Hybrid FEDformer-LSTM model for enhanced heave displacement prediction in offshore buoys

Buoys are a crucial structure used offshore, and the data acquired from them is essential for marine navigation, offshore engineering, coastal management, weather forecasting and wave energy research. To optimize wave energy extraction and guarantee the dependability of ocean-based structures, accurate heave displacement forecasting is essential. In order to enhance the estimation of heave displacement, a unique hybrid model that combines a Long Short-Term Memory (LSTM) network with the Frequency Enhanced Decomposition Transformer (FEDformer) is implemented. The proposed FEDformer – LSTM hybrid model competently captures long-range dependencies and non-linear temporal patterns in wave data by employing the frequency-domain decomposition powers of FEDformer and the temporal learning advantages of LSTM. Experimental data are retrieved from the buoy data of the National Institute of Ocean Technology (NIOT), which includes wave height, wind speed, and other data from key maritime areas. The hybrid model beats state-of-the-art forcasting algorithms and independent deep-learning techniques in terms of correlation metrics, Mean Absolute Error (MAE) and Root Meam Square Error (RMSE), affording to proportional tests carried out using real-world buoy datasets. The findings indicate that the FEDformer-LSTM model is more appropriate prediction model for the proposed application.