Quantification of wave-current-turbulence interactions through numerical modelling and data-driven method for ocean energy applications

Abstract

Wave-current interactions (WCI) play a critical role in shaping wave and tidal current energy resources, yet their neglect can lead to significant over- or underestimations, misrepresenting the complexities of the flow environment. Despite extensive theoretical exploration of WCI mechanisms, their application to real-world ocean settings remains limited. Long-term assessments using numerical models with two-way coupling of wave and tidal currents are rare, leaving gaps in understanding WCI under realistic conditions. Additionally, the quantification of turbulence enhanced by waves at actual sites is poorly understood. While machine learning has been widely applied to general wave predictions, it has yet to address parameters influenced by WCI. This PhD research addresses these challenges by investigating the impact of WCI on wave and tidal parameters in the Pentland Firth and Orkney Waters (PFOW), Scotland, UK, through three complementary methodologies: numerical modelling, wave-current decomposition, and machine learning. Numerical modelling was employed to analyse the effects of tidal currents on wave parameters and wave energy resources. Two models were developed: (1) a North Atlantic scale TOMAWAC wave-only model, which provided boundary conditions and wave parameters unaffected by tidal currents, and (2) a regional two-way coupled TOMAWAC-TELEMAC wave-current model, which simulated wave parameters accounting for WCI. Validation of both models was achieved using field measurements, including 10 years of data from Cefas WaveNet buoys (for the wave-only model) and 135 days of site measured Acoustic Wave and Current Profiler (AWAC) and Acoustic Doppler Current Profiler (ADCP) data (for the wave-current model). After that, eight representative sites across PFOW were analysed to assess tidal effects on wave parameters under Spring and Neap tides, Flood and Ebb phases. The enhanced wave breaking due to strong currents was frequently observed. Furthermore, a 10-year simulation (2014–2023) of both models produced wave maps incorporating tidal effects, revealing spatiotemporal variations in WCI phenomena across interannual, seasonal, and monthly scales. The second focus was the use of a novel wave-current decomposition method to quantify turbulence levels enhanced by waves. A side information assisted Empirical Mode Decomposition (EMD) method was introduced to separate wave and current components from combined velocity data. The validity of this method was demonstrated by comparing the derived wave heights and current velocity spectra with field measurements and theoretical benchmarks. The method was applied to ADCP data from three locations in PFOW, enabling the calculation of turbulence intensity (TI) for wave-only, current-only, and wave-current conditions across varying current velocities and wave heights. This analysis provided a comprehensive quantification of three-dimensional turbulence levels enhanced by waves in streamwise, transverse, and vertical directions. An empirical relationship between wave-induced, current-induced, and wave-current coupled turbulence was also proposed, offering a practical tool for estimating wave-induced turbulence levels. The final focus of the research involved machine learning methods to predict wave parameters. For deep-water, open-sea regions around northern Scotland where tidal currents are negligible, the spatiotemporal relationship between wind and waves was modelled using the Informer deep neural networks and the XGBoost machine learning algorithm. Ten years (2012–2021) of hourly wind data from ECMWF ERA5 and wave parameters from Cefas WaveNet buoys were used for training and verification, enabling accurate wave predictions for 2022. Models for typical and extreme weather conditions were developed to enhance prediction accuracy. Additionally, at PFOW regions where WCI are significant, the Informer model was used to predict waves under tidal effects. Input features were derived from the previously mentioned North Atlantic scale wave model and the regional scale wave-current model. Training on 2016 data enabled accurate predictions of 2017 wave conditions across different sites, demonstrating the model’s capability to capture wave-current interactions effectively. Overall, this thesis integrates numerical modelling, wave-current decomposition, and machine learning to provide a multifaceted quantification of WCI in real-world settings. Their interdependence on shared datasets underscores their internal synergy. The findings offer valuable insights and tools for addressing challenges in ocean engineering, particularly for wave energy development in wave-current environments, while providing extensive and robust datasets for future research.Wave-current interactions (WCI) play a critical role in shaping wave and tidal current energy resources, yet their neglect can lead to significant over- or underestimations, misrepresenting the complexities of the flow environment. Despite extensive theoretical exploration of WCI mechanisms, their application to real-world ocean settings remains limited. Long-term assessments using numerical models with two-way coupling of wave and tidal currents are rare, leaving gaps in understanding WCI under realistic conditions. Additionally, the quantification of turbulence enhanced by waves at actual sites is poorly understood. While machine learning has been widely applied to general wave predictions, it has yet to address parameters influenced by WCI.

