Prediction of temperature and CO concentration fields based on BPNN in low-temperature coal oxidation

Highlights: • Constructed a system to implement a process of low temperature oxidation of coal. • Selected back propagation neural network and Bayesian Regularization algorithm. • Predicted the temperature and CO concentration fields. To prevent coal spontaneous combustion, it is critical to accurately simulate and predict the low-temperature oxidation process of coal. In this study, an experiment system was constructed to investigate the temperature and gas concentration of two typical coals during low temperature oxidation. The back propagation neural network (BPNN) was proposed to simulate this process under six factors, including activation energy, void fraction, moisture content, air flow rate, stacking time and location of measuring point. The average relative error of oxygen concentration, temperature and CO concentration are 1.34 %, 1.09 % and 1.15 %, respectively. Besides, the statistical performance indicators are precise. Most importantly, the predicted temperature and gas concentration fields can be utilized to analyze coal spontaneous combustion process which allows us to track down the crucial factors on spontaneous combustion. By applying our BPNN modeling method, the odds of coal spontaneous combustion can be lowered.