Building energy models were widely used in building energy performance analysis and efficiency improvement, such as building demand response assessment and energy system optimization. However, building energy modeling was a time-consuming and complex process. Therefore, this paper proposed a hybrid building energy modeling method based on parameterized prototype models and rapid calibration. The main feature of the method was the rapid generation of target building energy models using the prototype building energy performance database. The parameterized prototype building energy model considered 13 uncertainty parameters. Based on the range of 13 uncertainty parameters, such as wall U-value and solar heat gain coefficient, 1000 simulations were performed by Monte Carlo sampling to generate the prototype building energy performance database. The target building energy model was quickly calibrated by learning from this database. It filtered the models that met the requirements based on the actual building measurement data. Based on this hybrid modeling method, the rapid modeling tool AutoBPS-Hybrid was developed in a ruby environment. Shopping mall buildings located in three different climate zones were selected as case studies for this study. The results showed that if only one building energy model meeting the percentage errors was needed, only 3–4 simulations were required. If it was necessary to match the real uncertainty parameter distributions, an average of about 53 simulations was needed. The building energy models were applied to plug load and lighting control in buildings. The shopping mall buildings in Harbin, Beijing and Chengdu could reduce energy consumption by 10.18–13.33 kWh/m2, 14.43–18.15 kWh/m2 and 11.54–14.69 kWh/m2 per year, respectively.