Abstract: The unbalanced spatial-temporal distribution of load demands in a distribution network may lead to photovoltaic generator (PVG) curtailment and load shedding. Dynamic reconfiguration provides a promising approach to solve those challenges by changing the power flow. However, the global dynamic reconfiguration model is difficult to solve and may result in large-scale power flow regulation with low incremental benefit. The two-stage optimal operation strategy based on fast identification of reconfiguration level was proposed. The identification method of reconfiguration level determined the "feeder-transformer-substation" level demand in each time period to prioritise local reconfiguration and minimize the PVG curtailment and load shedding. The segmented multi-level reconfiguration scheme was obtained after time division. Moreover, a two-stage optimal operation strategy was presented. In the first stage, the segmented multi-level reconfiguration aimed at minimizing the cost of switching operation, PVG curtailment, and load shedding. In the second stage, a stochastic programming model based on selected scenarios was established to address the uncertainties of PVGs and loads. The controllable load was optimized to make up for the deficiency of the first stage in improving power flow. Finally, the effectiveness of the identification method of reconfiguration level and two-stage optimal operation strategy was proved by the 148-bus system.