Application and Improvement of Ensemble Kalman Filter Method in Production Data Analysis

Production data analysis (PDA) is a subject to determine reservoir properties and predict future performance of single wells. It is normally conducted with either analytical methods or numerical history-matching methods. However, the analytical models normally have limited accuracy due to simplifying assumptions, while the history-match based on numerical simulations sacrifices computation load to increased accuracy. Therefore, in this study, the authors incorporated Ensemble Kalman Filter (EnKF) to improve PDA considering its superior efficiency in predicting system state and uncertainties. In this work, we applied the EnKF algorithm to single-well reservoir models to estimate permeability, skin factor, and drainage area. First, we tested the model accuracy after comparing the EnKF estimates to known reservoir properties. Next, we evaluated the cases with large estimation error and then adjusted the initial uncertainties and covariance of the static parameters. With confirmed improvements of property estimation in synthetic cases, the model is finally applied in a field study for further verifications. The results from this study confirmed that EnKF method could be an efficient solution for the modern PDA. They also indicate that accuracy issues are sometimes present when estimating skin factor and reservoir permeability simultaneously: large error exists in the property estimates, their uncertainties are overly reduced, and thus analysis and predicts are affected. By increasing the initial uncertainty bounds and adding minimum threshold values for the covariance, the property estimates could be improved, and reasonable uncertainty bounds are preserved. The methodology from this study is applicable for full-field evaluations.