Determination of quality differences and origin tracing of green tea from different latitudes based on TG-FTIR and machine learning.

Latitude differences can significantly affect the quality of tea, while in-depth research in this field is lacking. This study investigates green teas from different latitudes in China using thermogravimetric analysis coupled with infrared spectroscopy and machine learning methods. Industrial and elemental analyses were first conducted to characterize their physicochemical properties. Subsequently, pyrolysis experiments were carried out in a TG-FTIR analyzer at three heating rates (β: 10, 20, 30 ℃ min), and kinetic and thermodynamic parameters were calculated from thermogravimetric data using the model-free Kissinger, Ozawa, and Starink methods. Principal component analysis (PCA) was performed on the activation energy (Ea) values of tea samples from different latitudes for clustering analysis. Finally, linear and nonlinear predictive models of pyrolysis kinetic parameters were established using machine learning algorithms. The results showed that high-latitude teas exhibited lower mass loss percentages in the active pyrolysis zone and higher combustion efficiency compared to low-latitude teas. The average Ea values under different conversion rates (α) were higher for high-latitude teas than for low-latitude teas. PCA clustering based on Ea values effectively distinguished tea samples from different latitudes. Among the predictive models for pyrolysis kinetic parameters, the RF model based on SNV preprocessing achieved the highest accuracy, enabling the accurate prediction of pyrolysis kinetic parameters without relying on traditional formula calculations. In conclusion, this study comprehensively analyzed the differences in the pyrolysis processes of tea samples from different latitudes in China and provided new insights into tea origin traceability technology.