采空区煤自燃是诱发矿井火灾的主要因素之一，在矿井火灾中的占比高达90%。为实现采空区自燃的精准防治，需要准确获得采空区内部的高温点温度，以砚北煤矿为工程背景，分析其煤样氧化升温过程中产生的指标气体，建立煤自燃温度预测的指标体系，进而开展基于深度学习的多指标气体煤自燃温度预测研究。首先对砚北煤矿采集的煤样进行煤氧化升温实验，根据实验结果划分为单一气体指标与复合气体指标，分析各指标随温度上升的变化规律，进而确定合适的指标作为煤自燃温度预测指标；使用多源数据处理方法对煤自燃温度预测指标进行处理，应用库克距离法和多重插补法对数据进行清洗，并结合灰色关联度分析法建立煤自燃温度预测指标体系；使用Elman神经网络构建预测模型，确定模型结构与超参数后，进而建立煤自燃温度预测模型，获得煤氧化升温过程中温度的准确预测。

Spontaneous combustion of coal in goaf is one of the main factors inducing mine fire, which accounts for 90% of mine fire. In order to achieve accurate prevention and control of spontaneous combustion in goaf, it is necessary to accurately obtain the high temperature point temperature in goaf. Based on Yanbei Coal Mine as the engineering background, this paper analyzes the index gases produced during the oxidation and heating process of coal samples, establishes the index system of coal spontaneous combustion temperature prediction, and then carries out the research of multi-index gas coal spontaneous combustion temperature prediction based on deep learning. Firstly, the coal samples collected from Yanbei Coal Mine were tested for coal oxidation and temperature rise. According to the experimental results, they were divided into single gas index and compound gas index. The change law of each index with temperature rise was analyzed, and the appropriate index was determined as the prediction index of coal spontaneous combustion temperature. The prediction index of coal spontaneous combustion temperature is processed by multi-source data processing method, the data is cleaned by Cook distance method and multiple interpolation method, and the prediction index system of coal spontaneous combustion temperature is established by grey correlation analysis method. Elman neural network was used to build a prediction model, and after determining the model structure and hyper-parameters, a prediction model of coal spontaneous combustion temperature was established to obtain an accurate prediction of the temperature in the process of coal oxidation and heating.