PV Power Forecasting Using Deep Learning and Physical Models: Case Study of Morocco

Of all renewable energy sources, photovoltaic technology is the most immediate way to convert solar radiation into electricity. Although the penetration of renewable energies has increased in recent years, the problem of intermittency still persists due to the nature of the solar resource. Accurate solar power forecasting is crucial for operations and maintenance (O &M) and day-to-day operations monitoring in solar plants. This chapter proposes a hybrid day-ahead forecasting approach that combines deep learning with Numerical Weather Prediction (NWP) and electrical models. The performance of this model is compared to two other models: a WRF-Solar + electric model and an LSTM + regression model. The models are evaluated using Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE), and Mean Absolute Error (MAE) metrics. The proposed LSTM + WRF-Solar + Two diode model demonstrates improved forecasting accuracy compared to the other two models, with an improvement of 8.79% compared to the LSTM + regression model and 3.3% compared to the WRF-Solar + Two diode model.