Evaluation of HVAC performance in an office room bya virtual scenario is important when performing energysavinganalysis or peak power analysis. When evaluatingHVAC power by calibration, the framework in whicha simulator is made and simulation parameters are optimized,it is indispensable to minimize the engineeringcost while achieving high simulation speed and high accuracy.The aim of this paper is to propose a low-costHVAC calibration framework by using a small number ofeasily available sensor data. First, the HVAC power modelat an office room is approximated with a linear heat balanceequation. When multiple temperature time series inthe room can be used as input, the heat balance equationbecomes redundant. Thus, we propose a regression modelto solve the problem. According to the temperature-basedHVAC calibration results at a test room, the proposedmethod achieves results that are comparable to those ofconventional simulated annealing methods but more stable.