煤自燃问题一直是矿井安全生产亟需解决的难点，及时准确的预测预报是煤自燃防控的前提。煤在自燃的过程中，会产生一些具有规律性并易于检测的标志性气体，如CO、C2H4、C2H2、C2H6等。通过准确检测指标气体浓度的变化可以为煤自燃早期隐患识别与预警提供可靠判据，降低误报率和漏报率。 本文采用可调谐半导体激光吸收光谱法（TDLAS）实现气体的非接触式高灵敏度检测。以朗伯-比尔定理为基础，通过对HITRAN数据库进行分析，选择煤自燃标志性气体位于1627 nm处的C2H4，1532 nm处的C2H2，2332 nm处的CO进行研究。本文的主要研究内容如下： （1）驱动激光器对指标气体激光器波长特性实验研究。基于可调谐半导体激光光谱分析技术，采用1627 nm、1532 nm、2332 nm的DFB激光器分别作为C2H4、C2H2、CO气体浓度检测的窄带光源，设计气体激光器波长特性实验，分别测试激光器电流和温度的微小变化对激光器输出光波长的影响，并通过对电流波长调谐特性曲线进行补偿与验证，从而减小由失真带来的系统测量误差，提高系统检测精度。 （2）煤自燃烃类标志性气体吸收谱线交叉干扰修正方法研究。为了解决实际测量中煤自燃烃类标志性气体甲烷对乙烯吸收谱线交叉干扰的影响，设计交叉干扰实验装置，将CH4和C2H4的混合标气进行实验测试，将实测数据分为训练样本集和测试样本集两组；通过三种修正算法：BP神经网络、弹性网络回归和改进微粒群优化极限学习机（PSO-ELM）补偿算法，分别建立非线性补偿模型，利用训练后的模型对测试样本集进行测试，通过对比三种修正模型的测试精度可得，基于BP神经网络的交叉干扰修正模型修正效果最好，修正后C2H4浓度最大相对误差减小了10.373%，从而实现CH4对C2H4气体吸收谱线交叉干扰的修正。 （3）温度和压力波动引起的标志性气体吸收谱线畸变的修正方法研究。针对环境温度和压强对气体测量浓度的影响，搭建气体温度-压力TDLAS测试系统，测试煤自燃标志性气体C2H4、C2H2、CO浓度随气体温度和压力的变化特性与变化规律；根据测试结果使用支持向量机算法建立气体温度-压力修正模型并进行算法预测与修正，根据修正结果显示，补偿后C2H4浓度最大相对误差减小了13.524%；C2H2浓度最大相对误差减小了43.817%；CO浓度最大相对误差减小了17.776%，从而实现了煤自燃标志性气体检测温度-压力的修正。 煤自燃标志性气体激光光谱高精度检测误差修正方法研究为完善煤自燃预测技术体系提供坚实的理论基础，也有利于推动煤火灾害防控理论与技术的发展。

Coal spontaneous combustion has always been a difficult problem to be solved urgently in mine safety production. Timely and accurate prediction is the premise of coal spontaneous combustion prevention and control. In the process of coal spontaneous combustion some regularity and easy-to-detect marker gases, such as CO, C2H4, C2H2, C2H6 and so on, will be produced. By accurately detecting the change of gas concentration, this paper can provide reliable criteria for identifying and warning the early hidden dangers of coal spontaneous combustion, and reduce the rate of false alarm and missed alarm. In this paper, tunable diode laser absorption spectroscopy (TDLAS) is used to realize non-contact high sensitivity detection of gases. Based on Lambert-Bill theorem, through the analysis of HITRAN database, C2H4 at 1627 nm, C2H2 at 1532 nm and CO at 2332 nm，were selected as the marker gas of coal spontaneous combustion. The main contents of this paper are as follows: Experimental study on wavelength characteristics of target gas lasers by driving lasers. Based on tunable semiconductor laser spectroscopy technology, the wavelength characteristics of gas lasers were designed by using DFB lasers at 1627, 1532 and 2332 nm as narrow-band light sources for detecting C2H4, C2H2 and CO gas concentration, respectively. The effect of small changes on current and temperature on the output wavelength of laser is experimented, and the tuning characteristic curve of current wavelength is compensated and validated to reduce the system measurement error caused by distortion and to improve the detection accuracy of the system. Experimental study on cross-interference correction method for hydrocarbon absorption spectrum of coal spontaneous combustion. In order to solve the influence of methane on cross-interference of ethylene absorption spectra in the actual measurement of coal spontaneous combustion hydrocarbon marker gases, a cross-interference experimental device was designed, and the mixed standard gases of CH4 and C2H4 were tested. The experiment divides the measured data into two groups: training sample set and test sample set. Through three modified algorithms: BP neural network, elastic network regression and improved particle swarm optimization extreme learning machine (PSO-ELM) compensation algorithm, the non-linear compensation model is established, and the test sample set is tested with the trained model. By comparing the test accuracy of the three modified models, the cross-interference correction model based on BP neural network has the best effect. The maximum relative error of C2H4 concentration was reduced by 10.373% after correction, which realize the correction of the cross-interference of CH4 to the absorption spectrum of C2H4. Experimental study on the correction method about the spectral distortion of the signature gas absorption caused by the fluctuation of temperature and pressure. Aiming at the influence of ambient temperature and pressure on gas concentration measurement, a TDLAS gas temperature-pressure measurement system was set up to test the variation characteristics and rules of C2H4, C2H2 and CO concentration in coal spontaneous combustion. According to the test results, a gas temperature-pressure correction model is established by using support vector machine algorithm, and the algorithm is predicted and revised. The results show that after compensation, the maximum relative error of C2H4 concentration decreases by 13.524%, the maximum relative error of C2H2 concentration decreases by 43.817%, and the maximum relative error of CO concentration decreases by 17.926%. Thereby, the correction of temperature and pressure in the detection of coal spontaneous combustion marker gas is realized. The research on laser spectrum high precision detection error correction method of coal spontaneous combustion marker gas provides a solid theoretical basis on improving the prediction technology system of coal spontaneous combustion. Meanwhile，the study is also conducive to promoting the development of theory and technology in coal fire prevention and control.