This PhD research addresses these challenges by investigating the impact of WCI on wave and tidal parameters in the Pentland Firth and Orkney Waters (PFOW), Scotland, UK, through three complementary methodologies: numerical modelling, wave-current decomposition, and machine learning.

Numerical modelling was employed to analyse the effects of tidal currents on wave parameters and wave energy resources. Two models were developed: (1) a North Atlantic scale TOMAWAC wave-only model, which provided boundary conditions and wave parameters unaffected by tidal currents, and (2) a regional two-way coupled TOMAWAC-TELEMAC wavecurrent model, which simulated wave parameters accounting for WCI. Validation of both models was achieved using field measurements, including 10 years of data from Cefas WaveNet buoys (for the wave-only model) and 135 days of site measured Acoustic Wave and Current Profiler (AWAC) and Acoustic Doppler Current Profiler (ADCP) data (for the wave-current model). After that, eight representative sites across PFOW were analysed to assess tidal effects on wave parameters under Spring and Neap tides, Flood and Ebb phases. The enhanced wave breaking due to strong currents was frequently observed. Furthermore, a 10-year simulation (2014–2023) of both models produced wave maps incorporating tidal effects, revealing spatiotemporal variations in WCI phenomena across interannual, seasonal, and monthly scales.

The second focus was the use of a novel wave-current decomposition method to quantify turbulence levels enhanced by waves. A side information assisted Empirical Mode Decomposition (EMD) method was introduced to separate wave and current components from combined velocity data. The validity of this method was demonstrated by comparing the derived wave heights and current velocity spectra with field measurements and theoretical benchmarks. The method was applied to ADCP data from three locations in PFOW, enabling the calculation of turbulence intensity (TI) for wave-only, current-only, and wave-current conditions across varying current velocities and wave heights. This analysis provided a comprehensive quantification of three-dimensional turbulence levels enhanced by waves in streamwise, transverse, and vertical directions. An empirical relationship between wave-induced, current-induced, and wave-current coupled turbulence was also proposed, offering a practical tool for estimating wave-induced turbulence levels.

The final focus of the research involved machine learning methods to predict wave parameters. For deep-water, open-sea regions around northern Scotland where tidal currents are negligible, the spatiotemporal relationship between wind and waves was modelled using the Informer deep neural networks and the XGBoost machine learning algorithm. Ten years (2012–2021) of hourly wind data from ECMWF ERA5 and wave parameters from Cefas WaveNet buoys were used for training and verification, enabling accurate wave predictions for 2022. Models for typical and extreme weather conditions were developed to enhance prediction accuracy. Additionally, at PFOW regions where WCI are significant, the Informer model was used to predict waves under tidal effects. Input features were derived from the previously mentioned North Atlantic scale wave model and the regional scale wave-current model. Training on 2016 data enabled accurate predictions of 2017 wave conditions across different sites, demonstrating the model’s capability to capture wave-current interactions effectively.

Overall, this thesis integrates numerical modelling, wave-current decomposition, and machine learning to provide a multifaceted quantification of WCI in real-world settings. Their interdependence on shared datasets underscores their internal synergy. The findings offer valuable insights and tools for addressing challenges in ocean engineering, particularly for wave energy development in wave-current environments, while providing extensive and robust datasets for future research.