Improved History Matching and Uncertainty Quantification Using an Ensemble Based Approach, Case Study S-Field Sumatra
Author: Ainun N Chanif, Galih I Pambuko*, Kevin A Pratama
galih.imam@rda-indonesia.com
The purpose of reservoir modeling is to solve the inverse problem by building a 3D mathematical model of the reservoir using measured data. To minimize the investment risk and make the best decision for field development strategy, reservoir engineers have to build a reservoir model close to the actual condition. Understanding the field, including the uncertainties, is essential to make a trusted reservoir model prediction. The traditional approach to reservoir modeling and history matching is looking for one single base case model using a trial error approach to get a history-matched model. Reservoir modelers commonly use the conventional method. Therefore, it has a limitation in quantifying the model uncertainties that can lead to model inconsistencies between static and dynamic data measurements during the history matching process. A step-wise manner in modeling using a traditional approach can limit the interaction between different subsurface disciplines during the model generation and history matching. The generated model might perfectly match the current dynamic data measurements but fail to honor the static data and the geological concept. We applied an ensemble-based approach combined with machine learning software in S-Field to quantify the uncertainty parameters using a probabilistic approach during the history matching process, which enabled the automatic generation of multiple equiprobable realization models under uncertainties. Firstly, an initial ensemble of models is generated to capture the uncertainties in all parts of the modeling process. A dynamic modeling workflow is created to input dynamic modeling parameters. We introduce uncertainty in permeability endpoints and capillary pressure curves in all facies types. There is no SCAL data in S-Field, so permeability needs to be calculated using correlation, and it proved to be a challenge to find a suitable range. Creating the initial ensemble (CIE) process is done by doing facies modeling and petrophysical modeling, introducing concept probability for facies distribution to honor the existing geological concept. The match level difference between the observed data and generated model can be assigned from the initial ensemble result. Then the uncertainty parameters that should be modified in the history matching process are identified to achieve an initial ensemble that covers and follows the observed data’s trend. Finally, the history matching process is done by performing a computational step with several iterative methods to achieve an ensemble of reservoir models that match observed data. The proposed workflow and method can be the case solution to capture the uncertainties when we want to achieve reservoir models which consistently honor the static data and geological concept. Hence, the result of an ensemble reservoir model can be used to decide the best development strategy and minimize investment risk. Moreover, we can effectively utilize all available data consistently using the integrated workflow.
Keywords: reservoir, uncertainty, ensemble, history matching, probabilistic
View Full ManuscriptDownload