Quantitative application of 4D seismic data for updating thin-reservoir models
MetadataShow full item record
A range of methods which allow quantitative integration of 4D seismic and reservoir simulation are developed. These methods are designed to work with thin reservoirs, where the seismic response is normally treated in a map-based sense due to the limited vertical resolution of seismic. The first group of methods are fast-track procedures for prediction of future saturation fronts, and reservoir permeability estimation. The input to these methods is pressure and saturation maps which are intended to be derived from time-lapse seismic attributes. The procedures employ a streamline representation of the fluid flow, and finite difference discretisation of the flow equations. The underlying ideas are drawn from the literature and merged with some innovative new ideas, particularly for the implementation and use. However my conclusions on the applicability of the methods are different from their literature counterparts, and are more conservative. The fast-track procedures are advantageous in terms of speed compared to history matching techniques, but are lacking coupling between the quantities which describe the reservoir fluid flow: permeabilities, pressures, and saturations. For this reason, these methods are very sensitive to the input noise, and currently cannot be applied to the real dataset with a robust outcome. Seismic history matching is the second major method considered here for integrating 4D seismic data with the reservoir simulation model. Although more computationally demanding, history matching is capable of tolerating high levels of the input noise, and is more readily applicable to the real datasets. The proposed implementation for seismic modelling within the history matching loop is based on a linear regression between the time-lapse seismic attribute maps and the reservoir dynamic parameter maps, thus avoiding the petro-elastic and seismic trace modelling. The idea for such regression is developed from a pressure/saturation inversion approach found in the literature. Testing of the seismic history matching workflow with the associated uncertainty estimation is performed for a synthetic model. A reduction of the forecast uncertainties is observed after addition of the 4D seismic information to the history matching process. It is found that a proper formulation of the covariance matrices for the seismic errors is essential to obtain favourable forecasts which have small levels of bias. Finally, the procedure is applied to a North Sea field dataset where a marginal reduction in the prediction uncertainties is observed for the wells located close to the major seismic anomalies. Overall, it is demonstrated that the proposed seismic history matching technique is capable of integrating 4D seismic data with the simulation model and increasing confidence in the latter.