Open AccessAir pressure forecasting for the Mutriku oscillating‐water‐column wave power plant: Review and case study
Author(s) -
Marques Silva Jorge,
Vieira Susana M.,
Valério Duarte,
Henriques João C. C.,
Sclavounos Paul D.
Publication year - 2021
Publication title -
iet renewable power generation
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.005
H-Index - 76
eISSN - 1752-1424
pISSN - 1752-1416
DOI - 10.1049/rpg2.12289
Subject(s) - oscillating water column , support vector machine , computer science , wave power , energy (signal processing) , renewable energy , power (physics) , model predictive control , column (typography) , power station , control theory (sociology) , engineering , machine learning , artificial intelligence , control (management) , mathematics , wave energy converter , statistics , telecommunications , physics , electrical engineering , quantum mechanics , frame (networking)
Abstract The high variability and unpredictability of renewable energy resources require optimization of the energy extraction, by operating at the best efficiency point, which can be achieved through optimal control strategies. In particular, wave forecasting models can be valuable for control strategies in wave energy converter devices. This work intends to exploit the short‐term wave forecasting potential on an oscillating water column equipped with the innovative biradial turbine. A Least Squares Support Vector Machine (LS‐SVM) algorithm was developed to predict the air chamber pressure and compare it to the real signal. Regressive linear algorithms were executed for reference. The experimental data was obtained at the Mutriku wave power plant in the Basque Country, Spain. Results have shown LS‐SVM prediction errors varying from 9% to 25%, for horizons ranging from 1 to 3 s in the future. There is no need for extensive training data sets for which computational effort is higher. However, best results were obtained for models with a relatively small number of LS‐SVM features. Regressive models have shown slightly better performance (8–22%) at a significantly lower computational cost. Ultimately, these research findings may play an essential role in model predictive control strategies for the wave power plant.