Abstract: Exceeding the limits of harmful gases in underground mines is a core problem that restricts the safe production of coal mines, among which the abnormal concentrations of carbon monoxide （CO） and carbon dioxide （CO₂） not only directly threaten the lives and health of underground workers, but also may trigger secondary disasters such as coal spontaneous combustion, leading to the suspension of production at the working face or even the closure of the mine. In order to reveal the adsorption-desorption behaviours of these two gases in coal, this study takes coal samples from Xiaobaodang Coal Mine as the research object, and carries out CO and CO₂ adsorption-desorption experiments relying on the self-developed multivariate gas isothermal adsorption experimental device, analyses gas adsorption and desorption laws, and builds a coal spontaneous combustion prevention and control system based on this. The results show that the adsorption and desorption capacities of both CO and CO₂ are positively correlated with the equilibrium pressure, meaning that the higher the equilibrium pressure, the greater the adsorption and desorption amounts of the gases in coal. Their residual amounts exhibit differences: the residual CO amount is closely related to the equilibrium pressure stage, while the residual CO₂ amount decreases with the increase of equilibrium pressure. Due to its stronger adsorption capacity, CO₂ can “drive out” CO in underground environments, altering the gas distribution pattern. Meanwhile, the desorption process of CO in coal samples significantly lags behind the adsorption process. The study simulates the distribution characteristics of CO in the mining area, and delineates the spontaneous combustion hazardous area in the mining area. The HA-BP neural network is used to construct an early prediction system for spontaneous coal combustion, and the prediction results of the system for coal temperature have a very small deviation from the experimental coal temperature, with a regression coefficient of about 0.99. In view of the current actual situation of Xiaobaodang Coal Mine, it optimises the prediction system by using the big data mining technology to improve the prediction accuracy, constructes a multi-directional coal spontaneous combustion monitoring and early warning system device, and put forward the comprehensive prevention and control system of spontaneous combustion in the mining hollow zone and the technological solutions. Proposed comprehensive prevention and control system and technology programme for spontaneous combustion in the mining area. The results of the study reveal the adsorption-desorption characteristics of coal to CO and CO₂, improve the disaster prevention and control ability, and are of great practical significance and popularisation value for the safe and sustainable development of the coal industry.