Channel Reservoir Characterization by Ensemble Smoother with Selective Water Breakthrough Data

Abstract

Reservoir characterization is a process to figure out reservoir parameters of interest using available data. Ensemble Smoother (ES) method is one of the main methods of reservoir characterization in petroleum engineering. ES utilizes and assimilates all available data through a single global update. ES is fast but unstable with possible overshooting of parameters and distorting of parameters distribution.

In this research, a method called selective use of measurement data using ES for each well is suggested in order to improve its performances. The proposed method is to use oil production rates before water breakthrough and water cut rates after water breakthrough for each well.

ES is very sensitive to the initial ensemble members. Therefore, it is necessary to select reliable models, which are similar to the reference one for better stability in the assimilation step. In this study, PCA (Principle Component Analysis) and K-means Clustering are applied to get good reservoir models.

In order to check the superiority of the proposed method, 3 methods are compared in our research: ES with all data

ES with 100 initial models selected by PCA and K-means Clustering

and ES with 100 initial selected models and selective measurement data. 2D synthetic channelized reservoir models are applied with nine-spot waterflooding.

The proposed method can properly overcome the limitation of standard ES and conserve channel connectivity. this method manages high uncertainty ranges, gives the best reservoir characterization results and shows the reliable future performance estimations.