Key points are not available for this paper at this time.
Waves are emerging as a renewable energy resource, but the harnessing of such energy remains among the least developed in terms of renewable energy technologies on a regional or a global basis. To generate usable energy, wave heights must be predicted in near-real-time, which is the driving force for wave energy converters. This study develops a hybrid Convolutional Neural Network-Long Short-Term Memory-Bidirectional Gated Recurrent Unit forecast system (CLSTM-BiGRU) trained to accurately predict significant wave height (Hsig) at multiple forecasting horizons (30 minutes, 0.5H; 2 hours, 02H; 3 hours, 03H and 6 hours, 06H. In this model, convolutional neural networks (CNNs), long-short-term memories (LSTMs), and bidirectional gated recurrent units (BiGRUs) are employed to predict Hsig. To construct the proposed CLSTM-BiGRU model, historical wave properties, including maximum wave height, zero-up crossing wave period, peak energy wave period, sea surface temperature, and significant wave heights are analysed. Several wave energy generation sites in Queensland, Australia were tested using the hybrid deep learning CLSTM-BiGRU model. Based on statistical score metrics, scatterplots, and error evaluations, the hybrid CLSTM-BiGRU model generates more accurate forecasts than the benchmark models. This study established the practical utility of the hybrid CLSTM-BiGRU model for modelling Hsig and therefore shows the model could have significant implications for wave and ocean energy generation systems, tidal or wave height monitoring as well as sustainable wave energy resource evaluation where a prediction of wave heights is required.
Ahmed et al. (Thu,) studied this question.