Feature Selection Based on Grey Wolf Optimizer for Oil Gas Reservoir Classification
The classification of the hydrocarbon reserve is a significant challenge for both oil and gas producing firms. The factor of reservoir recovery contributes to the proven reservoir growth potential which leads to a good preparation of field development and production. However, the high dimensionality...
| Main Authors: | Al-Tashi, Q., Rais, H.M., Abdulkadir, S.J., Mirjalili, S. |
|---|---|
| Format: | Conference or Workshop Item |
| Institution: | Universiti Teknologi Petronas |
| Record Id / ISBN-0: | utp-eprints.29882 / |
| Published: |
Institute of Electrical and Electronics Engineers Inc.
2020
|
| Online Access: |
https://www.scopus.com/inward/record.uri?eid=2-s2.0-85097561507&doi=10.1109%2fICCI51257.2020.9247827&partnerID=40&md5=491de640e828ea65f6ece71c79c58ee1 http://eprints.utp.edu.my/29882/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: |
The classification of the hydrocarbon reserve is a significant challenge for both oil and gas producing firms. The factor of reservoir recovery contributes to the proven reservoir growth potential which leads to a good preparation of field development and production. However, the high dimensionality or irrelevant measurements/features of the reservoir data leads to less classification accuracy of the factor reservoir recovery. Therefore, feature selection techniques become a necessity to eliminate the said irrelevant measurements/ features. In this paper, a wrapper-based feature selection method is proposed to select the optimal feature subset. A Binary Grey Wolf Optimization (BGWO) is applied to find the best features/measurements from big reservoir data obtained from U.S.A. oil gas fields. To our knowledge, this is the first time applying the Grey Wolf Optimizer (GWO) as a search technique to search for the most important measurements to achieve high classification accuracy for reservoir recovery factor. The wrapper K-Nearest Neighbors (KNN) classifier is used to evaluate the selected features. In addition, to examine the efficiency of the proposed method, two recent algorithms namely: Whale Optimization algorithm (WAO) and Dragonfly Algorithm (DA) are implemented for comparison. The experimental results showed that, the proposed BGWO-KNN significantly outperforms benchmarking methods in terms of feature reduction as well as increasing the classification accuracy. The proposed method shows a great potential for solving the real oil gas problems. © 2020 IEEE. |
|---|