3D Channel Reservoir Characterization Using a Model Selection Scheme for Fast and Reliable History Matching
B. Kang, G. Kim and J. Choe
Event name: Saint Petersburg 2018
Session: Reservoir Modelling
Publication date: 09 April 2018
Info: Extended abstract, PDF ( 778.67Kb )
Price: € 20
In this study, we use ensemble Kalman filter (EnKF) for the optimization of the reservoir models. EnKF uses multiple models, which enable stochastic analysis of the parameters. However, analysis of a large number of models brings huge simulation costs. Also, even though the models are updated after applying EnKF, poorly-designed initial models still lead to wrong estimations.Our method is to select good initial models before applying the EnKF algorithm. We use principal component analysis (PCA) for hundreds of initial models to discover some common trends of the permeability distributions in the models. After projecting the models onto a 2D principal component plane, the model with the smallest error is selected as a representative and we choose 100 models near it for our history matching analysis. This process can reduce simulation time in EnKF as well as increase prediction quality on reservoir performances.We show our works in 3D channel reservoir case and the proposed method improves in both time and reliability of the analysis. What is more, since we use only a quarter number of initially generated 400 models, the simulation takes 75% less time when compared to the original case.