This paper proposes a day-ahead dispatch framework of thermostatically controlled loads (TCLs) for system peak load reduction. The proposed day-ahead scheduling framework estimates the user’s indoor thermal comfort degree through the building thermal inertia modelling. Based on the thermal comfort estimation, a dayahead TCL scheduling model is formulated, which consists of 3 stages: TCL aggregator estimate their maximal controllable TCL capacities at each scheduling time interval by solving a optimization model; ` the system operator performs the day-ahead system dispatch to determine the load shedding instruction for each aggregator; and ′ the TCL aggregators schedules the ON/OFF control actions of the TCL groups based on the instruction from the system operator. A heuristic based optimization method, history driven differential evolution (HDDE) algorithm, is employed to solve the day-ahead dispatch model of the TCL aggregator side. Simulations are conducted to validate the proposed model.