Explainable Machine Learning for Distribution Transformer Load Ratio Forecasting
8th Global Power, Energy and Communication Conference, GPECOM 2026, Naples, İtalya, 3 - 05 Haziran 2026, ss.1022-1027, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1109/gpecom70462.2026.11578502
- Basıldığı Şehir: Naples
- Basıldığı Ülke: İtalya
- Sayfa Sayıları: ss.1022-1027
- Anahtar Kelimeler: Distribution transformers, Load ratio, Machine learning, SCADA, SHAP, Short-term forecasting, XAI
- Çukurova Üniversitesi Adresli: Evet
Özet
Today, electricity is an essential part of everyday life, and the quality of the electrical grid significantly impacts human well-being. Transformer load ratio estimation was carried out using machine learning methods due to the significant impact of transformers on the electrical grid. In this paper, Supervisory Control and Data Acquisition (SCADA) and weather data were collected on a single transformer; eXtreme Gradient Boosting (XGBoost), Long Short-Term Memory (LSTM), OneDimensional Convolutional Neural Network (1D CNN), and Seasonal Persistence (SP) structures were evaluated. Explainable Artificial Intelligence (XAI) was applied to the predictive model that demonstrated the highest accuracy, achieving a maximum R2 value of 0.97, along with a minimum RMSE value of 1.97 and a minimum MAE value of 1.41. The variables that most affect the load ratio were determined. The findings provide recommendations for power downsizing or upsizing and transformer tap adjustments based on the predicted short-term load ratio derived from data-driven models.