced further by increasing the diversity of the data observed. We integrate two different types of data: time-lapse seismic (spatially distributed and of lower temporal periodicity) and production data (localized around wells and of high temporal periodicity). Production data provides an integrated response of the reservoir to fluid flow, while time-lapse seismic data yields a spatially distributed characterization of the changes in elastic velocities due to saturation and pressure variations. The reduction in the number of optimization variables by PCA allows the use of numerical derivatives of the cost function. Within a distributed computing framework these approximate derivatives can be calculated efficiently. We also consider derivative-free algorithms. We illustrate the methodology on a sector of the Stanford VI synthetic reservoir created for testing algorithms.

SEG Technical Program Expanded Abstracts 2003, 2003

A stochastic model is developed to estimate porosity ( ) and water saturation (S w ) using multiple sources of information, including borehole and S w measurements, seismic P-and Swave travel time, and inverted electrical conductivity ( ). Within the stochastic framework, both reservoir parameters and geophysical attributes at unsampled locations in the interwell volumes are considered as random variables and are estimated simultaneously by conditioning to available data. The focus of the inversion process is not on finding a bestfitting solution of the unknown parameters as with conventional methods, but rather on sampling from the joint probability density function of the unknowns. From the samples, various statistics of each variable can be inferred, such as mean, variance, confidence intervals, and even probability density functions. A synthetic case study based on well log data is presented. Results show that the stochastic inverse provides not only overall better estimates of and S w but also information about uncertainty in the estimation that cannot be obtained using conventional methods.

GEOPHYSICS, 2009

Hydrocarbon reservoirs are characterized by seismic, well-log, and petrophysical information, which is dissimilar in spatial distribution, scale, and relationship to reservoir properties. We combine this diverse information in a unified inverse-problem formulation using a multiproperty, multiscale model, linking properties statistically by petrophysical relationships and conditioning them to well-log data. Two approaches help us: (1) Markov-chain Monte Carlo sampling, which generates many reservoir realizations for estimating medium properties and posterior marginal probabilities, and (2) optimization with a least-squares iterative technique to obtain the most probable model configuration. Our petrophysical model, applied to near-vertical-anglestackedseismic data and well-log data from a gas reservoir, includes a deterministic component, based on a combination of Wyllie and Wood relationships calibrated with the well-log data, and a random component, based on the statistical charact...

Journal of Geophysics and Engineering, 2020

Time-lapse (4D) seismic inversion aims to predict changes in elastic rock properties, such as acoustic impedance, from measured seismic amplitude variations due to hydrocarbon production. Possible approaches for 4D seismic inversion include two classes of method: sequential independent 3D inversions and joint inversion of 4D seismic differences. We compare the standard deterministic methods, such as coloured and model-based inversions, and the probabilistic inversion techniques based on a Bayesian approach. The goal is to compare the sequential independent 3D seismic inversions and the joint 4D inversion using the same type of algorithm (Bayesian method) and to benchmark the results to commonly applied algorithms in time-lapse studies. The model property of interest is the ratio of the acoustic impedances, estimated for the monitor, and base surveys at each location in the model. We apply the methods to a synthetic dataset generated based on the Namorado field (offshore southeast Br...

73rd EAGE Conference and Exhibition - Workshops 2011, 2011

During the last couple of decades, there have been great advances in seismic inversion and lithology/fluid prediction. In the last few years, we have seen breakthroughs in integration of seismic methods, rock physics, spatial statistics and reservoir geology, allowing for more robust and realistic predictions of reservoir parameters from seismic data. The greatest advances have been made in academia and at research centers; now is the time to implement recent technologies into the workflows of the oil industry, in order to reduce exploration risk in frontier areas and boost oil recovery in existing fields. In this presentation, we give an overview of major breakthroughs in the last decade, and we suggest further extensions on how to integrate seismic inversion, rock physics, spatial statistics and geological knowledge during seismic reservoir prediction. In particular, we demonstrate that the uncertainties in lithology/fluids predictions can be reduced if geological trends are included as constraints in the inversion model. The lithology/fluid classification is constrained by depth trends and a Markov random field prior model for spatial coupling of the discrete lithology/fluids classes. A Bayesian method then combines seismic data, well observations, and prior information to predict lithology/fluid classes with associated uncertainties. The inversion approach is evaluated on a real case from the North Sea. The prior Markov random field makes it possible to identify complex structures in the lithology/fluid characteristics, by improving spatial continuity laterally. Furthermore, the posterior estimates of lithologies and fluids are used to constrain the estimation of continuous reservoir parameters like porosity. By honoring spatial continuity and vertical transitions in lithologies and fluids, we obtain sharper porosity sections and unravel details in the data not detectable from more conventional inversion algorithms. Our results also show better match with well log porosities than direct estimates of porosities from elastic parameters.