Real-time control of WECs based on NAR, NARX and LSTM artificial neural network

Li, Tuo and Zhang, Ming and Yu, Shuang-Rui and Yuan, Zhi-Ming; Chung, Jin S. and Buzin, Igor and Kawai, Hiroyasu and Liu, Hua and Kubat, Ivana and Peng, Bor-Feng and Reza, Ali and Sriram, Venkatachalam and Van, Suak Ho and Wan, Decheng and Yamaguchi, Satoru and Yan, Shiqiang, eds. (2022) Real-time control of WECs based on NAR, NARX and LSTM artificial neural network. In: The Proceedings of The Thirty-second (2022) International Ocean and Polar Engineering Conference. The Proceedings of the International Ocean and Polar Engineering Conference . International Society of Offshore and Polar Engineers, CHN, pp. 359-367. ISBN 9781880653814 (https://onepetro.org/ISOPEIOPEC/proceedings/ISOPE2...)

[thumbnail of Li-etal-ISOPE-2022-Real-time-control-of-WECs-based-on-NAR-NARX-and-LSTM-artificial-neural-network]
Preview
 Text. Filename: Li-etal-ISOPE-2022-Real-time-control-of-WECs-based-on-NAR-NARX-and-LSTM-artificial-neural-network.pdf
Accepted Author Manuscript
License: Strathprints license 1.0
Download (1MB)| Preview

Abstract

In this study, we aim to improve WECs’ performance for maximizing energy absorption through a sub-optimal method of phase control by latching is applied to the device. The forecasting of future wave force is required for the optimal control command deducted. An artificial neural network, namely LSTM (Long Short-Term Memory) is proposed to accurately predict the short-term wave force. The hydrodynamic properties of a point absorber is analyzed based on the 3D potential flow theory in frequency-domain. Cummin’s equation and a 4th-order state-space model are used to efficiently represent the hydrodynamic behavior of the WEC under irregular waves in time-domain. The Nonlinear Autoregressive artificial neural network(NAR-ANN) and NARx network are used to verify the method proposed in this paper. The simulation results show that the mean square error value, root mean square error value and R2 value based on the LSTM prediction model are better than those of the NAR prediction model. The prediction performance of LSTM is more suitable for processing the time series.

CORE (COnnecting REpositories)