Iterative History Matching Technique for Estimating Reservoir Parameters from Seismic Data

High uncertainties associated with unknown fluids and pressure distributions could hamper the history matching process. Accordingly, efforts have been made to evaluate these properties distributions from static and dynamic seismic attributes. Nevertheless, these new methodologies aim to develop quantitatively seismic data integration, which require a suitable association between Petro-Elastic and Flux Models parameters dealing with the inherent inter-dependence between simulation and rock physics parameters. In this paper it is presented an iterative workflow between two optimization processes. Firstly, a coupling between a PetroElastic Model and a reservoir simulator provides Acoustic Impedance (AI) values for two numerical models that have to be matched. The one that has to be honored provides the synthetic “observed” AI data and the other gives the calculated AI values. An objective function based on these is minimized through a Pre-Conditioned Conjugate-Gradient method under reservoir simulation constraints, in order to achieve saturation and pressure trends. As a second step, a history matching procedure where the saturation/pressure values derived from the former one are now the models to be honored. Then, this information is combined with flux lines in the parameterization phase delineating reservoir heterogeneities to define regions where permeability objective functions are optimized. The saturation/pressure trends, evaluated by this last step, are used to update the initial inputs at the first one, providing history matching constraints to the overall processes. This technique allowed to determine faults and permeability channels with less uncertainties and higher accuracy and decreasing in time-consuming of the history matching procedure was obtained. Moreover, this methodology improved the reservoir behavior predictions reliability, through quantitative seismic data integration in the history matching process. Certainly the major goals are the integration aspects of this proposed methodology combined with the obtained improvements in the history matching process. Introduction Simulation models have been used to estimate reservoir behaviour. The model is generally built from static geomodels often constrained to log and core data from wells in addition to pre-production seismic. The models are then modified to match static and dynamic well data, including fluid production rates and pressures. However, the non-uniqueness of the models can hamper the process due to missing information, particularly regarding to changes in the fluid distributions and pressures between wells. Comparison between 3D surveys can identify fluid and pressure changes. Thus, time-lapse seismic is a valuable tool to improve reservoir characterization and monitoring (Johann, 2006; Wu, 2005). Particularly, quantitative time-lapse saturation and pressure data constraints on numerical reservoir models might improve reservoir prediction reliability (Emerick, 2007; GOSSELIN, 2003; Mezghani, 2004). Risso (2007) underlines that applications of time-lapse information on the history match process through seismic attributes imply that at each step of the process it is necessary to deal with transformations between seismic attributes and saturation/pressure changes. Among the problems of this procedure are: high number of simulations, uncertainties relating the rock physics constraints, scale issues and the non-existence of a well established inversion method to derive saturation/pressure changes from seismic attributes. Tura and Lumley (1998; 1999) showed that cross-plotting time-lapse changes in seismic P and S impedance allowed the possibility to quantitatively estimate pressure or saturation in reservoir zones where only one of these properties changed