Estimation of Coal's Sorption Parameters Using Artificial Neural Networks

Area of Science:

• Geochemistry
• Computational Science

Background:

• Accurate determination of coal sorption parameters (maximal sorption capacity, effective diffusion coefficient) is crucial for understanding coal's behavior as a natural sorbent.
• Current methods for determining these parameters are time-consuming and require expensive, specialized equipment.
• Coal's sorption properties are vital for managing methane and carbon dioxide, key components in coalbed gas reservoirs and greenhouse gas emissions.

Purpose of the Study:

• To apply artificial neural networks (ANNs) for efficient and accurate determination of coal's sorption parameters.
• To estimate maximal sorption capacity and effective diffusion coefficient using readily available data from technical, densitometric, and petrographic analyses.
• To validate the ANN model's performance against traditional gravimetric methods.

Main Methods:

• Utilized feed-forward back-propagation networks (FNNs), a type of artificial neural network.
• Input data included technical and densitometric analyses, and petrographic composition of coal samples.
• Sorption parameters were estimated using regressive neural models and compared with gravimetric measurements.

Main Results:

• The ANN models demonstrated high compatibility with gravimetric measurements, with prediction errors of 6.1% for effective diffusion coefficient and 0.2% for maximal sorption capacity.
• High determination coefficients (0.982 and 0.999) and low standard deviation ratios (below 0.1) confirmed the models' strong predictive capabilities.
• The developed neural models accurately predicted coal's sorption properties.

Conclusions:

• Artificial neural networks offer a promising, efficient alternative to traditional methods for determining coal sorption parameters.
• The proposed ANN approach provides a cost-effective and less time-consuming solution for assessing coal's natural sorbent characteristics.
• This method has significant potential for describing the sorption properties of coal for methane and carbon dioxide.