煤炭是支撑当今世界经济发展的重要能源之一，但煤矿火灾依然是煤矿安全生产过程中的重大隐患。由于何家塔煤矿50105工作面采空区推采速度快，冒落空间较大，采空区内部浮煤遗留较多，工作面供风量相对较大但是漏风不便管理，易于造成采空区遗煤自燃。 本文依据何家塔煤矿50105工作面采空区易自燃煤层煤自然发火的特点，研究采空区遗煤自然发火规律，确定预测指标，建立煤自燃预测模型并验证其有效性。首先，选用西安科技大学大型实验台测定何家塔煤样自然发火表征参数，分析煤自燃过程中的指标气体、温度、耗氧速率、放热强度以及自然发火期等表征参数，得出何家塔矿煤样实验最短自然发火期为45天，实际自然发火期为46天；其次，对何家塔煤矿50105工作面采空区现场进行观测实验，得到了工作面采空区温度及CO浓度等特征参数分布规律；最后，结合灰色关联分析选择温度、CO为预测指标，结合实验和现场观测数据利用最小二乘法进行数据拟合，得到以温度和CO浓度为预测指标的预测模型，通过残差分析得出温度预测模型可靠性更高；再将温度预测模型应用到现场预测得到最短煤自然发火期为44天，较好的验证了煤自燃温度预测模型的有效性，该预测模型对相似条件的煤矿有一定的参考和借鉴作用。

Coal is one of the important energy sources supporting the economic development in the world today, but coal mine fires remain a major hidden danger in the safety production process of coal mines. Due to the high mining speed in the goaf of the 50105 working face of the Hejiata Coal Mine, the large amount of floating coal in the goaf remains, and the relatively large amount of air supply in the working face, but the management of the air leakage is inconvenient, and it is easy to cause the goaf coal spontaneous combustion. Based on the characteristics of natural spontaneous combustion of spontaneous coal seam coal in the goaf of 50105 working face in Hejiata coal mine, this paper studies the spontaneous combustion law of leftover coal in goaf, determines the forecasting index, establishes coal self-ignition prediction model and verifies its effectiveness. Firstly, the Xi'an University of Science and Technology used a large-scale experimental platform to measure the spontaneous combustion characterization parameters of the Hejiata coal, and analyzed the parameters such as the index gas, temperature, oxygen consumption rate, heat release intensity and natural firing period in the process of coal spontaneous combustion. The shortest spontaneous combustion period of the coal sample test for the tower coal mine is 45 days, and the actual spontaneous combustion period is 46 days. Secondly, observation experiments are conducted on the goaf of the 50105 working face of the Hejiata coal mine, and the temperature and CO concentration of the gob of the working face are obtained. Finally, combined with grey correlation analysis, temperature and CO were selected as forecasting indicators. The experimental and on-site observation data were combined with least-squares method to fit the data. Predictive models with temperature and CO concentration as predictors were obtained. The difference analysis shows that the reliability of the temperature prediction model is higher; the application of the temperature prediction model to the on-site prediction to obtain the shortest coal spontaneous combustion period of 44 days is a good example of validating the validity of the coal self-ignition temperature prediction model. The prediction model is similar. The conditions of the coal mine have certain reference and reference.