DWT-LSTM combined tidal prediction model

Accurate tidal data are important for coastal city planning and maritime vessel navigation safety, but traditional tidal reconciliation and analysis methods require a large amount of data, while the more commonly used interpolation methods can only analyse the existing data and cannot predict future data. In order to solve the above problems, a tidal prediction model based on long short-term memory network (LSTM) is proposed. The tidal data from Weizhou Island, China and Clearwater Beach, USA were selected for experimental validation, and the R
²
at Weizhou Island reached 0.9741, while the RMSE was 0.14m, with a maximum prediction error of 0.27m. The R
²
at Clearwater Beach reached 0.9596, while the RMSE reached 0.05m, with a maximum error of 0.31m. Therefore, the model can effectively improve the accuracy of tidal prediction and provide a solution for tidal prediction in different sea areas.