A Photovoltaic Power Output Prediction Method Based on an Extended Similar-day Time Scale

With the vigorous development of the new energy industry, solar energy is being used more extensively and deeply. However, photovoltaic power generation exhibits characteristics such as randomness, volatility, anti-peak and intermittency, making its power output uncertain and highly influenced by weather factors. This uncertainty poses challenges for power system scheduling and secure operation. Therefore, establishing accurate photovoltaic power generation models based on climate factors is of paramount importance. This paper introduces a novel photovoltaic power output prediction method based on an extended similar-daytime scale. The method takes into account the photovoltaic power output throughout the year, effectively selecting similar days or even similar hours that can accurately capture the characteristics of the predicted day’s photovoltaic power output through a comparison of meteorological conditions. Additionally, it combines the use of neural network algorithms and relevant statistical knowledge, effectively addressing the unbounded and irregular nature of meteorological data for complex weather types in terms of time scale. The model proposed in this paper balances prediction speed and accuracy and holds significant practical value.