Optimization of the electricity generation of a wave energy converter using deep reinforcement learning
Document Type
Article
Publication Date
1-15-2022
Abstract
Ocean wave energy is one of the sustainable energy sources which continues to attract increasing research interests. Traditionally, the model-based controls of Wave Energy Converters (WECs) are derived from the linear hydrodynamics. Although showing promising performance, in practice, the actual electricity produced is significantly lower than predicted. On the other hand, it is extremely cumbersome to derive a model-based control from very complex dynamics. In this paper, a Deep Reinforcement Learning (DRL) control is proposed which is model-free and therefore makes it possible to be designed from a global point of view. To validate the performance of the control, a point absorber WEC with a direct-drive power take-off (PTO) unit is simulated. The results show the proposed DRL control outperforms the model-based controls in terms of wave power production (improved from 24% up to 152%). Furthermore, the results show the DRL control achieves the best power quality in terms of the operation efficiency and power variation (improved from 23% up to 84%). Finally, the performance of different controls is further validated with real ocean conditions (at PacWave) and the DRL can consistently shows the best performance in terms of both power production and power quality.
Publication Title
Ocean Engineering
Recommended Citation
Zou, S.,
Zhou, X.,
Khan, I.,
Weaver, W.,
&
Rahman, S.
(2022).
Optimization of the electricity generation of a wave energy converter using deep reinforcement learning.
Ocean Engineering,
244.
http://doi.org/10.1016/j.oceaneng.2021.110363
Retrieved from: https://digitalcommons.mtu.edu/michigantech-p/15574