Data-driven applications for wind energy analysis and prediction: The case of “La Haute Borne” wind farm
Abstract
The current paper investigates several methods of predicting wind energy generation for the onshore “La Haute Borne” wind farm. The hybrid model has been developed to get short-term power forecasts using both historical in-situ measurements available from ENGIE and Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2) global reanalysis dataset.
Meteorology is connected with chemistry because weather is the state of air, determined by pressure and temperature variations. This means that enriching historical measurements with weather data has a potential to get better predictions. It was shown that adding three extra meteorological parameters – pressure, humidity, and temperature – allowed to reach a higher accuracy compared with cases when weather parameters were completely ignored or used partially; this was proved by applying multivariate, one-step Long Short-Term Memory (LSTM) networks.
Finally, the paper explains how to apply LSTM networks for the day-ahead forecasts. In particular, two state-of-the-art models were investigated – a base LSTM network and a more advanced method which combines convolutional and LSTM layers through the CNN-LSTM approach. Results showed the latter network reached higher accuracy for both 12- and 24-h forecasts and performed faster than an ordinary LSTM network. A significant advantage of both methods deals with their light structure which allows running models on the Central Processing Unit (CPU).
Keywords
Onshore wind farm
Meteorology
Short-term forecasts
Deep learning
LSTM
CNN-LSTM
Description
Indexed in scopushttps://www.scopus.com/authid/detail.uri?authorId=57226096001 |
Article metrics10.31763/DSJ.v5i1.1674 Abstract views : | PDF views : |
Cite |
Full Text![]() |
Conflict of interest
“Authors state no conflict of interest”
Funding Information
This research received no external funding or grants
Peer review:
Peer review under responsibility of Defence Science Journal
Ethics approval:
Not applicable.
Consent for publication:
Not applicable.
Acknowledgements:
None.