Prediction model for spontaneous combustion temperature of coal based on PSO-XGBoost algorithm

Area of Science:

• Geosciences and Environmental Science
• Chemical Engineering
• Computational Science

Background:

• Accurate prediction of coal spontaneous combustion temperature is vital for preventing mine disasters.
• Existing models often lack the precision needed for effective early warning systems.
• Thermodynamic disasters like coal spontaneous combustion and gas explosions pose significant risks in goaf areas.

Purpose of the Study:

• To develop a robust predictive model for coal spontaneous combustion temperature in goaf.
• To identify key gaseous indicators influencing spontaneous combustion.
• To enhance the accuracy and reliability of early warning systems for mine safety.

Main Methods:

• Collected 381 datasets from 9 coal types through programmed temperature experiments and industrial analysis.
• Utilized Pearson correlation coefficient for feature selection, identifying O2, CO, CO2, C2H4, C3H8, and various gas ratios as key indicators.
• Applied Particle Swarm Optimization (PSO) to tune the XGBoost regressor, creating the PSO-XGBoost model.

Main Results:

• The proposed PSO-XGBoost model demonstrated superior predictive accuracy and robustness compared to other models (PSO-RF, PSO-SVR, XGBoost, RF, SVR).
• Tenfold cross-validation confirmed the model's strong performance, fault tolerance, and universal applicability.
• Key input indicators identified include O2, CO, CO2, C2H4, C3H8, and specific gas ratios like C3H8/CH4 and CO2/CO.

Conclusions:

• The PSO-XGBoost model offers a highly accurate and reliable method for predicting coal spontaneous combustion temperature.
• The identified gaseous indicators provide valuable insights into the combustion process.
• This model significantly improves the potential for effective monitoring and early warning systems in mining operations.