Thermal stimulation methods have been investigated to enhance gas extraction from the high maturity shale recently. In this paper, a Longmaxi shale core sample, with vitrinite reflectance = 2.8%, was selected for canister tests to investigate the gas desorption process first at 55 °C and then switched to 110 °C. A mathematical model (self-similarity) was proposed to mimic the gas decaying process in nano-sized pore system. Gas adsorption/desorption incorporated advection and Knudsen flow were considered in this model. For high Knudsen number flow, the effective diffusion coefficient of the gas in the model was assumed to be a power law of density gradient. From the experiment, when heating temperature was raised from 55 °C to 110 °C, more adsorbed gas started desorbing to the porous media. It was found that the gas production rate increased significantly (20 times) and the cumulative gas production also raised considerably (11 times). By selecting an appropriate exponent m for effective diffusion coefficient, the gas production rate decayed as 1t at 55 °C and declined as 1t2 at 110 °C, which fit well with experimental data. The scaling results indicated that Knudsen flow was the dominant source of the gas flow in the whole heating process. In conclusion, Knudsen flow is the dominant gas transport form during the heating process. Thermal stimulation can significantly promote gas production rate. Moreover, the developed mathematical model can well predict the gas production rate in the heating process.