Abstract

Accurate forecasting of photovoltaic (PV) power generation is crucial for optimizing energy management and enhancing grid stability. While hybrid models combining Numerical Weather Prediction (NWP) with deep learning techniques and electrical models have shown promising results, improving their forecasting accuracy remains a key challenge. This study investigates the use of hybrid forecasting models for PV power prediction, focusing on the integration of WRF-Solar outputs with machine learning approaches such as Long Short-Term Memory (LSTM) and electrical models. The study evaluates model performance over a 24-hour forecast horizon using standard metrics, including mean absolute error and root mean square error. The results highlight the balance between prediction accuracy and model efficiency, providing insights into the effectiveness of combining numerical weather predictions with machine learning approaches for improving PV power prediction systems.