In the first part of this paper, a method to build a mathematical model of weather data is proposed. The elements selected to be modeled are temperature, global solar radiation, and absolute humidity. They are decomposed into deterministic and stochastic components, which are respectively modeled as a Fourier series and a time series model, such as an ARMA model. As an example, the parameters and the characteristics of the model of Tokyo are shown with the comparison to the original record. Secondary, a new calculation method of stochastic air-conditioning load is proposed based on the assumption that both the weather record and building thermal response can be approximated as output of a liner system. It is explained that the method can estimate the mean and the variance of heat load under the condition of intermittent air-conditioning without hour by hour load simulation.
