The collision reaction of H + O2 = OH + O is a pivotal step in combustion. To investigate the influence of singlet oxygen on this reaction, we computed potential energy surfaces (PESs) for all six lowest states using high-level ab initio methods and coupled them with embedded atom neural network (EANN) fitting. By integrating quasi-classical trajectory (QCT) with trajectory surface hopping (TSH) based on the fitted PESs, we simulated the dynamics of both ground- and excited-states to derive the reaction rate constants for the forward and reverse processes. The results reveal that the forward reaction facilitates radical generation, promoting combustion reactions. Furthermore, calculations of reverse reaction rate constants indicate that all electronic states ultimately yield ground-state oxygen, leading to radical deactivation and exerting an inhibitory effect on combustion processes.