Innovative Simulation History Matching Approach Enabling Better Historical Performance Match and Embracing Uncertainty in Predictive Forecasting
Emad Elrafie, Mohammed Agil, Tariq Abbas, Boy Idroos and François-Michel Colomar
Event name: SPE/EAGE Reservoir Characterization & Simulation Conference
Session: Reservoir Characterization & Simulation
Publication date: 19 October 2009
Organisations: SPE, EAGE
Info: Extended abstract, PDF ( 614.78Kb )
The purpose of the simulation history match phase in a study is to achieve a simulation model calibrated to historical performance for predictive production forecasting while preserving reservoir understanding in terms of reservoir characterization and fluid flow mechanisms. The classical history match simulation approach involves running a number of history match simulation cases with modified simulation model variables to obtain only one of the many probable match models to the field data. Undoubtedly, the conventional simulation history match approach does not normally handle the uncertainty of all model variables, nor the possibility to identify and carry forward a set of multiple equi-probable history match model scenarios to predictive forecasting. Furthermore, the conventional history match approach lacks a rigorous mechanism to ensure that the original reservoir characterization and understanding is preserved after achieving only one of the many probable history match models. This paper presents an innovative history match approach as part of Saudi Aramco’s integrated “Event Solution1” study workflow. This approach was developed to enable faster simulation history match under uncertainty, in terms of static and dynamic variables. The history matching process is performed with the aid of assisted history matching software 2 that tracks the match quality of hundreds of history match cases and analyzes the impact of each variable and its range of uncertainty on model match quality to historical field data. Finally, a proxy (statistical History Match solution surface including all uncertainty variables) is created that combines model learnings to provide directional guidance to a most likely history match model design. As the history match process progresses, history match variables are characterized into three distinct categories; (1) critical variables to history match, (2) non critical variables to history match but with significant impact on prediction, and (3) non critical variables to history match but with less impact on prediction. The impact of the variables on prediction is concluded by concurrently running prediction runs under uncertainty. The uncertainty range of the variables categorized in groups (1) and (3) are set to a single realizations or narrower range of uncertainty for each variable while group (2) variables are carried forward with a more restricted range of uncertainty (defined by history match quality analysis) setting the stage for prediction under uncertainty modeling. This paper presents the application of an innovative history match approach that provides all project stakeholders with a shared understanding of critical and non critical uncertainties (static and dynamic) in history match as carried forward to prediction runs under uncertainty.