The change of humidity and temperature field of concrete is closely related to its long-term time-varying performance of shrinkage and creep. Therefore, accurate prediction of temperature and humidity in concrete plays an important role in studying the service performance of concrete structures. In the study, the mechanism of coupled moisture and heat (M&H) transfer in concrete is systematically analyzed from three scales: micro-nano scale, mesoscale, and macroscale; meanwhile, a multi-scale theoretical model of coupled M&H transfer in concrete is established. In addition, a novel simulation method of bidirectional coupled of M&H is proposed and the corresponding ABAQUS subroutine is compiled. Based on the existing experiments, the concrete temperature and humidity data for 180 days were further monitored, which verified the feasibility of the proposed method. Finally, the prediction results of M&H transfer are analyzed and compared in three cases: uncoupling, unidirectional coupling, and bidirectional coupling. For the temperature prediction, the early hydration temperature rise is underestimated under the uncoupled model and the unidirectional coupled model that only considering the influence of temperature on humidity, and the prediction error of unidirectional coupled model that only considering the influence of humidity on temperature is relatively small. For the humidity prediction, both uncoupled and unidirectional coupled models underestimate the water consumption, and the closer to the surface of the concrete, the greater the error. Through experiments and numerical simulation analysis, it is suggested that the bidirectional coupled effect should be considered in the analysis of wet-heat field of concrete. This study provides a new way for the modeling and simulation of coupled M&H of concrete, and lays the foundation for the long-term performance research of infrastructure.