Optimized Random Forest Method for 3D Evaluation of Coalbed Methane Content Using Geophysical Logging Data.

Area of Science:

• Geosciences
• Petroleum Engineering
• Data Science

Background:

• Accurate coalbed methane (CBM) content evaluation is vital for efficient CBM exploration and development.
• Traditional core sample analysis is costly; geophysical logging offers a cost-effective alternative but faces challenges due to complex, nonlinear relationships with CBM content.
• Advanced prediction methods are needed to overcome the limitations of existing techniques.

Purpose of the Study:

• To develop and validate an advanced model for predicting CBM content using geophysical logging data.
• To improve the accuracy and efficiency of CBM content evaluation in the No. 3 coal seam of the Qinshui Basin.
• To construct a 3D CBM content model for enhanced exploration and reserve evaluation.

Main Methods:

• Utilized geophysical logging data and 148 laboratory core samples from the No. 3 coal seam.
• Developed a Simulated Annealing-Genetic Algorithm-Random Forest (SA-GA-RF) model for CBM content prediction.
• Validated the SA-GA-RF model using test data and new well data, comparing its performance against BPNN, LSSVM, ELM, and MR methods.

Main Results:

• The SA-GA-RF model achieved an average relative error of 13.13% on the test dataset, outperforming other benchmark methods.
• Demonstrated strong generalizability in new wells and improved model-building efficiency.
• Successfully constructed a 3D CBM content model, enabling detailed identification of high gas content areas and layers.

Conclusions:

• The SA-GA-RF model provides a highly accurate and efficient method for evaluating CBM content from geophysical logging data.
• The developed 3D CBM content model offers superior characterization for CBM exploration, reserve evaluation, and production optimization compared to traditional 2D methods.
• This approach offers significant advantages for the effective management of coalbed methane resources.