Abstract: The tidal levels in tidal river reaches exhibit complex spatiotemporal variations, making accurate simulation challenging. In order to improve the simulation accuracy, a tide level simulation method for tidal reach considering multiple influencing factors is proposed. This method uses the derivative dynamic time warping (DDTW) technique to determine the average propagation time of tidal levels between upstream and downstream locations. The ensemble empirical mode decomposition (EEMD) is utilized to decompose the various influencing factors data of tidal levels, extracting high-frequency and low-frequency component information. By combining deep residual shrinkage networks (DRSN), bidirectional long short-term memory networks (BILSTM), and attention mechanisms, an intelligent model (DRSN-BILSTM-Attention) is developed, which effectively captures the spatiotemporal features of the decomposed components. The Tianshenggang tidal level station on the Yangtze River was used as the research focus. Results show that the proposed method significantly improves the precision of tidal level simulation, achieving an average accuracy rate of 92.1% for three-hourly simulations. Compared to single machine learning approaches, the proposed method demonstrates superior capabilities in extracting spatiotemporal data features. This method offers a basis for flood prevention, disaster mitigation, and water regulation in coastal areas.