A review on coal spontaneous combustion prediction based on machine learning

The coal mining industry confronts a formidable challenge posed by spontaneous combustion disasters. As one of the crucial steps of prevention, spontaneous combustion prediction enables the early detection of potential risks, thereby facilitating the implementation of timely and effective preventive measures to ensure safety throughout the coal mining process. Machine learning methods, adept at analyzing and handling the intricate relationships between spontaneous combustion and various predictive indicators, have been extensively studied in the field of coal spontaneous combustion prediction. Therefore, this paper provides a comprehensive review of the research on machine learning in coal spontaneous combustion prediction, analyzes the difficulties in prediction and prospects its development direction. Firstly, the paper provides an analysis of primary predictive indicators utilized in forecasting spontaneous combustion temperature, hazard levels, and other aspects. Additionally, briefly discussing feature engineering techniques applied to spontaneous combustion prediction. Then, it analyzes the research progress in the field of coal spontaneous combustion prediction, including the application of artificial neural networks, support vector machines, cluster analysis, and ensemble learning machines. It summarizes the optimization algorithms applied to coal spontaneous combustion prediction. In addition, it also raises the existing problems in the current research, including the selection of prediction indicators being questionable, as well as the limited number of training samples and their inability to fully reflect the actual field conditions. Finally, the future directions for machine learning-based spontaneous combustion prediction are outlooked including following aspects. In terms of the data required for prediction models, it is proposed to improve the quantity and quality of samples. In terms of indicator selection for coal spontaneous combustion prediction, it is proposed to establish an adaptive optimization model for coal spontaneous combustion prediction indicators. In terms of coal spontaneous combustion prediction technologys, it is proposed to develope large model for predicting coal spontaneous combustion disasters. With the continuous improvement of computing performance, large language models tailored for all hazards in coal mines may be proposed, which can further improve the prevention and mitigation capabilities against coal mine accidents.