Deep learning framework based on ITOC optimization for coal spontaneous combustion temperature prediction: a coupled...

Area of Science:

• Mining Engineering
• Chemical Engineering
• Artificial Intelligence

Background:

• Coal spontaneous combustion (CSC) is a major safety risk in mines, necessitating accurate temperature prediction for early warning systems.
• Understanding coal oxidation and pyrolysis is vital for assessing combustion risk and developing effective control strategies.

Purpose of the Study:

• To develop and validate a novel deep learning framework for accurate temperature prediction in coal spontaneous combustion.
• To identify key gas indicators highly correlated with CSC temperature for improved predictive modeling.

Main Methods:

• Analysis of programmed heating experimental data and coal oxidation-pyrolysis coupled reaction mechanism.
• Pearson correlation analysis to identify six key gas indicators (O₂, CO, C₂H₄, CO/ΔO₂, C₂H₄/C₂H₆, C₂H₆).
• Development of an Improved Tornado Optimization with Coriolis force (ITOC) strategy integrated with a CNN-BiGRU-CBAM deep learning model for hyperparameter optimization and prediction.

Main Results:

• The ITOC algorithm demonstrated superior accuracy and convergence stability compared to five heuristic algorithms.
• The proposed ITOC-CNN-BiGRU-CBAM model achieved high accuracy on the test set with R² of 0.9738, MAPE of 4.1254%, MAE of 6.2740, and RMSE of 12.4735.
• Field validation in multiple coal mines confirmed the model's strong generalization and engineering adaptability, with predicted temperatures closely matching measurements.

Conclusions:

• The ITOC-CNN-BiGRU-CBAM model provides a robust and accurate method for predicting coal spontaneous combustion temperatures.
• This intelligent framework offers a promising theoretical and practical solution for early warning and precise prevention of CSC hazards in mining operations.
• The identified key gas indicators serve as reliable predictors for assessing coal oxidation stages and combustion risks.