Neural Network Based Irradiance Mapping Model of Solar PV Power Forecasting Using Sky Image

作者：Fei Wang, Xinxin Ge, Zhao Zhen, Hui Ren, Yajing Gao, Da‐Shuai Ma, Miadreza Shafie‐khah, João P. S. Catalào · 年份：2018 · DOI：10.1109/ias.2018.8544694 · 被引用次数：22 · 研究领域：Solar Radiation and Photovoltaics、Photovoltaic System Optimization Techniques、Solar Thermal and Photovoltaic Systems

Due to the stochastic fluctuant characteristic of solar irradiance, large-scale grid-connected photovoltaic (PV) power plants can bring great difficulties to the operation of the power system. In order to fulfil the sky images based ultra-short term PV power forecasting and enhance the grid consumptive ability of PV power, an accurate model that can map sky images to corresponding surface solar irradiance is very significant. Therefore, in this paper a neural network based irradiance mapping model of solar PV power forecasting using sky image is proposed. First, we combine the theoretical calculation of extraterrestrial solar irradiance and atmospheric optical thickness to establish the clearance surface irradiance model. Second, the sky images observed by total sky imager are processed to extract image features related to solar irradiance. Third, a neural network based irradiance mapping model is built and trained using historical sky images and solar irradiance data. Simulation results show that the proposed model can map sky image features to surface solar irradiance accurately in different weather conditions.