Proposes a cooling load prediction modeling method based on multivariate nonlinear regression for shopping malls. Describes the cooling load characteristics of the shopping malls at different time using actual energy consumption coefficient. Establishes a prediction model and increases the prediction modeling speed while avoids the fluctuation of predicted values accuracy due to incomplete training samples. Carries out a simulation experiment with the measured cooling load data of a shopping mall in Guangzhou. Obtains a multivariate nonlinear fitting equation between cooling load and its influencing factors. Reveals the influence of various factors on cooling load, and the simulation results demonstrate the effectiveness and feasibility of this method.