Short-Term Electricity Demand Forecasting Using AI-Based Models for Green and Sustainable Energy Management
The International Symposium for Production Research 2025 (ISPR2025), İstanbul, Türkiye, 9 - 11 Ekim 2025, ss.0-1, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1007/978-3-032-22784-3
- Basıldığı Şehir: İstanbul
- Basıldığı Ülke: Türkiye
- Sayfa Sayıları: ss.0-1
- İstanbul Üniversitesi-Cerrahpaşa Adresli: Evet
Özet
This study focuses on enhancing short-term electricity demand forecasting to improve grid stability, support cost-effective operations, and facilitate re-newable energy integration in Türkiye’s power sector. To address the chal-lenges posed by nonlinear dynamics and seasonal variability in electricity consumption, four advanced AI-based forecasting models; Random Forest (RF), XGBoost, ExtraTrees, and Prophet were evaluated for 24-hour-ahead hourly demand prediction using multi-year datasets enriched with lagged demand and calendar-based features. The empirical results demonstrate that the ExtraTrees model achieved the highest forecasting accuracy, with R² = 0.889, MAE = 1,069 MW, RMSE = 1,489 MW, and MAPE = 3.04%. The RF and XGBoost models exhibited comparably strong performance, while the Prophet model despite its ability to capture seasonal and trend components, showed relatively lower accuracy (R² = 0.797, MAPE = 4.64%). These find-ings underscore the effectiveness of tree-based ensemble models in capturing short-term fluctuations and complex seasonal patterns of electricity demand. The study highlights the potential of AI-driven forecasting to improve re-newable energy scheduling, enhance operational efficiency, and strengthen the resilience of Türkiye’s electricity grid. Aligning these advanced forecast-ing approaches with sustainability goals contributes to accelerating Türkiye's transition to a low-carbon and environmentally friendly electricity sector.