Multi-Variant Selection from History Matching to Prediction in Probabilistic Dynamic Model: A Case Study
Abstract
Multiple history matching approach to quantify remaining oil saturation distribution uncertainty is possible and practical to perform with technological advancement nowadays. This paper presents the selection method to optimize number of forecast variants while preserving the uncertainty. The results are low estimate, best estimate and high estimate of remaining oil saturation distribution for development scenario design and wide-covered representative variants to be used for prediction stage.
Different model variants considered match with historical data are selected based on few criteria such as field total liquid production, field total oil production, region pressure and well total oil production. To be able to measure how similar or different one variant to another, cluster analysis was performed. The clustering was based on significant parameters to objective function which were used to build proxy-equations. In this study, Multi-Dimensional Scaling (MDS) method was used to visualize the result in twodimensional space. Each variant represented by a point and the distance between points show their degree of similarity. After grouping process finished, different remaining oil saturation distribution were analyzed from P90, P50 and P10 quantile models.
There were 400 variants on final history match stage and then 127 variants were selected based on production and pressure profile similarity. Subsequently, number of forecast variants were optimized to 20 representative variants which cover the range of uncertainty. Three quantiles were selected from cumulative distribution function of these variants to be used for subsurface risk management in designing infill well location or waterflood pattern.
This paper demonstrates the application of extended uncertainty analysis by combining static modelling to dynamic simulation uncertainty variables in Limau Barat Limau Tengah Field which applicable in most development fields. Assisted history matching algorithm used in this study were determined for full field simulation. This paper also introduces the application of probabilistic remaining mobile oil saturation maps in assessing subsurface risk for better decision making in field development.
