Real-time Tide Prediction Based on An Hybrid HA-WANN Model Using Wind Information

Observed tidal level is a kind of signals in essence. The fluctuations of tide affect the maritime activities and human lives in offshore and coastal zone, including the waterfront distances in shallow waters, islands, and the lands with gentle slope, especially in locations having shallow water depths. Harmonic analysis is traditionally employed for long-term tide predictions, but it becomes unreliable if non-periodic meteorological events predominate. In addition of other signal processing methods, this study proposed a hybrid model for hourly tidal level forecasting by combining ANN with wavelet decomposition using SLA and wind shear velocity components as inputs. Six mother wavelets and nine data sets are computed in HA-WANN model in order to acquire the best models, and CE, r
²
, MAE, RMSE and MSRE indexes are used to measure the superiority of all models. Results obtained from this study indicate that the conjunction of wavelet decomposition and artificial intelligence models can definitely improve the accuracy of tidal level prediction, and the usage of SLA and wind shear velocity can yield better capability than HA and original ANN models.