Mazuyer, Antoine (Universite de Lorraine- GeoRessources) | Giot, Richard (Universite de Lorraine- GeoRessources) | Cupillard, Paul (Universite de Lorraine- GeoRessources) | Conin, Marianne (Universite de Lorraine- GeoRessources) | Thore, Pierre (Total SA)
The aim of this study is to estimate the initial stress in reservoirs before production using 3D calibrated geomechanical models. We propose an inverse method for estimating stress. Wellbore data can be integrated in a Mechanical Earth Model in order to estimate stresses nearby wells. It yields a first rough estimation in the whole reservoir by a simple interpolation which is not in equilibrium with the external forces and boundary conditions. From this rough stress field, the inversion aims at finding a physically acceptable stress state (i.e.: in equilibrium with the external forces and boundary conditions) that fit the local stresses wells. The forward problem is ensured by a Finite Element Analysis which is able to take into account structures such as faults, which have a significant influence on the stress magnitude and orientation. Inverse loop stops when the stress computed near wells matches the one estimated using borehole data. The uncertainties on the boundary conditions, elastic parameters and the first stress estimation are taken into account with a stochastic approach. In this study, faults are built with a volumetric representation of the core and damage zone by introducing elastic parameters variations within. This representation is possible because only small deformations are expected.
Subsoil raw materials exploitation generally induces stress changes. For instance, depletion disturbs the mechanical equilibrium, yielding a stress change in the reservoir and in the overlying geological layers (the overburden). It could have dramatic consequences such as borehole instability (Zoback et al., 2003) and fault reactivation, which can lead to unexpected oil and gas leaks or migration (Wiprut and Zoback, 2002).
Stress estimation is important during field exploitation to avoid these problems and to anticipate measures to stabilize wells using different drilling and casing techniques (Wilson et at., 1999). Depletion can be monitored and it is possible to estimate the relative stress change during the exploitation. The goal of this paper is to introduce a new approach to estimate the absolute stress in the reservoir and the overburden before exploitation.
We have performed a set of inversions on a thinly bedded clastic reservoir from the West of Shetland. The inversion techniques comprise a data-driven deterministic inversion (contractor), a data-driven stochastic inversion (contractor), a model-driven deterministic inversion (proprietary), and a model-driven stochastic inversion (proprietary). The variability between the results obtained from these different inversion techniques is far greater than the variability between the realizations obtained from a single stochastic inversion. This shows that the main uncertainty associated with the seismic inversion is the parameterization of the inversion itself.
Asnaashari, Amir (Université Joseph Fourier Grenoble) | Brossier, Romain (Université Joseph Fourier Grenoble) | Garambois, Stéphane (Université Joseph Fourier Grenoble) | Virieux, Jean (Université Joseph Fourier Grenoble) | Audebert, François (TOTAL E&P) | Thore, Pierre (TOTAL E&P)
Summary I compare different sets of solutions obtained with three stochastic algorithms to a highly multimodal inverse problem: the seismic to well tying of Time-Lapse Data. The problem is to find the perturbations of velocity and density in a layered model which explain the differences between the base and the monitor surveys. Several almost equivalent solutions exist especially if the thicknesses of the layers are thin. To explore the solution space I used three stochastic based optimizers: Simulated Annealing, Genetic Algorithms and Covariance Matrix Adaptation Evolution Strategy. In this paper, I compare the results obtained by the different algorithms not only in terms of speed and fitness but also in the way they offer the widest range of acceptable realizations.
Summary We present an analysis of the Kurtosis method used to estimate the phase of the wavelet without well log information. The kurtosis is a high order statistics which preserves the phase information. This paper aims to test the reliability of the Kurtosis method based on synthetic and real seismic data. We have addressed in our tests a number of factors (e.g. Our observations on the real data examples show that the phase estimated by this technique is influenced by the geology.
Summary The approach presented here is a fast track method for extracting 4D signal in complex media. It is based on a transformation which maps the data from the real space to a pseudo 1D space where the current 4D inversion techniques are still valid.
We present a new technique for inverting 4D seismic data constrained by dynamics and geology. The inversion is first performed at well positions where all the constraints are set and afterwards extended to the full 3D dataset. The geological and dynamical constraints are set in the model definition i.e. a layered description of the geology (with permeable and non permeable layers) which may be different at each well. This information is then propagated concurrently from each well to the whole dataset. The way the inversion is posed prevents from side lobes effect and enables to discriminate density and velocity effects (P in the case of post-stack data and P&S in the case of prestack). The more reliable information is the P velocity since it affects both reflectivity and travel time.
Asnaashari, Amir (Université Joseph Fourier & CNRS) | Brossier, Romain (Université Joseph Fourier & CNRS) | Garambois, Stéphane (Université Joseph Fourier & CNRS) | Audebert, François (TOTAL E&P) | Thore, Pierre (TOTAL E&P) | Virieux, Jean (TOTAL E&P)