The quantification of uncertainty in aggregated distributed photovoltaic power generation is crucial for decision control in power systems. A data-driven aggregation quantification method for power uncertainty in distributed photovoltaic power generation was proposed. Firstly, considering the spatial-temporal distribution characteristics of distributed photovoltaics, an uncertainty aggregation model based on spatial-temporal graph convolutional neural networks was proposed, which can effectively mine the spatial-temporal characteristics of data. Secondly, aggregated uncertainty was quantified and characterized without assuming its probability distribution through the combination of spatial-temporal graph convolutional neural networks with quantile regression models. In addition, considering the cross problem of quantiles, a quantile mapping method to avoid the cross problem of quantized results with aggregated uncertainty was proposed. Finally, the effectiveness of the proposed method was verified based on the IEEE-33 bus system.