Thermostatically controlled loads (TCLs) on the demand side have been a vital energy resource in smart grids. To efficiently utilize the large-scale TCLs and enhance the flexibility of micro-community systems, this paper proposes a distributed coordinated control strategy based on the distributed model predictive control (MPC). To achieve the adaptive coordinated control among TCLs and consider user comfort constraints, a distributed dual-layer internal control strategy based on MPC is established on a scalable communication network. This strategy achieves the efficient utilization of TCLs in a distributed manner and notably improves the convergence speed through sparse network communication between neighbors. For external resource utilization of TCLs, a multi-timescale scheduling framework is proposed to realize the pre-allocation of electricity. Furthermore, the feasibility of the proposed distributed coordinated control strategy is confirmed through comparative case analysis.