Accurate detection of CH₄ is essential to prevent gas explosion and ensure safe production. However, the gas detection technology based on tunable diode laser absorption spectroscopy (TDLAS) has a large error due to temperature change. This paper explored the CH₄ detection based on TDLAS technology and the temperature compensation method, analyzed the impact of temperature on CH₄ absorption line, and finally eliminated the impact of environmental temperature on the CH₄ detection through algorithm compensation model. This study used TDLAS technology’s principle and theory to design the transmitter unit, absorption cell, signal receiver unit and data processing unit. A CH₄ detection system based on TDLAS technology was established, the concentration of CH₄ at different ambient temperatures (10~50 ℃) was measured, and the effect of temperature change on the intensity and half width at half-maximum of CH₄ absorption line at 1.653 μm was analyzed. In order to eliminate the influence of temperature on CH₄ detection and improve the compensation effect, the particle swarm optimization (PSO) was employed to optimize the optimal weight and the threshold of back propagation neural network (BPNN). The PSO-BP temperature compensation model of CH₄ was established, which overcame the characteristics of slow convergence rate and easy to fall into local optimum of the BPNN The result indicated that: (1) Based on TDLAS technology, the CH₄ detection concentration dropped with the increasement of ambient temperature, the relative error range within the whole experimental temperature was 4.25%~12.13%. The relationship between CH₄ detection concentration and temperature under different ambient temperatures can be expressed as a cubic polynomial; (2) The absorption intensity and half width at half-maximum of CH₄ gas decrease with the increase of temperature relationship between it and temperature was a monotonous decreasing function. The relative change rate of temperature on the absorption line intensity of CH₄ gas was greater than the half width. The absorption line intensity of CH₄ gas was more susceptible to the temperature change; (3) The absolute mean error (MAE) of the BPNN and PSO-BP model test samples were 12.88% and 1.81%, the mean absolute percentage error (MAPE) were 2.3% and 0.3%, the root mean square (RMSE) were 15.96% and 2.69%, and the correlation coefficient R² were 0.980 6 and 0.999 6, respectively. By establishing the PSO-BP temperature compensation model, the compensation effect was mostly distributed within the error range of ±1.0%, and MAE, MAPE, RMSE, R² and another evaluation indexes were greatly improved. It has a certain reference significance to improve the accurate detection of CH₄ in the mine with TDLAS technology.