Transfer Learning-Based Prediction of Heating Energy Consumption Using Urban Building Energy Models (UBEM)

Canli I., UÇAR Ş., Gursoy F. E., Halacli E. G., Iseri O. K., YAVUZ F., ...Daha Fazla

31st International Workshop on Intelligent Computing in Engineering, EG-ICE 2024, Vigo, İspanya, 3 - 05 Temmuz 2024, ss.580-589 identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Basıldığı Şehir: Vigo
  • Basıldığı Ülke: İspanya
  • Sayfa Sayıları: ss.580-589
  • Orta Doğu Teknik Üniversitesi Adresli: Evet

Özet

To estimate the energy demand of buildings, machine learning (ML) using datasets generated by Urban Building Energy Modeling (UBEM) has emerged as a potent approach. However, collecting large training datasets for each urban context is challenging, and developing large UBEMs for each city is too resource-intensive. Transfer learning (TL) can address these limitations by transferring knowledge from an existing model to train a model in target cities with limited datasets. In this paper, we present an approach for TL-based prediction of heating energy consumption on an urban scale based on three existing UBEMs of cities with different climatic conditions and thermal characteristics of buildings. Our experiments show that TL does provide better estimation compared to training ML models from scratch. Furthermore, our results suggest that this gain in accuracy is even higher if the target city has a substantially low number of buildings to train an ML model.