The variability and fluctuations in distributed photovoltaic power generation due to changes in weather，seasons，and pollution or shading of photovoltaic components pose challenges. Addressing the randomness and volatility issues in photovoltaic power generation，this paper proposes an optimization of the rescheduling strategy for a photovoltaic power generation system based on big data mining. The objective is to reduce the probability of branch flow overload，utilizing the overload probability index to characterize system risks. Data from photovoltaic sensors，including the all-day light sensor Lufft WS501 and precision spectral average radiometer Eppley，are collected and recorded. Leveraging big data mining techniques，an iterative optimization scheduling model is employed to mitigate the risk of power congestion. Simulation verification on the IEEE39 nodes demonstrates that，compared to traditional scheduling，the overload probability through branch 26 decreases from 0.444 0 to 0.013 8，and through branch 15 decreases from 0.447 0 to 0.032 7，significantly reducing the overload probability. This approach accurately collects photovoltaic generation data，improves the reliability of the power system，and provides an effective method for addressing the randomness and fluctuations in photovoltaic power generation，holding crucial significance for the stable operation of power systems.