A New Pulsed Neutron-Gamma Density Logging Method Based on Gamma-Ray Spectra and Machine Learning.

Area of Science:

• Geophysics
• Petroleum Engineering
• Nuclear Geophysics

Background:

• Formation bulk density is crucial for oil and gas reservoir evaluation.
• Pulsed neutron-gamma density (NGD) logging is a safer, eco-friendlier alternative to gamma-gamma density (GGD) logging.
• NGD accuracy is challenged by interferences like pair production, neutron transport, and borehole conditions.

Purpose of the Study:

• To analyze NGD interference factors, particularly formation chemistry's impact.
• To develop an improved NGD density calculation model integrating gamma-ray spectra.
• To leverage machine learning for enhanced density prediction and borehole correction.

Main Methods:

• Analysis of interference factors using path diagrams.
• Integration of gamma-ray spectra into the density calculation model.
• Application of machine learning regression algorithms for density prediction and correction.

Main Results:

• The developed model significantly improves density prediction accuracy, reducing root-mean-square errors from >0.03 g/cm³ to <0.01 g/cm³.
• The machine learning approach with gamma-ray spectra outperforms the conventional four-detector NGD method, even with a single detector.
• Machine learning enables single-step density prediction and borehole correction, improving workflow and applicability.

Conclusions:

• Integrating gamma-ray spectra and machine learning offers a robust solution for accurate NGD formation density measurements.
• This advanced method enhances cost-effectiveness, resolution, and applicability in geophysical exploration.
• The technique demonstrates considerable practical potential for reservoir evaluation, even without standoff information.