Forecasting Electricity Generation and Assessing 2030 Energy Transition Scenarios: Evidence from Türkiye and Germany

Abstract

Accurate electricity generation forecasting plays an important role in production planning, resource allocation, supply security, and long-term investment decisions. Since electricity systems require a continuous balance between supply and demand, forecasting errors may negatively affect operational efficiency and increase imbalance costs. At the same time, with the acceleration of the energy transition, not only short-term generation forecasts but also the evaluation of how future generation structures may evolve under different scenarios has become increasingly important. In this study, deep learning and boosting-based models were compared for Türkiye and Germany under 7-, 30-, and 365-day forecasting horizons, and the best-performing models were then used to generate BAU projections for 2030, which were evaluated together with the ST and EC scenarios reported by ENTSO-E. The findings showed that the GRU model produced more successful and consistent results for Türkiye, whereas boosting algorithms delivered stronger forecasting performance for Germany. The scenario comparisons also revealed that the BAU structure remained more dependent on fossil-based resources, while the ST and EC scenarios placed a clearer emphasis on the transition toward renewable energy. In this way, the study provides an integrated framework that supports both the evaluation of forecasting performance and the assessment of long-term energy planning alternatives.