With the intensification of the global energy crisis, the proportion of renewable energy in the power system is increasing. Demand response can enhance the regulation capacity of new power systems by introducing flexible controllable loads. In this paper, the control problem of flexible thermostatically controlled loads (TCLs) in demand response is studied, and a distributed hierarchical control strategy is proposed based on a multi-agent consensus algorithm. In this strategy, the optimal power distribution problem for the TCL aggregator is considered in the upper optimization layer, and the distributed coordination mechanism for TCLs is considered in the lower coordination layer. The convergence of the upper distributed optimization algorithm and the lower coordinated control algorithm is proved theoretically for the uncertainty problem of users′ electricity consumption behavior. Numerical simulation experiments are carried out in MATLAB environment to verify the effectiveness of the proposed strategy. Compared with the centralized control methods, the proposed distributed hierarchical control strategy can reduce the communication burden of the manage unit of demand response event and protect users' privacy.