Robust reservoir identification by multi-well cluster analysis of wireline logging data

Area of Science:

• Geoscience
• Petroleum Engineering
• Data Science

Background:

• Accurate rock type identification is crucial for hydrocarbon exploration and production.
• Traditional clustering methods like K-means are sensitive to noise and initial centroid selection.
• Well log data (gamma ray, density, sonic, photoelectric index, resistivity) are essential for characterizing subsurface formations.

Purpose of the Study:

• To introduce and validate a novel Most Frequent Value (MFV) based clustering method for enhanced rock type identification using multi-dimensional well log data.
• To demonstrate the robustness and noise rejection capabilities of the MFV method compared to K-means.
• To apply the MFV clustering workflow to identify lithological and petrophysical characteristics in clastic reservoirs.

Main Methods:

• Application of a Most Frequent Value (MFV) based clustering technique to multiple well log types (natural gamma ray, bulk density, sonic, photoelectric index, resistivity).
• Utilizing a histogram-based selection method for optimal initial cluster center placement.
• Employing a weighted Euclidean (Steiner-) distance and automated weighting for cluster element analysis.
• Processing synthetic and real-world well log data, including single and multi-well datasets.

Main Results:

• The MFV clustering method demonstrated high noise rejection and efficient cluster recognition, even with outlying and missing data.
• The method proved more reliable than K-means, showing less sensitivity to initial centroid selection.
• Application to Hungarian Miocene clastic reservoirs successfully identified lithological and petrophysical characteristics, validated by core permeability and independent log analyses.

Conclusions:

• The proposed MFV clustering method offers a robust and accurate approach for rock type identification from well log data.
• This technique significantly improves upon traditional methods by handling noisy and incomplete datasets effectively.
• The workflow provides a valuable tool for detailed subsurface characterization in hydrocarbon exploration and reservoir management.