To address the challenges of scheduling complexity and uncertainty in distribution networks caused by the integration of massive distributed photovoltaics

a stochastic scheduling optimization method based on two-level clustering spatio-temporal characteristics is proposed. Firstly

spatial clustering clustering is performed by integrating the geographical location and electrical distance of PV nodes to characterize the spatial coupling relationship. Secondly

temporal clustering is conducted based on dynamic time warping (DTW) distance and historical output curves to extract typical temporal characteristics. Finally

stochastic scenarios are constructed using typical curves obtained from clustering and error bounds

and a stochastic optimal scheduling model for distribution networks is established

which coordinates generation cost

curtailment penalty

and carbon emission cost. Simulation results show that the proposed method significantly improves calculation efficiency while ensuring scheduling accuracy. The decision time is reduced from 477.73 s to 22.17 s

the total operating cost is reduced by 3.17%

and the PV utilization rate reaches 99.62%. Furthermore

the method effectively reveals the boundaries between system economics and accommodation capacity under different penetration levels.