This paper proposes a distributed real-time state estimation (RTSE) method for the combined heat and power systems (CHPSs). First, a difference-based model for the heat system is established considering the dynamics of heat systems. This heat system model is further used along with the power system steady-state model for holistic CHPS state estimation. A cubature Kalman filter (CKF)-based RTSE is developed to deal with the system nonlinearity while integrating both the historical and present measurement information. Finally, a multi-time-scale asynchronous distributed computation scheme is designed to enhance the scalability of the proposed method for large-scale systems. This distributed implementation requires only a small amount of information exchange and thus protects the privacy of different energy systems. Simulations carried out on two CHPSs show that the proposed method can significantly improve the estimation efficiency of CHPS without loss of accuracy compared with other existing models and methods.