Wind Power Forecasting With LSTM and Comparison With Different Machine Learning Algorithms: A Case Study of Southwestern Turkey

Abstract

Wind is a renewable energy resource but its intermittent nature poses some issues for transmission system operators, wind farms, and the electricity market. To address these issues, wind power forecasting is crucial. In this study, the data were obtained from a turbine anemometer in the 1-year time horizon of 10 min for wind power forecasting. Models are designed with univariate long short-term memory, multivariate long short-term memory, multilayer perceptron (MLP), extreme gradient boosting (XGBOOST), convolutional neural networks (CNNs), long short-term memory, gated recurrent unit (GRU), and random forest (RF) algorithms. Outliers in the data are cleaned with the isolation forest algorithm. Hyperparameter optimization of the models, designed with machine learning algorithms for a selected month from each season (January, April, July, and November) was made with Bayesian optimization. The success of the models was measured by mean absolute error, root mean square error, and determination coefficient. Among the designed models, for January, April, and November, the multivariate long short-term memory model gave the most successful results with 0.926, 0.961, and 0.946 determination coefficient values, respectively. For July, the XGBOOST model provided the most successful result with 0.916 value of the determination coefficient

Keywords:

• wind energy
• wind power forecasting
• LSTM
• outlier detection
• hyperparameter tuning

AUTHOR CONTRIBUTION

M.A.Y. is in charge of the wind turbine data analysis, with O.B. providing direction. The idea was conceived and carried out by M.A.Y., who also carried out the tests. Both authors contributed equally to the final draft of the work. After reviewing and giving their approval, each contributor published the manuscript.

AVAILABILITY OF DATA AND MATERIALS

Not applicable.

DISCLOSURE STATEMENT

The authors declare no competing interests.

Additional information

Funding

Not applicable.

Notes on contributors

Mehmet Ali Yelgeç

Mehmet Ali YELGEÇ is presently a PhD candidate at Isparta University of Applied Sciences’ Department of Electrical and Electronics Engineering. 2015, in Eskisehir, Turkey, he graduated with a Bachelor of Science in electrical and electronics engineering from Eskisehir Osmangazi University. He graduated in 2022 from Isparta University of Applied Sciences in Isparta, Turkey, with a Master of Science in electrical and electronics engineering. He works at the Turkey Electricity Transmission Co. as a shift engineer at the moment, and his research interests include artificial intelligence and renewable energy.

Okan Bingöl

Okan BİNGÖL obtained his degrees in Bachelor’s, Master’s, and Doctoral programs in Electrical Education from Gazi University in Ankara, Turkey. At Isparta University of Applied Sciences, he is now Head of the Department of Electrical and Electronics Engineering and a Professor in the Faculty of Technology. His research interests include electric power components and systems.