This work presents a comparative study of Deep Learning models for short-term forecasting of key electrical and thermal variables in PV inverters. Using data from a gridconnected PV plant, models are trained with different types of neural networks. Results show that hybrid CNN+LSTM and TCN architectures achieve the highest predictive accuracy across multiple metrics (MAE, RMSE, R2). The proposed approach supports an assessment of the condition of the inverter, enabling anomaly detection and predictive maintenance of photovoltaic systems for accurate production forecasting to avoid potential economic damage.
Comparative Analysis of Neural Network Models for PV Inverter Variable Prediction / Notarpietro, L.; Leuzzi, R.; Monopoli, V. G.; Bianco, G. L.; Canino, A.; Bizzarri, F.. - (2025), pp. 925-929. ( 14th International Conference on Renewable Energy Research and Applications, ICRERA 2025 aut 2025) [10.1109/ICRERA66237.2025.11283727].
Comparative Analysis of Neural Network Models for PV Inverter Variable Prediction
Notarpietro L.;Leuzzi R.;Monopoli V. G.;
2025
Abstract
This work presents a comparative study of Deep Learning models for short-term forecasting of key electrical and thermal variables in PV inverters. Using data from a gridconnected PV plant, models are trained with different types of neural networks. Results show that hybrid CNN+LSTM and TCN architectures achieve the highest predictive accuracy across multiple metrics (MAE, RMSE, R2). The proposed approach supports an assessment of the condition of the inverter, enabling anomaly detection and predictive maintenance of photovoltaic systems for accurate production forecasting to avoid potential economic damage.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
