Abstract: Spontaneous combustion of coal is one of the common and extremely hazardous concealed disasters in underground coal mines. Its early stage is accompanied by a slow temperature rise, and conventional electrical or point-type temperature monitoring methods are hardly capable of detecting weak thermal signals with high precision. To effectively monitor the temperature during the early stage of coal spontaneous combustion, a temperature measurement method based on the thermal deflection effect of a bi-material cantilever beam is proposed. A bi-material cantilever beam is composed of two materials with different thermal expansion coefficients, and its free end can generate thermally induced displacement. The deflection of this thermally induced displacement of the bi-material cantilever beam is detected on the basis of the Fabry-Perot interferometry principle, thereby realizing temperature inversion. A uniform temperature field was applied within the temperature range of 0 ℃ to 150 ℃. The Abaqus finite element software was adopted to conduct thermal-mechanical structural coupling simulations on the rectangular bi-material cantilever beam model, focusing on parameters such as coating thickness, drilling parameters of the substrate layer and the coating layer, as well as different structural forms of the cantilever beam. The simulation results indicate that the “zigzag-shaped” beam structure exhibits the optimal performance, with a temperature sensitivity of 0.576 μm/℃. Based on the established Fabry-Perot interferometric displacement measurement platform, the calibration tests on the thermal deflection displacement of the bi-material cantilever beam were carried out. Analysis was conducted on the relationship between the cavity length at the free end of the cantilever beam and the fiber movement distance, as well as the stability of temperature-displacement. The linear relationship between the fiber movement distance and the cavity length variation was verified. The verification results show that the root-mean-square error of fitting is ±0.15 μm and the displacement standard deviation in the repeatability tests is 0.90 μm. Combined with the simulated sensitivity of the “zigzag-shaped” beam, the temperature measurement error of the system can reach ±0.26 ℃ within the range of 0-150 ℃, which verifies the high precision and stability of the system in micro-displacement measurement. This method features passivity, fast response speed and high sensitivity. It can be easily networked via optical fibers, making it suitable for temperature monitoring in the initial stage of coal spontaneous combustion. It thus provides an effective technical means for the early warning of coal spontaneous combustion.