云图纹理分析结合SVM的地表太阳辐射预测  

On Ground Solar Radiation Forecasting Combined with Textural Feature Extraction and SVM. Computer Engineering and Applications

作　　者：郁云[1] 许昌[2] 曹潇[3] 魏瑾[1] 徐坚[1] 

机构地区：[1]南京信息职业技术学院信息服务学院,南京210046 [2]河海大学能源与电气学院,南京211100 [3]中国电力科学研究院新能源研究中心,北京100192

出　　处：《西南师范大学学报（自然科学版）》2017年第12期76-82,共7页Journal of Southwest China Normal University(Natural Science Edition)

基　　金：南京信息职业技术学院基金项目(YK20140601);科技部中丹国际合作项目(2014DFG62530);江苏省自然科学基金面上项目(BK20131369);国家电网公司科学技术项目(NY71-16-034)

摘　　要：太阳能资源的间歇性与不稳定性为地表太阳辐射的准确预测带来很大的挑战.本研究通过对全天空图像进行纹理特征分析,并结合支持向量机模型,实现了对地表太阳辐射的预测.首先,基于全天空图像,利用数字图像处理技术提取与太阳辐射相关的图像纹理特征,包括反差、熵、灰度相关与能量;然后,结合图像特征与辐射衰减系数建立回归模型;最后,基于支持向量机模型实现地表太阳辐射的预测.实验结果表明:本文算法的预测精度明显优于基于统计的传统算法,同时优于基于云团遮挡预测的算法,为准确预测复杂天气条件下的地表太阳能辐射提供重要的参考依据.The intermittency and instability of the solar energy resource brings great challenge to the accurate forecasting of the surface solar irradiation.In this work,based on texture characteristics analysis of sky image collected by TSI combined with SVM model,a new methods for the forecasting of ground solar radiation was presented.First,the texture characteristics related to solar irradiation were extracted from sky images through image processing technologies,including contrast,entropy,grayscale and energy.Second regression model was built between image characteristics and irradiation reduction coefficient.And finally,the SVM model was used to forecasting ground solar radiation.The experiment results indicated that：the forecasting accuracy of the method based on texture characteristics tends to provide better a performance than traditional forecasting method and forecasting method based on cloud block movement prediction.It can provide important reference for the accurate forecasting of surface solar irradiation on complex climate conditions.

关 键 词：地基云图 支持向量机 地表太阳辐射 纹理特征 云辐射衰减 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]