Short-term prediction of photovoltaic energy generation by intelligent approach

► ANN correlates metrological data and energy generated by PV. ► ANN hidden neurons number estimated by rule of thumb is justified statistically. ► Three parameters predict real-time and up-to-20-min lapse PV energy generated. ► The result offers a 95% confidence level of prediction crucial to power management. Population growth and quickly depleting fossil fuel reserves are creating demand for the development and use of renewable energy resources such as solar energy. The evaluation and forecasting of energy demands have become concerns for facility managers, and predicting energy generation plays a critical role in power-system management, scheduling, and dispatch operations. A reliable energy supply forecast helps to prevent unexpected loads and provides vital information for decisions made on energy generation and purchase. However, study of energy generation prediction by the photovoltaic (PV) system has been limited over the years, especially concerning short-term predictions. This study will adopt the artificial neural network (ANN) to mimic the nonlinear correlation between the metrological parameters and energy generated by the PV system. It aims to find that short-term prediction performance is comparable with real-time prediction performance when ahead solar angles are applied to the predictions.