A novel selection method of seismic attributes based on gray relational degree and support vector machine

Area of Science:

• Geophysics
• Petroleum Geoscience
• Machine Learning Applications

Background:

• Seismic attribute selection is crucial for accurate reservoir prediction.
• Current methods for effective seismic attribute selection remain a challenge.
• Reliability and credibility of seismic attributes directly impact prediction accuracy.

Purpose of the Study:

• To present a novel, two-hierarchical method for seismic attribute selection.
• To improve the accuracy of reservoir prediction, specifically for coalbed methane (CBM) content.
• To establish a robust approach for identifying the most relevant seismic attributes.

Main Methods:

• A novel two-hierarchical selection method combining gray relational degree (GRD) and support vector machine (SVM).
• Primary selection using GRD to assess relationships between seismic attributes and reservoir parameters, and among attributes themselves.
• Secondary selection employing SVM for interactive error verification with training samples.

Main Results:

• The GRD-SVM method successfully identified reliable seismic attributes for CBM content prediction in a real-world case study.
• Selected attributes included instantaneous amplitude, instantaneous bandwidth, instantaneous frequency, and minimum negative curvature.
• Predicted CBM content showed strong consistency with measured data, validating the method's effectiveness.

Conclusions:

• The proposed GRD-SVM method effectively selects seismic attributes for reservoir prediction.
• This approach significantly improves prediction accuracy, as demonstrated by the CBM content case study.
• The GRD-SVM method offers a practical and reliable tool for seismic attribute selection in geological applications.