Summary In this work, we propose an ensemble-based seismic history matching approach to predict reservoir properties, i.e. porosity and permeability, with uncertainty quantification, using both production and time lapse seismic data. To avoid the common underestimation of uncertainty in ensemblebased optimization approaches, and to make the computation feasible, we introduce the convolutional autoencoder to reparameterize seismic data into a lower dimensional space. We then apply the Ensemble Smoother with Multiple Data Assimilation to optimize an ensemble of reservoir models using the production and re-parameterized seismic data. The proposed methodology is tested on a 2D synthetic case. The inversion results indicate that the method can largely improve the characterization of reservoir models compared to the history-matching scenario with production data only.
In this work, we focus on a Bayesian inversion method for the estimation of reservoir properties from seismic data and we study how the inversion parameters, such as rock-physics and geostatistical parameters, can affect the inversion results in terms of reservoir performance quantities (pore volume and connectivity). We apply a Bayesian seismic inversion based on rock-physics prior modeling for the joint estimation of facies, acoustic impedance and porosity. The method is based on a Gibbs algorithm integrated with geostatistical methods that sample spatially correlated subsurface models from the posterior distribution. With the ensemble of multiples scenarios of the subsurface conditioned to the experimental data, we can evaluate two quantities that impact the production of the reservoir: the reservoir connectivity and the connected pore volume. For each set of parameters, the inversion method yields different results. Hence, we perform a sensitivity analysis for the main parameters of the inversion method, in order to understand how the subsurface model may be influenced by erroneous assumptions and parameter settings.
Presentation Date: Monday, October 15, 2018
Start Time: 1:50:00 PM
Location: 206A (Anaheim Convention Center)
Presentation Type: Oral
Summary In this work, we propose a stochastic nonlinear inversion framework for PP and PS seismic data based on the ensemble smoother with multiple data assimilations (ES-MDA) to estimate elastic reservoir properties with uncertainty quantification. The ES-MDA is an iterative ensemble-based data assimilation method that generates an ensemble of solutions of the inverse problem. In our approach, it is applied to a seismic inversion problem in which the full Zoeppritz equations, without linearization, are used to improve the inversion accuracy. The ensemble of updated reservoir realizations obtained by assimilating seismic data allows evaluating the associated model uncertainty. To avoid the model uncertainty be underestimated in the ensemble-based approach, we propose to apply the ES-MDA in a lower-dimensional data space obtained by the re-parameterization of PP and PS seismic data using the singular value decomposition (SVD).
The variations of dynamic reservoir properties cause the change in seismic response. During the production phase, time-lapse seismic data can be used to monitor water saturation and pressure changes. The prediction of water saturation and pressure conditions from seismic datarequires physical model to link their changes to variations in elastic properties. The empirical models commonly used constant empirical coefficients in the reservoir. However, in first part of the work, we show that different porosity, saturation, and pressure in in-situ conditions can affect the model coefficients. We then propose a new rock physics model to compute the changes in reflectivity due to thechanges in saturation and pressure, accounting in-situ reservoir conditions. The model is then integrated in a Bayesian inversion method to predict water saturation and pressure changes directly from the amplitude difference of time-lapse seismic data. We apply the proposed method to a synthetic dataset and obtain accurate results.
Presentation Date: Tuesday, October 16, 2018
Start Time: 8:30:00 AM
Location: 209A (Anaheim Convention Center)
Presentation Type: Oral
In this paper we propose a new workflow to perform Petrophysical Joint Inversion (PJI) of surface to surface seismic and Controlled Source ElectroMagnetic (CSEM) data, to recover reservoir properties (clay volume, porosity and saturation). Seismic and CSEM measurements provide independent physical measurements of subsurface that complement each other. In the case of well-logs, the basis of the PJI training dataset, taking advantage of such complementarity is straightforward. Indeed, elastic and electric measurements of earth properties sense the same earth volume at much the same scale. When applying the training dataset to the surface data derived geophysical attributes, the order of magnitude gap in between the scale at which those elastic and electric attributes represent the earth undermines dramatically PJI validity. Various CSEM inversion constraining methods (regularization breaks, prejudicing, use of an a priori model etc) help to reconcile seismic and CSEM resolution, but they are usually proven to be insufficient or inaccurate. In addition to these methods, we suggest adding a further downscaling step, so the recovered electric attribute resolution can be adequate with respect to the seismic one, hence fit for purpose. Such downscaling is designed to be consistent in electrical attribute space via transverse resistance within a rockphysics framework. The workflow will be demonstrated on a case study.
