Disclosed is a method for predicting the cooling load for efficient operation of a heat accumulation system by obtaining a prediction function regarding outdoor air temperature and specific humidity from meteorological office data, predicting the outdoor air temperature and specific humidity by using the prediction function and the highest and lowest temperatures of the weather forecast, and predicting the cooling load based on the sensible heat load coefficient, outdoor air coefficient, sensible heat load constant, and latent heat load constant, which are obtained from the building design data. The cooling load can be predicted without using a complicated mathematical model and with no reference to past operation data regarding the target building, but solely based on four air-conditioning design values of the building and the highest and lowest temperatures of the next day, which can be easily obtained from the weather forecast of the meteorological office.