Abstract: The 12.5 m deep-water channel in the Yangtze Estuary has brought great economic and social benefits into play.At the same time, the problems of large amount of channel back-siltation and its highly concentrated temporal and spatial distribution are prominent, and the pressure of channel water depth maintenance is large.A large amount of maintenance and dredging force needs to be invested every year.The monthly back-siltation intensity of the deep-water channel in the Yangtze Estuary is large and changes obviously with time and space.Therefore, how to accurately predict the channel back-siltation is an important technical problem.Based on the measured hydrological data from 2016 to 2018,this paper screened the main influencing factors of the channel back-siltation, and established a BP neural network high-precision prediction model of the deep-water channel back-siltation in the Yangtze River Estuary under the action of multiple influencing factors.The number of hidden layers and the number of neurons in each layer of the training and prediction network are compared and recommended.The hydrological data of the long series of the Yangtze River Estuary from 2016 to 2018 are selected for prediction model training, and the data of 2019 is selected to verify the prediction model, which verified that the model has a high prediction ability and spatial distribution prediction accuracy of channel back-siltation.The results will provide important support for the scientific management of channel maintenance and the rational scheduling of dredging ships.