Enhancing Geological Interpretation Efficiency and Accuracy Using Convolutional Neural Networks: A Case Study from Balsam Field, Nile Delta | ||||
| Scientific Journal of Faculty of Science, Menoufia University | ||||
| Volume 28, Issue 2, December 2024, Page 39-48 PDF (1.55 MB) | ||||
| Document Type: Original Article | ||||
| DOI: 10.21608/sjfsmu.2024.317987.1008 | ||||
 View on SCiNiTO | ||||
| Authors | ||||
Ali Abdel Baset  1; Moahmed Mahmoud Abu -El Hassan2; Mohamed Farouk Abu-Hashish3
| ||||
| 1El Wastani Petroleum Company (WASCO), Cairo, Egypt | ||||
| 2Geology Department, faculty of Science, Menoufia University | ||||
| 3Geology Department, Faculty of Science, Menoufia University | ||||
| Abstract | ||||
| In the oil and gas industry, reducing costs and improving data interpretation accuracy pose significant challenges, especially in mature fields like Balsam Field located onshore in the Nile Delta. These challenges are particularly evident during critical decisions on drilling new wells within geological units defined by conventional sedimentological studies. This study focuses on the application of Convolutional Neural Network (CNN) techniques, known for their exceptional performance in pattern recognition and classification, to predict borehole image facies efficiently and accurately within the Qawasim Formation at Balsam Field. The workflow comprises five major steps: data collection, preprocessing, CNN model training, testing, and evaluation. The dataset used includes 1350 images categorized into three labeled facies types (cross-laminated, laminated, and massive). The trained CNN model employs convolutional and max-pooling filters for feature extraction, followed by fully connected neural network layers for classification. The model achieved a significant accuracy of 82%, demonstrating its effectiveness in rapid facies prediction to support real-time decision-making and cost reduction strategies in borehole image analysis. Moreover, this adaptable model can be extended to other clastic reservoirs, offering quick and accurate geological models essential for future field development planning and production optimization. The application of deep learning, as illustrated in this study, enhances both efficiency and accuracy in borehole image interpretation, thereby reducing geological study costs and minimizing risks associated with reservoir modeling. | ||||
| Keywords | ||||
| Convolutional Neural Network (CNN); Borehole image facies; Qawasim Formation; Machine learning; Nile Delta | ||||
|
Statistics Article View: 128 PDF Download: 106 |
||||
