Reinforcement learning based control for arrays of point absorber wave energy converters

The aim of this project is to use single-agent and multi-agent reinforcement learning to obtain optimal control strategies for individual point absorber wave energy converters and arrays of these. This will guarantee maximum power on the one hand, and minimum power oscillation on the other. The approach will also involve multi-objective optimisation.
First, we will analyse the response of individual point absorbers and arrays to different sea states and wave conditions. It is envisaged that the work would be applicable to point absorbers with different PTO technologies, but as a reference, a hydraulic PTO will be assumed. Second, different reinforcement learning architectures will be explored, for example cooperative multi-agent control.
Reinforcement learning will be used to address two control objectives: (a) optimising power generation of individual point absorbers and arrays and reducing overall power oscillation; (b) robust operation, via reinforcement learning algorithms, to track modelling uncertainties and disturbances.