煤层火灾，是指埋藏在地下的煤层自燃或被采煤生产引燃后，随时间逐渐扩散成大规模的煤田火灾，又称煤火。煤火不仅会对煤矿安全造成严重的影响，造成不可再生的煤炭资源浪费，更对人类赖以生存的大气环境造成无法估量的破坏。因此，如何发现地下煤火，尤其是在其燃烧初期确定火源点的位置，是煤田火区灾害成功治理的关键。 本文通过分析露头煤田的煤体特点，搭建松散煤体实验台，模拟地下火源的燃烧升温过程，同时结合实验煤体燃烧数据，建立传热模型。采用红外热成像技术，从热传导理论、光学、煤矿石燃烧及导热等多方面进行分析研究，给出了一种GPSO混合优化反演算法。该算法结合了萤火虫(GSO)与粒子群(PSO)两种算法的优点，在求解精度及速度上都有了提升。通过构建实验煤体中的点热源扩散方程，作为GPSO算法的正演模型，结合煤体表面的热红外图像，反演火源点位置。本文构造的GPSO混合优化算法与常用的传统算法相比，避免了有限差分法和正则化方法离散化所引起的模型阶次误差，从而提高了反演的准确性。同时，文中对GSO、PSO算法和GPSO混合算法的反演结果做了较为全面的比较、分析，为实地火源探测提供了重要理论依据。

Coal seam fire, refers to the coal seam spontaneous combustion or be buried in the ground, ignition after coal mining production and gradually spread over time into a large-scale coalfield fire, also called coal fire. Coal fire will not only cause serious influence on coal mine safety, cause the non-renewable resources of coal waste, more to the survival of humans on atmospheric environment incalculable consequences. So how to find underground coal fire, particularly in its early to determine the location of the fire point, burning coal fire area is the key to the successful governance of the disasters. In this paper, through analyzing the characteristics of the coal outcrop coal, structures, loose coal test bench, the simulation of the underground fire burning heating process, combined with coal combustion experimental data at the same time, the heat transfer model is established. By using infrared thermal imaging technology, from the theory of heat conduction, the optical, stone coal combustion and thermal analysis, to construct a GPSO hybrid optimization inversion algorithm. The algorithm combines the firefly (GSO) and particle swarm (PSO), the advantages of both methods have the promotion in the precision and speed. By building the experimental coal point heat source diffusion equation, the positive evolution model initialization GPSO algorithm, combined with the coal on the surface of the internal temperature field of thermal infrared image media information, inversion of ignition point location. The paper constructs the GPSO hybrid optimization algorithm is compared with the traditional algorithm commonly used, to avoid the finite difference method and regularization method order error caused by the discretization model, so as to improve the accuracy of the inversion. At the same time, this paper GSO, PSO algorithm and GPSO hybrid algorithm inversion results did a comprehensive comparison and analysis. Provide the important theory basis for field fire detection.