This study proposed a surface reaction model (SRM) and a volume reaction model (VRM), which do not rely on the reaction model functions (f(α) functions) to determine the kinetic parameters of coal oxidation. Based on the results of thermogravimetric analysis experiments for coal oxidation under different oxygen concentrations (1-21%), the activation energy (E) and pre-exponential factor (A) in the kinetics-controlled regime were determined using SRM, VRM, and the traditional model-based method (MBM) with three selected f(α) functions. It was found that the selected f(α) functions have great influence on the E obtained by MBM, with an average variation of 24%. For SRM and VRM, the decreasing oxygen concentrations (from 21 to 3%) mainly reduces the reactivity of coal oxidation by decreasing A, that is, the active sites. Moreover, the reaction rates of coal oxidation in the kinetics-controlled regime were predicted. Compared with the traditional MBM, the proposed VRM and SRM demonstrate better predictive performance for the reaction rates. Of these, SRM is the optimal model in this study, achieving an average prediction error below 10%.