Microgrids formed locally within AC/DC distribution network possess autonomous operation capabilities. Traditional uncertainty optimization methods struggle to flexibly handle the uncertain behaviors of these special nodes with dual-source and load characteristics, making it difficult to achieve precise alignment between scheduling plans and actual operations. To address this, an adaptive robust optimization strategy for AC/DC distribution network is proposed, embedding microgrid autonomous operation into a dual-time-scale scheduling framework. First, an optimization model for the partitioned operation of an AC/DC distribution network incorporating microgrids is established. Second, an operational architecture based on adaptive robust optimization is introduced, accounting for source-load uncertainties and potential topological changes in microgrids. This framework formulates a robust scheduling plan for the worst-case scenario while dynamically adjusting energy storage, electricity procurement, and power exchange plans. It also executes autonomous microgrid grid-connected or islanding switching based on real-time system conditions. Finally, the effectiveness of the proposed method is validated using a typical case study: through coordinated distribution and microgrid operations, the method avoids congestion in the DC distribution network during midday peak hours, reduces voltage deviations, effectively mitigates the conservatism of the plan, lowers the total system cost by 4.2%, and reduces losses by 3.3%.