Objective: Understanding the pyrolysis characteristics of tobacco powder is crucial to controlling the formation of harmful substances and enhancing the safety and quality of heated cigarettes, thus providing a scientific basis for optimizing the design of heated cigarettes. Although the extant studies mainly explored the pyrolysis characteristics and product distributions of some tobacco powders, they barely considered the development of a reliable, component-specific kinetic model for the pyrolysis of tobacco powder in heated cigarettes. To address this research gap, we explored the pyrolysis process and the kinetic parameters of tobacco powder in heated cigarettes, followed by the development of a reliable model. Methods: First, employing thermogravimetric (TG) analysis, the thermal weight loss of the sample (tobacco powder, 10±0.02 mg) was studied under a purge gas (nitrogen) at a flow rate (50.0 mL/min). Following a pre-dehydration step (heating to 100 ℃, holding for 30 min, and cooling to room temperature), the sample was heated from 50 ℃ to 500 ℃ at four heating rates: 5, 10, 15, and 20 ℃/min. Next, the resulting TG and derivative TG (DTG) curves were recorded. Second, we calculated the apparent activation energy for the pyrolysis of tobacco powder using three model-free fitting methods: the Kissinger-Akahira-Sunose (KAS), Friedman (FR), and Flynn-Wall-Ozawa (FWO) methods. Finally, we proposed a four-step pseudo-component kinetic model covering the four-stage pyrolysis of tobacco powder: water evaporation, hemicellulose pyrolysis, cellulose pyrolysis, and lignin pyrolysis. Additionally, a genetic algorithm was employed to optimize 19 kinetic parameters (e.g., activation energy, pre-exponential factor, reaction order, and stoichiometric coefficient) in the model to considerably minimize the error between calculated and experimental DTG values. Results: Tobacco powder pyrolysis proceeded in four stages, illustrated using the 20℃/min heating rate: 1) dehydration (50—208 ℃), where the mass loss (8.73%) was mainly due to the evaporation of bound water in the tobacco powder; 2) hemicellulose pyrolysis (208—291 ℃), where the mass loss reached 19.20% owing to hemicellulose decomposition; 3) cellulose pyrolysis (291—372 ℃), where the mass loss was 22.50%, accounting for the maximum weight-loss rate (0.381 wt%/℃) at 325 ℃; 4) lignin pyrolysis (372—500 ℃), where the mass loss was 14.40%, marking a gradual decrease in the weight-loss rate owing to the complex structure and high thermal stability of lignin. Notably, the 10 ℃/min heating rate yielded the highest weight-loss rate (0.390 wt%/℃) and lowest residual mass (35.0%). Furthermore, the apparent activation energy calculations were highly reliable, with the correlation coefficients (R²) of all three model-free fitting methods exceeding 0.97. Particularly, the KAS, FWO, and FR methods yielded activation energy ranges of 200.89—519.66, 198.11—505.40, and 198.17—505.53 kJ/mol, respectively, exhibiting high correlation throughout the process. As the conversion rate increased, the activation energy exhibited a "first up, then down, then up again" trend. In detail, the activation energy stabilized at approximately 200 kJ/mol when the conversion rate was less than 0.1 (the dehydration stage), then increased rapidly to 249 kJ/mol as the conversion rate increased from 0.1 to 0.3 (hemicellulose pyrolysis). It stabilized again at approximately 250 kJ/mol at a conversion rate of 0.3—0.5 (cellulose pyrolysis) before decreasing to 219 kJ/mol as the conversion rate reached 0.6 (initial lignin pyrolysis). Afterward, it increased rapidly as the conversion rate exceeded 0.6 (lignin decomposition into phenols and aromatic compounds). The kinetic model, optimized using the genetic algorithm, converged after 120 iterations (with a deviation of 5.00%). Notably, the optimized activation energies for the four pseudo-components were 62.00 kJ/mol (water), 118.83 kJ/mol (hemicellulose), 217.31 kJ/mol (cellulose), and 192.13 kJ/mol (lignin). Further, the model accurately simulated the experimental data, with the TG and DTG curves achieving high R² values (>0.94) for all four heating rates. Conclusions: We clarified the four-stage pyrolysis characteristics of tobacco powder in heated cigarettes. We established that the three model-free fitting methods can be used to reliably calculate the conversion-rate-dependent apparent activation energies for the pyrolysis of tobacco powder, reflecting the differences in energy requirements across the pyrolysis stages. Furthermore, our kinetic model can accurately simulate the pyrolysis of tobacco powder, providing key theoretical support for improving the quality of heated cigarettes.