Abstract: In the operation and management of photovoltaic charging stations, capacity optimization configuration is mainly based on user charging behavior, ignoring the randomness of photovoltaic power generation, resulting in poor peak shaving and valley filling effects. Therefore, a multi-objective optimization configuration method for photovoltaic storage and charging stations considering photovoltaic uncertainty is proposed. Using the improved K-means method to cluster historical photovoltaic data and generate typical photovoltaic output scenarios considering photovoltaic uncertainty. Starting from three aspects: Economic operating costs, annual abandonment and loss of load costs, and compensation for peak and valley power supply on the grid side in uncertain scenarios, set optimization configuration goals for photovoltaic charging stations. Finally, the third-generation non dominated sorting genetic algorithm is used for search to obtain optimization configuration decisions that meet the optimization objectives. The experimental results show that using this method to optimize the configuration of photovoltaic charging stations reduces the peak valley difference of charging load to 750 kW.