| Machine learning and seismic attributes analysis to improve stratigraphic interpretation and gas channel detection in the Scarab field, offshore Nile Delta, Egypt. |
| Paper ID : 1046-ISCHU |
| Authors |
|
Tarek Mahmoud Khalifa *1, Amir Ismail2, Islam Ali Mohamed3, Amin Esmail Khalil2 1Helwan University. 2Helwan University 3Rashid Petroleum Company |
| Abstract |
| The western offshore Nile Delta of Egypt is one of the important hydrocarbon reservoirs globally. The reservoir is composed of vertically stacked channels where channels marked 1 and 2 are considered as the main reservoir. Unfortunately, these channels are disconnected, and their reservoir's architecture is hard to identify using conventional seismic interpretation techniques. This situation is inherited because of the thin and discontinuous nature of the reservoir sands and the presence of gas chimneys, which can create false positives when identifying hydrocarbon reservoirs. To resolve this problem, the present used adopted two unsupervised machine learning techniques, namely Kohonen's self-organizing maps (SOMs) and principal component analysis (PCA). PCA is used to reduce a large set of seismic attributes to a smaller set of more meaningful attributes containing most of the variation in the data. The output of the PCA serves as the input to the SOM, a type of neural network that can be used to cluster data into meaningful groups. SOMs accurately clustered meta-attributes into six groups, each representing a different lithology calibrated to well logs. Traditional techniques are not able to distinguish between different facies. The output of the SOMs was also able to discriminate between six major lithological facies, including siltstone, shaley sand, mudstone/shale 1, mudstone/shale 2, wet sand, and gas sand, and determine the gas/water contact on the seismic data. Moreover, the results demonstrate the potential of unsupervised machine learning techniques to improve the accuracy of seismic interpretations and extract subtle changes in the stratigraphy of deep marine channels. This is a significant finding, as it has the potential to reduce exploration and development risk and reveal more information about reservoir distribution and heterogeneity in the geological subsurface systems. |
| Keywords |
| Self-organizing maps, machine learning, seismic attributes, Principal component analysis, unsupervised machine learning, Scarab field. |
| Status: Abstract Accepted (Oral Presentation) |