Comparison the Performance of Different Optimization Methods in Artificial Intelligence Based Electricity Production Forecasting

Kaysal K., Hocaoglu F. O., Öztürk N.

10th International Conference on Smart Grid, icSmartGrid 2022, İstanbul, Türkiye, 27 - 29 Haziran 2022, ss.236-239 identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/icsmartgrid55722.2022.9848724
  • Basıldığı Şehir: İstanbul
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.236-239
  • Anahtar Kelimeler: energy forecasting, optimization, short term load forecasting
  • Gazi Üniversitesi Adresli: Evet

Özet

© 2022 IEEE.The purpose of this paper is to examine the performance of the LSTM method in Turkey's electricity production estimation and to determine the optimization technique that provides the best performance in the LSTM estimation method. In this study as a short term one hour's production forecast was made. For the forecasting, the last four years' real-time production data were analyzed hourly. Also, after the LSTM architecture was established in the study, the most suitable optimization method for the network was tried to be determined. The performance of each of them was calculated according to MAPE and R2 score criteria and the results were compared by using four different optimization techniques without changing the architectural structure created. According to the results obtained from Adam, Adamax, Nadam and RMSprop optimization techniques, proposed optimization method showed the best performance for this study with a 98% success rate.