<p>The uncertainties of renewable distributed generation power output and load demand have created significant challenges to the scheduling of grid-connected microgrids and have restricted the carbon emission reduction capability of microgrid. To address the problems posed by source-load uncertainties of microgrid and carbon emissions on scheduling, this paper proposes a two-stage robust optimal scheduling model for microgrid based on data-driven segmented multi-interval uncertainty sets. Firstly, the source-load predicted power and error data are segmentally fitted based on a Gaussian mixture model, which is combined with a box-type uncertainty set to establish a multi-interval uncertainty set for the power partition segments, which is used to characterize the wind, photovoltaic and load uncertainties more accurately; then, a robust optimal scheduling model for microgrid considering segmented multi-interval uncertainty sets and carbon emissions is constructed to minimize the microgrid operation cost and carbon emission cost. The model is solved using an improved Columns and Constraints Generation (CC&amp;G) algorithm. Finally, the effectiveness of the proposed model is verified by case studies. The results show that the total cost of the proposed method is reduced by 8.15%, 3.13%, and 12.10% compared to single-interval RO, multi-interval RO, and ellipsoidal RO, respectively, and the solution time is reduced by about 100 s compared to SO. This indicates that the proposed model can better balance the economy and robustness of MG. In addition, the robustness and economy of the scheduling scheme can be flexibly adjusted by choosing appropriate uncertain budget parameters as needed during the actual formulation of the scheduling scheme.</p>