Machine learning approaches in predicting the wind power output and turbine rotational speed in a wind farm

İLHAN A., TÜMSE S., BİLGİLİ M., Sahin B.

Energy Sources, Part A: Recovery, Utilization and Environmental Effects, cilt.46, sa.1, ss.12084-12110, 2024 (SCI-Expanded) identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 46 Sayı: 1
  • Basım Tarihi: 2024
  • Doi Numarası: 10.1080/15567036.2024.2392890
  • Dergi Adı: Energy Sources, Part A: Recovery, Utilization and Environmental Effects
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, ABI/INFORM, Aerospace Database, Agricultural & Environmental Science Database, Applied Science & Technology Source, CAB Abstracts, Communication Abstracts, Compendex, Computer & Applied Sciences, Environment Index, Greenfile, INSPEC, Metadex, Pollution Abstracts, Veterinary Science Database, Civil Engineering Abstracts
  • Sayfa Sayıları: ss.12084-12110
  • Anahtar Kelimeler: Artificial intelligence, climate change, machine learning, turbine rotational speed, wind power
  • Çukurova Üniversitesi Adresli: Evet

Özet

Accurate wind energy forecasting has become increasingly important to effectively manage the energy produced by wind turbine power plants and optimize their operational performance. In this study, several artificial intelligence techniques are recommended to simulate turbine rotation speed to predict wind energy production 10 minutes ahead. Four tools are employed: The fuzzy c-means (FCM) approach of adaptive neuro-fuzzy inference system (ANFIS), long short-term memory (LSTM), grid partitioning (GP) method of adaptive neuro-fuzzy inference system and subtractive clustering (SC) algorithm of adaptive neuro-fuzzy inference system were used for predictions. These methods use historical data as input for the physical parameters to be estimated and estimate the subsequent value as the output. Thus, using only historical data, future values of the considered parameter can be easily predicted without the need for other physical parameters, such as meteorological data or data regarding the design of the mechanical installation, or without solving complex differential equations containing many unknowns. In the study, wind power and rotor rotational speed data from one wind turbine operating in a wind farm is taken into consideration. Among 34 models, LSTM is shown to perform best in capturing real observed wind turbine parameters. In the estimations of wind output power of wind turbine, it is reported that the rates of evaluation criteria were computed as 136.04 kW MAE, 242.64 kW RMSE, and 0.9711 R. Besides, in the predictions of turbine rotational speed, it is notified that 0.72 rpm MAE, 1.16 rpm RMSE, and 0.9452 R values were computed. On the other hand, among the generated ANFIS models, ANFIS-FCM model yields best accurate results with 138.69 kW MAE, 244.40 kW RMSE and 0.9708 R values in wind power, 0.73 rpm MAE, 1.17 rpm RMSE and 0.9451 R values in turbine rotor rotation.