Coal Identification Based on Reflection Spectroscopy and Deep Learning: Paving the Way for Efficient Coal Combustion...

Area of Science:

• Geoscience
• Materials Science
• Artificial Intelligence

Background:

• Coal is a critical global energy source, requiring accurate identification for efficient utilization in mining, combustion, and pyrolysis.
• Current coal identification methods may lack the speed and accuracy needed for dynamic industrial processes.
• Optimizing coal energy conversion relies on precise classification based on its type and properties.

Purpose of the Study:

• To develop a rapid and accurate coal identification approach using deep learning and reflection spectroscopy.
• To create a novel model, RS_PSOTELM, integrating convolutional neural networks (CNN) and extreme learning machines (ELM) for spectral data analysis.
• To enhance the model's performance through particle swarm optimization (PSO) for parameter tuning.

Main Methods:

• Field collection and preprocessing of spectral data from diverse coal samples.
• Development of the RS_PSOTELM model: CNN for feature extraction and ELM for classification.
• Optimization of ELM parameters using particle swarm optimization (PSO) to improve identification accuracy.

Main Results:

• The RS_PSOTELM model demonstrated a high accuracy of 98.3% in coal identification tasks.
• The approach proved capable of quick and precise identification of different coal categories.
• Successful extraction of effective spectral features and accurate classification were achieved.

Conclusions:

• The proposed RS_PSOTELM method offers a low-cost, efficient, and reliable solution for coal identification.
• This technique supports optimized coal utilization in mining, combustion, and pyrolysis stages.
• The study paves the way for enhanced energy conversion efficiency through advanced coal classification.