Learning tidal currents
Lead Research Organisation:
University of Oxford
Department Name: Engineering Science
Abstract
Tides occur due to the changing gravitational movement of the Moon and Sun relative to the Earth. As astronomical movements are highly predictable the tides should also be predictable. This is one of the key advantages of tidal stream energy (a rapidly developing source of renewable energy). The existing methods which are used to predict tidal movements perform very well for predicting water levels and slow moving currents, but often perform very badly on fast flowing tidal streams of the type in which we areinteresting in placing tidal turbines. This project will address this by applying methods from the machine learning community to the analysis of fast flowing tidal streams. This will produce an algorithm which will allow users from the oceanographic and tidal energy community to greatly improve the prediction of tidal currents at any point indefinitely far into the future. Thus a robustprediction of the performance of tidal stream turbines can be obtained. In the rapidly growing area of tidal stream energy, accurate knowledge of the tidal currents is vital for: robust predictions of energy yield; for the calculation of loads and the design of the turbine; and to give confidence to investors.
Planned Impact
This proposal aims to address one of the key areas of uncertainty preventing the development of tidal stream energy. The method currently used to predict tidal currents (harmonic analysis) has been shown to be inadequate for analysis of many candidate sites for tidal stream turbines. This has lead to inaccurate resource assessment and sub-optimal design of turbines. A consequent lack of confidence from investors could significantly hamper the development of the tidal stream industry in the UK. This proposal will facilitate a major improvement in the way tidal currents are analysed and predicted, improving our fundamental understanding of these phenomena and delivering improvedengineering and decision making in the tidal stream industry.
The methods developed will be made freely available to users . Potential users come from across the oceanographic and tidal stream energy community. The methods developed will be valuable beyond the tidal energy community; uses include navigation information for shipping, analysis of sediment transport, pollutant modelling and environmental loading on structures. This project's contributions will hence have broad applicability and impact.
The methods developed will be made freely available to users . Potential users come from across the oceanographic and tidal stream energy community. The methods developed will be valuable beyond the tidal energy community; uses include navigation information for shipping, analysis of sediment transport, pollutant modelling and environmental loading on structures. This project's contributions will hence have broad applicability and impact.
Publications
Dias F
(2017)
Analytical and computational modelling for wave energy systems: the example of oscillating wave surge converters.
in Acta mechanica Sinica = Li xue xue bao
Sarkar D
(2018)
Prediction of tidal currents using Bayesian machine learning
in Ocean Engineering
Sarkar D
(2019)
Spatiotemporal Prediction of Tidal Currents Using Gaussian Processes
in Journal of Geophysical Research: Oceans
Sarkar D.
(2016)
A machine learning approach to the prediction of tidal currents
in Proceedings of the International Offshore and Polar Engineering Conference
Description | This project used techniques from the machine learning community to predict tidal currents. Although we had to use a slightly different methodology to that originally expected the project was successful on improving the previous state of the art techniques. More significantly, we also developed a further technique for predicting tidal currents across a large spatial area which is not sampled continuously - essentially we could get as good predictions when using one moving sampling point within a tidal race as, using traditional techniques, you sampled at every spot. Two journal papers are under consideration. |
Exploitation Route | We hope the techniques developed will be of use to industry and academia. They will be particularly useful for those developing tidal energy sites. |
Sectors | Energy |
Description | This project was an early example of the application of machine learning to ocean data. It was perhaps ahead of its time in some respects which meant output has not been directly picked up as had been hoped even though the work was published in world-leading journals. There is increased interest in this topic, including recently in discussion with a tidal turbine developer who is supporting our recently submitted programme grant. |
First Year Of Impact | 2022 |
Sector | Energy |