关键词: 钱塘江涌潮;潮时预报;深度学习;卷积神经网络;BP神经网络;

Abstract: Accurate tidal time forecast of Qiantang River tidal bore is of great significance to the management of moisture-proof safety，tidal sightseeing，the development and utilization of water resources，transportation and navigation，as well as the construction of wading projects.The traditional empirical prediction method based on similar tide analysis can not accurately analyze the influence of some short-term random factors，and the prediction accuracy has some uncertainty.Therefore，a tidal time forecast method for Qiantang River tidal bore based on two-dimensional convolutional neural network is proposed.The influence factors such as lunar calendar date，wind power and direction，upstream water inflow，river channel topography，and the time difference between the previous day and the next day were selected as feature vectors to build a feature set.A two-dimensional convolutional neural network is adopted to extract effective features from the input feature matrix，and a fully connected neural network is used to establish the tidal time forecast model of Qiantang River tidal bore.The particle swarm optimization algorithm is used to optimize the parameters and realize the interval day prediction of Qiantang River tidal time.Finally，the tidal time forecast results at Cangqian Station，a representative station of Qiantang River，from 2009 to 2017 showed that the proposed tidal time forecast model of Qiantang River based on convolutional neural network has high prediction accuracy and reliability.

Key words: Qiantang River tidal bore；tidal time forecast；deep learning；convolutional neural network；BP neural network