Long short-term memory (LSTM) neural network and adaptive neuro-fuzzy inference system (ANFIS) approach in modeling renewable electricity generation forecasting


BİLGİLİ M., Yildirim A., ÖZBEK A., ÇELEBİ K., Ekinci F.

INTERNATIONAL JOURNAL OF GREEN ENERGY, cilt.18, sa.6, ss.578-594, 2021 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 18 Sayı: 6
  • Basım Tarihi: 2021
  • Doi Numarası: 10.1080/15435075.2020.1865375
  • Dergi Adı: INTERNATIONAL JOURNAL OF GREEN ENERGY
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, PASCAL, Aerospace Database, CAB Abstracts, Communication Abstracts, Compendex, Environment Index, Geobase, Greenfile, INSPEC, Metadex, Pollution Abstracts, Veterinary Science Database, Civil Engineering Abstracts
  • Sayfa Sayıları: ss.578-594
  • Anahtar Kelimeler: Deep learning, renewable electricity generation, adaptive neuro-fuzzy inference system (ANFIS), long short-term memory (LSTM), short-term forecasting
  • Çukurova Üniversitesi Adresli: Evet

Özet

Renewable energy sources are developing rapidly worldwide because they are unlimited and permanent, available in every country and also eliminate foreign dependency. In this respect, accurate renewable electricity generation (REG) forecasting is essential in a country's energy planning in relation to its development. In this study, two different data-driven methods such as adaptive neuro-fuzzy inference system (ANFIS) with fuzzy c-means (FCM) and long short-term memory (LSTM) neural network were applied to perform one-day ahead short-term REG forecasting. In addition, short-term hydropower electricity generation (HEG), geothermal electricity generation (GEG), and bioenergy electricity generation (BEG) forecasting were also made using these methods. The correlation coefficient (R), root-mean-square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) were used as evaluation criteria. The values predicted by the ANFIS-FCM and LSTM models were compared with the actual values by evaluating their errors. According to the test results obtained in terms of MAPE evaluation criteria, the best estimation model was obtained for GEG. The lowest MAPE values were found to be 7.20%, 7.46%, 1.63%, and 2.46% for REG, HEG, GEG, and BEG estimates, respectively. The results showed that both ANFIS and LSTM models presented satisfying performances in daily REG prediction, and the ANFIS and LSTM models gave almost identical results.