Datapages, Inc.Print this page

Advanced Assisted Calibration Solution in Basin Modelling

Mathieu Ducros1, Jean-Marie Laigle1, and Stephane Bellin2
1Beicip-Franlab, 232 av Napoleon Bonaparte, 92502 Rueil-Malmaison cedex, France
2Beicip Inc, 1880 South Dairy Ashford #630, Houston, TX 77077, USA

One of the basin modeling major objectives consists in explaining the local spatial variations of measurements. In this direction a calibration process, which is a time consuming and hit-or-miss affair, is usually conducted by the modeler. It requires strong skills and rigorous methodology to be performed in a time frame compatible with operational constraints. As dynamic reservoir modeling simulators already benefit successfully from tools helping reservoir engineers in realizing assisted history matching, we propose to explore and demonstrate their added value in the calibration process for basin modeling.

More specifically, we intend to use a methodology which consists in 1) modeling the mismatch between simulated and measured data through an objective function 2) realizing a sensitivity analysis which determines the most influential parameters and 3) handling an inversion loop using advanced parameterization techniques to tune these parameters till reducing the mismatch to a satisfactory value.

The simultaneous calibration of both pressure and porosity is one of the most challenging steps of a basin modeling study. Our methodology has been successfully applied to constrain model parameters in order to fit pressure-related output variables on observation wells. The studied case features some of the classical difficulties basin modelers may face with (thin tight shale layers, faults, salt diapirisms and unknowns on sand connectivity among others). The parameters under optimization impact both the structural definition of the model and the fluid flow parameterization. Consequently, it has been necessary to chain in a single optimization loop the backward modeling, which allows getting the structural evolution through time thanks to a backstripping algorithm, and the forward thermal and fluid flow modeling.

The main steps of this workflow were: 1) The definition of an objective function based on the weighted least square method to measure the discrepancy between simulated results and measured porosity and pressure. At this stage, we had to weight measurement respectively to the confidence we had in their quality. As the information is not uniformly distributed spatially, it has been necessary to balance the formulation of the objective function in order to give an equal weight to each geological formation or well. 2) The parameters to be optimized have been formerly identified through a sensitivity analysis, conducted on a selection of parameters thanks to an experimental design and response surface technique. The most influential factors were kept to perform the final optimization on. 3) The realization of the assisted calibration process using gradient-based optimization algorithms led to a set of values fitting very accurately the measurements of both pressure and porosity.

Starting up from a comprehensive initial model, and after a step of identification of the main controlling factors, we demonstrate through this methodology a significant speed up in the calibration process without compromising on the results quality. In addition, the benefit of user’s expertise has been maximized at each step of the workflow. And eventually, in addition to the calibrated model, an in-depth interpretation of the outcomes of the optimization workflow have brought new qualitative and quantitative information on the basin physics such as spatial range of impact of a geological object (fault) or parameter (seal resistance).

References

Feraille, Roggero, Reis, (2007): Advanced history matching solutions: an integrated field case application, First break, 93.
Feraille, Roggero, Reis, (2003), Integration of dynamic data in a mature oil field reservoir model to reduce the uncertainty on production forecasting, AAPG annual meeting.

Simulated mud weight (in kg/m3) in respectively Well A, Well B and Well C resulting for the pre-calibration (we aimed at obtaining the main trends).

Simulated mud weight (in kg/m3) in respectively Well A, Well B and Well C after realizing the calibration loop.

 

 

AAPG Search and Discovery Article #90091©2009 AAPG Hedberg Research Conference, May 3-7, 2009 - Napa, California, U.S.A.