Photovoltaic Power Forecasting based on Artificial Neural Network and Ultraviolet Index

The accuracy of photovoltaic (PV) power generation forecast can seriously affect the penetration ability of PV power into the existing power grid, which is one of the key approaches to achieve emission peak, as well as realize carbon neutrality. In the conventional forecasting methods, Global Horizontal Irradiation (GHI), Diffuse Horizontal Irradiance (DHI), temperature, wind speed, rainfall, etc. are considered as the mainly factors to forecast the PV output power, but ignore the impact of PV power generation caused by the whole PV system’s decay over the 25– 30 years lifecycle. The ultraviolet (UV) index, which reflects the quantity of 10–400 nm irradiation, has a strong correlation with such decay and power generation. This paper proposes a novel PV power forecasting model that involving UV index in an artificial neural network, using Adam method to optimize the training process with the Kerastuner employed for optimization of the hyperparameters. Experiments demonstrate that the proposed model achieves more precise performance than conventional methods.