Attributing to the influences of complex environmental factors on tides, there are often large errors in tidal prediction. To improve the accuracy of tide prediction in practical application, we apply the multi-module combinatorial prediction method to tidal prediction by taking consideration of the correlation between tidal level changes and air pressure, air temperature. By jointly modeling air pressure and air temperature data and tidal data, the empirical mode decomposition (EMD) method can extract features that affecting dynamic tidal changes. The harmonic constant method is firstly employed to predict the periodical changes of tides, then the air pressure and air temperature data are processed by the EMD method to extract the features that affecting the dynamic tidal changes. Then, the difference between the predicted value by harmonic method and the measured values was decomposed at multiple levels, and the long and short-term memory network (LSTM) model was used to predict the dynamic tidal changes between the measured tidal level and the periodical changes predicted by harmonic method. In the model, the effects of air pressure and air temperature on tides are considered and adjusted by the model parameters adaptively to accommodate the effects of complex environmental factors on tides. Finally, the tidal prediction results of the harmonic constant model are combined with those of the modular method. Tidal prediction simulation was conducted based on the Tidal stations of Canaveral and Old Port Tampa, the results show that the multi-module combination prediction method exhibits better performance than the traditional harmonic analysis method, providing a new approach for tidal prediction in areas of Marine safety management, Marine science research, etc.