Abstract: With the rapid development of microgrid systems with multiple intelligent buildings integrated, thermostatically controlled loads（TCLs） have gradually become an important demand-response resource. However, its distribution dispersion and response uncertainty also put forward higher requirements for the regulation. To address the distribution dispersion of TCLs, this paper puts forward a TCLs aggregation scheme based on the building’s resistance-capacitance（RC） model. Specifically, the RC heat transfer model is linearized at first, and then the TCLs are aggregated according to the law of large numbers, thence the building’s physical structural parameters can be taken into account. This TCLs aggregation scheme not only adapts to the differentiated modeling requirements of buildings but also has preferable calculation efficiency. On the basis of the proposed TCLs aggregation scheme, a two-stage robust optimization model applicable to the integrated building microgrid system is established, which considers the uncertainties of aggregated TCLs and distributed renewable generators, and is solved by the column and constraint generation algorithm. Results show that the proposed TCLs aggregation scheme can reach the accuracy requirements of microgrid scheduling operation under diverse outdoor temperatures and solar radiation. Meanwhile, the established two-stage robust optimization model can effectively reduce the power regulation of tie lines under real circumstances, thence enhancing the stability and robustness of the microgrid system.