Data integration is a key step in reservoir modeling and characterization. If done properly, the integration of different sources of information about the subsurface petro-elastic properties of interest allow a better description of the reservoir while accounting for existing uncertainties. Under this scope, geostatistical seismic inversion techniques have been successfully applied to integrate seismic reflection and well data for seismic reservoir characterization. Depending on the available seismic reflection data, these geostatistical modeling techniques allow inferring acoustic and/or elastic spatial distributions of the the subsurface. The resulting elastic models are then used as secondary variables to infer the petrophysical properties (e.g. facies, porosity) of the reservoir. Consequently, these petrophysical models are not directly constrained by the existing seismic reflection data and the uncertainties related with the seismic inversion problem are not properly propagate during the entire geo-modeling procedure. This work introduces a framework to incorporate rock physics modeling within conventional pre-stack iterative geostatistical seismic inversion methodologies. The proposed technique allow inverting from pre-stack seismic reflection data directly for facies, porosity and pore fluid. This technique is based on three main concepts: i) the model parameter space is perturbed using stochastic sequential simulation and co-simulation; ii) a statistical rock physics modeling technique is used to link the elastic and petrophysical domain within the inversion procedure; and iii) the use of a global optimizer based on cross-over genetic algorithms driven by the mismatch between real and synthetic seismic reflection from iteration to iteration. This work illustrates the successful application of the proposed iterative geostatistical seismic inversion technique to a real dataset. The retrieved petro-elastic models are simultaneously, and consistently, conditioned by the well-log data, seismic reflection data and the reservoir geology (i.e. pore geometry) as expressed by a rock physics model.
Summary In this work, we develop an innovative stochastic nonlinear inversion method based on the Ensemble Smoother algorithm and data re-parameterization to estimate elastic as well as rock and fluid properties from seismic. The Ensemble Smoother is an iterative stochastic optimization method and in our work, it is applied to simultaneously update an ensemble of prior geostatistical models until a satisfactory match between the predicted seismic response and the measured seismic data is achieved. The model uncertainty is assessed from the multiple equiprobable posterior realizations. The proposed method is appropriate for nonlinear inverse problems, and it can be applied to the prestack AVA inversion and petrophysical inversion based on the exact Zoeppritz equations and nonlinear rock physics models. Yet, the proposed optimization method typically underestimates the uncertainty of the inverted results.
One of the emerging technologies in geophysics is the stochastic inversion of geophysical data for the prediction of rock and fluid properties. The probability distribution of the geophysical properties of interest can be computed using a probabilistic inverse method. The integration of stochastic inverse methods and geophysical modeling allows generating multiple reservoir models of rock and fluid properties that honor the geophysical measurements. Stochastic approaches allow sampling multiple solutions from the posterior distribution of the model parameters and quantifying the uncertainty in the model parameter predictions. Stochastic inversion algorithms can be applied to seismic inversion problems as well as petrophysical inversion problems. In this work, we discuss analytical and numerical approaches, as well as their advantages and disadvantages.
Presentation Date: Monday, September 25, 2017
Start Time: 1:50 PM
Presentation Type: ORAL
The goal of seismic reservoir characterization is to estimate rock and fluid properties from seismic data. The solution of the seismic inverse problem is based on the forward models that describe physical relationships between lithology and fluid parameters and their seismic response. If the forward models indicate nonlinear physics relations between the variables, advanced inversion methods, such as stochastic optimization algorithms, should be adopted to predict the reservoir properties. However, these methods are generally time consuming. In this paper, we propose an inversion approach that combined linearized AVO modeling and linearized rock physics relations. The rock physics model is based on Nur's critical porosity model and Gassmann's equations and its linearization is based on first-order Taylor series approximation. The linearization is computed with respect to solid elastic moduli, solid density, fluid bulk modulus, fluid density and porosity. Indeed, the rock physics model is almost linear in these properties and the linearization of the model provides a good approximation. The combined forward model is then used in a Bayesian inversion workflow for the estimation of the above-mentioned rock and fluid properties from pre-stack seismic data and well logs. In the Bayesian inversion, we assume the prior model and the error term to be distributed as Gaussian distribution to derive the analytical solution of the Bayesian inverse problem. The reservoir properties of interest, i.e. porosity, mineral volumes, and fluid saturations, are then computed from the inversion results. The proposed method was first validated on a synthetic dataset and then applied to a field dataset with satisfactory inversion results.
Presentation Date: Tuesday, September 26, 2017
Start Time: 8:55 AM
Presentation Type: ORAL
Summary In this work, we propose a seismic inversion method for the joint estimation of facies and elastic velocities from pre-stack seismic data based on a geostatistical approach. The objective of the propos ed inversion methodology is to obtain the posterior distribution of P-wave velocity, S-wave velocity and density and to simultaneously classify the lithology conditioned by seismic data. The inversion algorithm is a sequential Gaussian mixture inversion developed based on Bayesian linearized AVO inverse theory and sequential geostatistical simulations. To mathematically represent the multimodal behavior of elastic properties due to their variations within different facies, we adopt a Gaussian mixture distribution for the prior model of the elastic properties and use the prior probability of the facies as weights of the Gaussian components of the mixture . The solution of the inverse problem is achieved by deriving the explicit analytic al expression for the posterior distribution of the elastic properties and facies.