Calculation Method Of Surface Irradiance Based On Sky Image And Atmospheric Transmission Model

With the growing energy crisis and environmental pollution,solar energy as a renewable energy,is getting more and more attention.Solar radiation is the primary factor affecting the power of photovoltaic plant.Because of its attenuation by a variety of atmospheric factors,photovoltaic power is intermittent and volatile.Large-scale grid-connected photovoltaic System caused a huge impact on the power grid,which poses a challenge to the safe operation of the power grid.In order to promote the absorption of photovoltaic power for utility grid and improve the accuracy of the ultra-short-term global horizontal irradiance prediction or photovoltaic power prediction,accurate calculation of global horizontal irradiance using total sky images has a very important significance.Accurate calculation of global horizontal irradiance requires a comprehensive consideration of factors that affect solar radiation.In this paper,we analyzed astronomical and atmospheric factors that resulted in periodic and stochastic variations in solar radiation respectively.We focus on the sky image data,estimate the attenuation of the cloud on the solar radiation,and optimize the irradiance neural network model.Firstly,the atmospheric factors such as cloud,water vapor,aerosol and atmospheric molecules was evaluated at the degree of attenuation of solar radiation,the time scale of the influence of solar radiation and the spatial scale of the distribution.The clear sky surface irradiance model was established using the optical thickness of the direct radiation and scattered radiation in the atmosphere,combined with the theoretical calculation of extraterrestrial radiation.Secondly,the sky image taken by the total sky imager need to remove the noise such as support arm images,shading images and images around the sky environment in the sky image,and the sky image distortion caused by the hemispherical mirror should also be reduced.Then the fixed threshold method of red and blue ratio was used to identify the thin cloud,opaque cloud and sky in the sky image.And according to the solar zenith angle and azimuth angle,we got the block map corresponding to the sky image.Finally,the relevant factors such as global horizontal irradiance of the clear sky,opaque cloud,thin cloud,the temperature and the wind speed are introduced as the input of the back propagation neural network model.The results of the combination of different input variables were compared to determine the optimal input variable combination,and the model structure are optimized using cross-validation algorithm,and the output was the actual global horizontal irradiance.In this paper,the model was trained by the data collected by National Renewable Energy Laboratory.The experimental results show that the method has high accuracy in calculating the global horizontal irradiance in cloudy environment.