A Deep Learning Solution to optimise the control of Tidal Stream Conversion Devices with considerations of the marine environment

This PhD project aims to investigate a deep-learning solution for preserving the marine ecosystem when using tidal stream turbines.

Tidal stream turbines have the potential to generate a significant amount of energy, with the global tidal stream energy potential estimated to be around 120-400 GW.

Tidal turbines themselves are devices that convert the energy in marine currents into useful energy in a nearly identical manner to wind turbines.

However, tidal turbines may have a significant impact on the marine ecosystem, both through the risk of collision with marine species and the level of operational noise generated.

Not only do tidal turbines have an effect on the environment, but the wake dispersion caused by nearby wildlife can also affect the power output of the turbine. As such, control systems have to be in place in order to keep the turbine working at an optimal power output.

In this project, deep learning will be used to detect the presence of marine biodiversity by using information captured by the turbine's sensors, such as power, thrust, angular velocity, and flow speed, to avoid endangering the marine ecosystem prior to large-scale deployment.

Therefore, the aim of this PhD will be to classify and numerically quantify how flow disturbances from marine life affect tidal turbine arrays, and then devise deep-learning architectures to detect marine life's presence before a collision occurs and control its operation accordingly.