LSTM-Transformer-Seq2Seq Hibrit Yaklaşımıyla Geliştirilmiş Kısa Vadeli PV Güç Tahmini

Yıl 2026, Cilt: 8 Sayı: 1 , 1 - 11 , 30.04.2026

Bahtiyar Taşdemir

https://doi.org/10.46387/bjesr.1786127

https://izlik.org/JA86TB89SY

Öz

Fotovoltaik (PV) enerji üretimi, hava koşullarına bağlı olarak değişkenlik gösterdiğinden doğrusal olmayan ve karmaşık bir yapıya sahiptir. Bu durum güvenilir tahmin yöntemlerine olan ihtiyacı artırmaktadır. Bu çalışmada, fotovoltaik üretiminin daha doğru tahmin edilmesini sağlamak için Uzun Kısa Dönem Bellek (LSTM), Transformatör ve Seq2Seq modellerini birleştiren hibrit bir derin öğrenme modeli önerilmiştir. Önerilen modelde, geçmiş fotovoltaik üretim verileri uzun kısa dönem bellek tabanlı bir kodlayıcı kullanılarak işlenirken, meteorolojik veriler bir Transformer kodlayıcı kullanılarak analiz edilmiştir. Elde edilen özellik vektörleri birleştirildi ve gelecek tahminleri oluşturmak için Seq2Seq tabanlı bir kod çözücü yapısına beslendi. Model performansı Ortalama Mutlak Hata (MAE), Ortalama Kare Hatanın Kökü (RMSE) ve Ortalama Mutlak Yüzde Hata (MAPE) ölçütleri kullanılarak değerlendirilmiş ve sırasıyla 18.9 kW, 217.7 kW ve %9.36 değerleri elde edilmiştir. Sonuçlar, geliştirilen hibrit modelin fotovoltaik enerji tahmininde yüksek doğruluk sağladığını ve mevcut yöntemlere kıyasla üstün performans sergilediğini göstermektedir.

Anahtar Kelimeler

Tahmin , Hibrit , Fotovoltaik güç , Yenilenebilir enerji

Short term PV Power Forecasting Enhanced with an LSTM-Transformer-Seq2Seq Hybrid Approach

https://izlik.org/JA86TB89SY

Öz

Photovoltaic (PV) energy production has a non-linear and complex structure, as it varies depending on weather conditions. This situation increases the need for reliable prediction methods. In this study, a composite deep learning approach that integrates Long Short-Term Memory (LSTM), Transformer, and Seq2Seq models has been proposed to enable more accurate forecasting of photovoltaic production. In the proposed model, historical photovoltaic production data was processed using a long short-term memory-based encoder, while meteorological data was analyzed using a Transformer encoder. The resulting feature vectors were combined and fed into a Seq2Seq based decoder structure to generate future predictions. Model performance was evaluated using the Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE) metrics, yielding values of 18.9 kW, 217.7 kW, and 9.36%, respectively. It was observed that the hybrid model offers reliable accuracy in photovoltaic energy forecasting and performs better than previously used models.

