Due to climate change and urban layout, urban waterlogging disasters are becoming increasingly severe, posing a serious threat to the stable power supply of the distribution system. In order to minimize the impact of flood disasters, it is urgent to explore urban flood disaster prediction models to achieve distribution equipment risk prediction. However, the existing hydrodynamic model-based method has high computational complexity and is difficult to guarantee the timeliness of large-scale flooding simulation forecast. The data-driven model-based method has insufficient training data, which is insufficient to meet the requirements of fast and accurate urban waterlogging warnings. To this end, a rapid waterlogging prediction model based on multimodal data fusion is proposes. This method generates training data through a hydrodynamic model to solve the problem of insufficient training data and integrates image data such as elevation maps with rainfall sequence time series data to improve prediction accuracy. Furthermore, Guilin City is used as the research object to verify the effectiveness of the proposed method. The experimental results show that the proposed method maintains high accuracy while reducing computational complexity. This method can provide a reference for risk assessment of distribution terminals.