Comparison of CNN and CNN-LSTM Based Deep Learning Models in Photovoltaic Power
Öz
This study presents a systematic and controlled comparison of Convolutional Neural Networks(CNN) and hybrid CNN–LSTM architectures for photovoltaic(PV) power forecasting. In recent years, hybrid deep learning models have often been adopted under the assumption that increased architectural complexity leads to improved forecasting performance. However, this assumption has not been rigorously validated for PV power generation, which is primarily governed by localized and periodic temporal patterns. The analysis is conducted using one year of monthly aggregated meteorological and electrical production data collected in 2024 from three real-world PV systems located in different climatic regions of Türkiye. The systems include two grid-connected installations with capacities of 72.8 kWp and 49.6 kWp, as well as a 3.3 kWp battery-integrated hybrid system. To ensure a fair comparison, both models were trained and evaluated under identical conditions, including the same datasets, preprocessing pipeline, training configuration, and evaluation metrics. Forecasting performance was assessed using Mean Absolute Percentage Error (MAPE) and the coefficient of determination (R²). The results demonstrate that the standalone CNN model consistently outperforms the CNN–LSTM architecture across all systems, achieving MAPE values between 2.06% and 2.92%, while the CNN–LSTM model yields higher errors ranging from 4.19% to 4.87%. These findings suggest that PV power generation at monthly resolution is largely characterized by repetitive and localized temporal patterns that can be effectively captured through convolutional feature extraction. The inclusion of LSTM layers introduces additional model complexity without providing measurable performance improvements.
Anahtar Kelimeler
- Photovoltaic Power Estimation
- Photovoltaic Power Plants
- Deep learning
- Convolutional neural networks
- CNN–LSTM
- Data Augmentation
- Monthly Forecasting
Destekleyen Kurum
Etik Beyan
Teşekkür
Kaynakça
- Ahmed, R., Sreeram, V., Mishra, Y., & Arif, M. D. (2020). A review and evaluation of the state-of-the-art in PV solar power forecasting: Techniques and optimization. Renewable and Sustainable Energy Reviews, 124, 109792.
- T.C. Enerji ve Tabii Kaynaklar Bakanlığı. (2023). Türkiye Ulusal Enerji Planı. https://enerjiverimliligi.enerji.gov.tr/template/dist/pdf/T%C3%BCrkiye_Ulusal_Enerji_Plan%C4%B1.pdf, (Erişim tarihi: 01.10.2025).
- Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). Imagenet classification with deep convolutional neural networks. Advances in Neural Information Processing Systems, 25.
- Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural Computation, 9(8), 1735–1780.
- Agga, A., Abbou, A., Labbadi, M., El Houm, Y., & Ali, I. H. O. (2022). CNN-LSTM: An efficient hybrid deep learning architecture for predicting short-term photovoltaic power production. Electric Power Systems Research, 208, 107908.
- Korkmaz, D. (2021). SolarNet: A hybrid reliable model based on convolutional neural network and variational mode decomposition for hourly photovoltaic power forecasting. Applied Energy, 300, 117410.
- Mellit, A., Pavan, A. M., Ogliari, E., Leva, S., & Lughi, V. (2021). Deep learning neural networks for short-term photovoltaic power forecasting. Renewable and Sustainable Energy Reviews, 149, 111315.
- Zang, H., Cheng, L., Ding, T., Cheung, K. W., Wei, Z., & Sun, G. (2020). Short-term global horizontal irradiance forecasting based on a hybrid CNN-LSTM model with spatiotemporal correlations. Renewable Energy, 160, 26–37.
Ayrıntılar
Birincil Dil
İngilizce
Konular
Yenilenebilir Enerji Sistemleri
Bölüm
Araştırma Makalesi
Yayımlanma Tarihi
30 Ekim 2026
Gönderilme Tarihi
15 Ocak 2026
Kabul Tarihi
31 Mart 2026
Yayımlandığı Sayı
Yıl 2026 Cilt: 13 Sayı: Sürdürülebilirlikte Güncel Yaklaşımlar Özel Sayısı