Anahtar Kelimeler

Forecast , Hybrid , Photovoltaic power , Renewable energy

Kaynakça

• W.-C. Tsai, C.-S. Tu, C.-M. Hong, and W.-M. Lin, “A Review of State-of-the-Art and Short-Term Forecasting Models for Solar PV Power Generation,” Energies (Basel), vol. 16, no. 14, p. 5436, Jul. 2023.
• I. K. Bazionis, M. A. Kousounadis‐Knousen, P. S. Georgilakis, E. Shirazi, D. Soudris, and F. Catthoor, “A taxonomy of short‐term solar power forecasting: Classifications focused on climatic conditions and input data,” IET Renewable Power Generation, vol. 17, no. 9, pp. 2411–2432, Jul. 2023.
• K. J. Iheanetu, “Solar Photovoltaic Power Forecasting: A Review,” Sustainability, vol. 14, no. 24, p. 17005, Dec. 2022.
• A. H. Eşlik, O. Sen, and F. Serttaş, “Güneş ışınımı tahmini için CNN-LSTM modeli: Performans analizi,” Gazi Üniversitesi Mühendislik Mimarlık Fakültesi Dergisi, vol. 39, no. 4, pp. 2155–2162, May 2024.
• O. Taşdemir, “Photovoltaic Power Prediction with Teaching Learning Based Optimization Algorithm,” Gazi University Journal of Science Part A: Engineering and Innovation, vol. 11, no. 4, pp. 780–791, Dec. 2024.
• Y. Dai, Y. Wang, M. Leng, X. Yang, and Q. Zhou, “LOWESS smoothing and Random Forest based GRU model: A short-term photovoltaic power generation forecasting method,” Energy, vol. 256, p. 124661, Oct. 2022.
• M. Yang, M. Zhao, D. Huang, and X. Su, “A composite framework for photovoltaic day-ahead power prediction based on dual clustering of dynamic time warping distance and deep autoencoder,” Renew Energy, vol. 194, pp. 659–673, Jul. 2022.
• D. El Bourakadi, H. Ramadan, A. Yahyaouy, and J. Boumhidi, “A novel solar power prediction model based on stacked BiLSTM deep learning and improved extreme learning machine,” International Journal of Information Technology, vol. 15, no. 2, pp. 587–594, Feb. 2023.
• F. Yiğit, İ. Karali, and A. Kabul, “Experimental and ensemble learning-based prediction of heat pipe and aluminum fin cooling effects on PV panel power output for a sustainable future,” Thermal Science and Engineering Progress, vol. 66, p. 104067, Oct. 2025.
• A. Sardarabadi, A. Heydarian Ardakani, S. Matrone, E. Ogliari, and E. Shirazi, “Multi-temporal PV power prediction using long short-term memory and wavelet packet decomposition,” Energy and AI, vol. 21, p. 100540, Sep. 2025.
• H. Dai, Z. Zhen, F. Wang, Y. Lin, F. Xu, and N. Duić, “A short-term PV power forecasting method based on weather type credibility prediction and multi-model dynamic combination,” Energy Convers Manag, vol. 326, p. 119501, Feb. 2025.
• S. Peng et al., “Prediction of wind and PV power by fusing the multi-stage feature extraction and a PSO-BiLSTM model,” Energy, vol. 298, p. 131345, Jul. 2024.
• G. Piantadosi, S. Dutto, A. Galli, S. De Vito, C. Sansone, and G. Di Francia, “Photovoltaic power forecasting: A Transformer based framework,” Energy and AI, vol. 18, p. 100444, Dec. 2024.
• S. Cui, S. Lyu, Y. Ma, and K. Wang, “Improved informer PV power short-term prediction model based on weather typing and AHA-VMD-MPE,” Energy, vol. 307, p. 132766, Oct. 2024.
• O. Das, D. Dahlioui, M. Hamza Zafar, N. Akhtar, S. Kumayl Raza Moosavi, and F. Sanfilippo, “Ultra-short term PV power forecasting under diverse environmental conditions: A case study of Norway,” Energy Conversion and Management: X, vol. 27, p. 101072, Jul. 2025.
• J. N. Abed, A. A. Abdoos, and A. Ebadi, “A new hybrid intelligent method for short-term photovoltaic power forecasting based on combination of ELM, VMD and PCA,” Electric Power Systems Research, vol. 250, p. 112104, Jan. 2026.
• M. Yang, Z. Guo, D. Wang, B. Wang, Z. Wang, and T. Huang, “Short-term photovoltaic power forecasting method considering historical information reuse and numerical weather forecasting,” Renew Energy, vol. 256, p. 123933, Jan. 2026.
• Y. Su, M. Zhang, L. Cao, Y. Chen, and Y. Tian, “Spatio-temporal Graph Neural Network with Fourier features for multi-site photovoltaic power forecasting,” Electric Power Systems Research, vol. 251, p. 112171, Feb. 2026.
• H. Feng, Y. Shi, M. Ren, W. Zhang, J. Zhang, and Y. Zhao, “Distribution-Free photovoltaic power probability density forecasting based on hybrid deep learning,” Solar Energy, vol. 300, p. 113769, Nov. 2025.
• G. Van Houdt, C. Mosquera, and G. Nápoles, “A review on the long short-term memory model,” Artif Intell Rev, vol. 53, no. 8, pp. 5929–5955, Dec. 2020.
• Vaswani, Ashish, Noam M. Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser and Illia Polosukhin. “Attention is All you Need.” Neural Information Processing Systems (2017).
• Y. Dai, X. Yang, and M. Leng, “Optimized Seq2Seq model based on multiple methods for short-term power load forecasting,” Appl Soft Comput, vol. 142, p. 110335, Jul. 2023.