Supervisor: Prof David Ingram (University of Edinburgh), Dr Qing Xiao (University of Strathclyde)
Sponsoring Company: Wave Energy Scotland
PhD Student: Enrico Anderlini
The Student has graduated and the project is complete
The design of an effective control strategy can significantly increase the energy absorption characteristics of wave energy converters, thus improving their economic feasibility. Most state-of-the-art control schemes rely on a model of the device dynamics, thus being affected by modelling errors. Not only does this result in lower energy absorption, but it can also result in damage to the machine in extreme waves, where modelling errors are greatest. For this reason, artificial intelligence algorithms have been investigated as a possible means to make existing control strategies for wave energy converters model free.
In particular, as part of this project, reinforcement learning and neural networks are studied in detail. Applications include time-averaged approaches, such as resistive and reactive control, as well as real-time control strategies, such as latching and declutching control. In the absence of experimental testing, these studies are performed using linear and non-linear models of different types of devices, including point absorbers and oscillating water columns. Both single- and multi-body units are analysed.
By removing the dependence of existing control strategies on models of the system dynamics, it is hoped that these schemes will result in an increase in energy absorption of current wave energy technologies without incurring any hardware costs. This is expected to improve the levellised cost of energy and take wave energy converters closer to economic viability.