Data-Driven Approach for the Prediction of In Situ Gas Content of Deep Coalbed Methane Reservoirs Using Machine...

Area of Science:

• Earth Science
• Geology
• Petroleum Engineering

Background:

• In situ gas content is vital for coalbed methane (CBM) exploitation and recovery.
• Deep CBM resources offer significant potential but face challenges in acquiring in situ gas content data due to complex geology.

Purpose of the Study:

• To develop an efficient and accurate method for predicting the in situ gas content of deep CBM using well logging data.
• To integrate multiple algorithms to overcome limitations of traditional prediction models.

Main Methods:

• Developed a novel prediction model by integrating Gray Relational Analysis (GRA) and Genetic Algorithm (GA) with a Back-Propagation Neural Network (BPNN).
• Utilized GRA to identify optimal input parameters for the BPNN.
• Employed GA to optimize the initial weights and thresholds of the BPNN.

Main Results:

• The combined GRA-GA-BPNN model demonstrated superior performance compared to the standalone BPNN model.
• Achieved a 77.60% reduction in Mean Square Error (MSE) compared to the BPNN model.
• Validated the model's reliability with an average relative error of only 4.62% in deep CBM wells in the Ningwu Basin.

Conclusions:

• The GRA-GA-BPNN model effectively enhances prediction accuracy and stability for deep CBM in situ gas content.
• The model exhibits high robustness and strong generalization ability, reducing reliance on experimental measurements.
• This approach holds significant practical application value for deep CBM resource assessment and development.