Semi-Automatic History Matching Using the Pilot Point Method Including Time-Lapse Seismic Data

Abstract

In this paper, we evaluate the benefit of using time-lapse seismic in reservoir simulation and handle some of the associated challenges. The problem is tackled from a reservoir engineering perspective, focusing specifically on the use of time-lapse seismic in history matching. A semi-automatic history-matching procedure is presented that uses both production and time-lapse seismic data as conditioning data. The actual integration of the production and time-lapse seismic data is handled in the objective or misfit function, which expresses the misfit between modeled and observed reservoir behavior. The pilot point method was chosen to reduce the large history matching parameter space1–5. The optimum permeability values at the pilot points were determined using the Levenberg-Marquardt optimization algorithm. The semi-automated method is tested on a 2D synthetic reservoir. The results indicate that for a case with abundant production data, time-lapse seismic sometimes yields only minor improvements in the final history matched model. However, when production data is limited, time-lapse seismic improves the history-matched permeability fields considerably by correctly identifying and locating the major heterogeneities of the permeability field. In addition, the sensitivity studies show that if pilot points are located in or near heterogeneities, the automatic history-matching algorithm has a better chance of correctly locating the heterogeneities. Furthermore, a seismic noise sensitivity analysis shows that seismic noise impedes accurate determination of the history-matched parameters. The results of this research demonstrate that a history-matching procedure including time-lapse seismic can improve reservoir characterization and modeling. It can be used as a reservoir management tool to optimize in-fill well locations, enhanced oil recovery projects, and economic evaluations. The semi-automated procedure allows engineers to work more efficiently with larger reservoir models and extensive dynamic conditioning data sets. Introduction 4D seismic is the process of repeating 3D seismic surveys over a producing reservoir in time-lapse mode. It has a potentially huge impact in reservoir engineering because it is the first technique that may allow us to directly image dynamic reservoir properties such as fluid movement, pressure build-up, and heat flow in a reservoir in a true volumetric sense. Seismic images are sensitive to spatial contrasts in two distinct types of reservoir properties:non-time-varying static geology properties such as lithology, porosity, shale content; andtime-varying dynamic fluid-flow properties such as fluid saturation, pore pressure and temperature. Given a single 3D seismic survey, representing a single snapshot in time of the reservoir, the static geology and the dynamic fluid-flow contributions to the seismic image are non-uniquely coupled and therefore difficult to separate. For example, it may be impossible to distinguish an oil-water contact from a horizontal depositional boundary in a single seismic image. However, with 4D seismic surveys, examining the difference between time-lapse 3D seismic images allows the non-time-varying geologic contributions to cancel, resulting in a direct image of the time-varying changes caused by the reservoir fluid flow. Potential applications of 4D seismic include mapping fluid contacts with time, estimating pressure compartmentalization and fault seal, and locating bypassed oil.

Advanced Integrated Workflows for Incorporating Both Production and 4D Seismic-Related Data into Reservoir Models

“…However, there is still no definite matching methodology and several issues remain to be addressed. 4D seismic can be accounted for under various forms: amplitudes (Arenas et al, 2001) or acoustic impedances derived from amplitudes in a preliminary step , impedances at base and monitor times or impedance differences between base and monitor times, impedances given in depth (Gosselin et al, 2001) or time (Fornel et al, 2007; domain. All of these may significantly impact the number of data, but also their informative content.…”

Impact of time-lapse seismic data for permeability estimation

“…One may assume that the time-lapse seismic data have already been inverted into differences in elastic properties and then use these differences as data in the inversion loop for the reservoir flow properties, see, e.g., [2,4,10,15,33,35]. This effectively decouples the seismic wave propagation modeling from the inversion loop for the flow properties.…”

“…A relatively high level of uncertainty must, however, be anticipated since pseudoseismic data results from an inversion process. In previous works using pseudoseismic data, generally positive results are obtained, but in most of these works, a modest uncertainty (smaller than 10% relative uncertainty) in the pseudoseismic data is assumed, or the assumed error level is unspecified (a counter example is found in [4], where the impact of different error levels on the pseudoseismic part of the objective function is investigated).…”

Mapping of Reservoir Properties and Facies Through Integration of Static and Dynamic Data

“…The type of seismic data used in the objective function has varied among the researchers. Huang et al (1997) used amplitude difference or other seismic attributes difference while Arenas et al (2001) used velocity difference. Landa and Horne (1997) assumed that saturation changes could be obtained directly from time-lapse surveys.…”

“…Because the number of data and model parameters can be quite large in history matching problems which include time-lapse seismic data, it is common to reduce the number of model parameters by using "pilot points" (Arenas et al, 2001;van Ditzhuijzen et al, 2001) or to divide model into zones with similar properties (Huang et al, 2001).…”