With the increasing complexity of distribution network topology, there is a problem that a high proportion of renewable energy output prediction errors will lead to a decline in the accuracy of distribution network optimal scheduling decisions. Thus a two-stage secure and low-carbon operation optimization method considering multi-level reconfiguration is proposed. First, the influence of two kinds of multi-level switches on reconfiguration and two-stage operational framework of a distribution network are analyzed. Secondly, a day-ahead and intra-day two-stage operation optimization method is proposed to realize an optimal operation solution by coordinating the multi-level switches’ state, the energy storage system and industrial load demand response. Then, the day-ahead global low-carbon economic scheduling and intra-day rolling security economic scheduling models are constructed by considering the photovoltaic output prediction errors over different time scales. Finally, numerical examples verify that the proposed model can effectively support the secure and low-carbon operation of complex distribution networks with high photovoltaic penetration.