Abstract: Usually, only a very limited number of buildings in a region are equipped with structural monitoring arrays. In order to improve the accuracy of earthquake damage simulation for building groups with limited monitoring data, a framework for updating the model parameters of the building groups based on Bayesian parameter learning algorithm was established. In this framework, the model parameters of all buildings are updated through updating the empirical calculation formula of the model parameters based on the limited monitoring data. In this study, the fundamental period that best represents the dynamic characteristics of the structure was selected, and the updating algorithm for the fundamental period of the building groups was derived under the above framework. A parameter analysis of the updating method for the fundamental period was conducted by using a virtual building group. And the proposed method was further validated by using measured data from real building groups. The results show that this method can effectively improve the accuracy of the estimated fundamental period of the building groups; The proportion of monitoring buildings that meet the minimum requirements is closely related to error of the initial average value of empirical formula parameters, the initial coefficient of variation of empirical formula parameters, and the standard deviation of empirical formula errors. It is recommended that the proportion of monitoring buildings be set at a minimum of 4%. This study provides a feasible method for calibrating and updating building group models based on the data of limited number of monitored buildings.