服务于区域光伏预测的天空图像K-means云空辨识模型

发布时间：2018-07-12 16:13

本文选题：初始聚类中心 + 天空图像　；参考：《华北电力大学学报(自然科学版)》2017年06期

【摘要】：地基天空图像的云空辨识及云团预测是研究区域电网光伏发电功率分布与变化的前提,对支撑调度提高光伏发电消纳比例具有重要意义。首先分别采用较高的红蓝分量比值和较低的红蓝分量比值作为固定阈值分割地基天空图像,依次提取辨识结果中的天空像素点和云像素点的位置信息并获取原图像中对应位置的天空像素点和云像素点的RGB值;其次对获得的天空像素和云像素求均值并将各自均值中的红蓝分量相除获取初始聚类中心;然后使用K-means算法,利用加权欧式距离计算每一个聚类样本与聚类中心之间的距离,通过数次迭代得到聚类结果,进而将聚类结果还原成矩阵得到地基天空图像的云空辨识结果图;最后利用云南某光伏电站全天空成像仪TSI-VIS-J1006采集的天空图像进行仿真,结果表明该方法较固定阈值法的收敛速度更快、聚类精度更高,能够有效实现地基天空图像的云空辨识。
[Abstract]:The cloud space identification and cloud cluster prediction of ground-based sky image is the premise of studying the distribution and variation of photovoltaic power generation in regional power grid. It is of great significance to support dispatching and improve the absorption ratio of photovoltaic power generation. At first, the higher ratio of red and blue component and the lower ratio of red and blue component are used as fixed threshold to segment the ground-based sky image. The position information of sky pixel and cloud pixel in the identification result is extracted in turn, and the RGB value of sky pixel and cloud pixel of corresponding position in the original image is obtained. Secondly, the average value of sky pixel and cloud pixel is obtained and the initial clustering center is obtained by dividing the red and blue components of the respective mean. Then, the distance between each cluster sample and the clustering center is calculated by using the weighted Euclidean distance using K-means algorithm. The clustering results are obtained by several iterations, and then the clustering results are reduced to matrix to obtain the cloud-space identification results of the ground-based sky images. Finally, the sky images collected by the all-sky imager TSI-VIS-J1006 of a photovoltaic power station in Yunnan are simulated. The results show that the proposed method has faster convergence speed and higher clustering accuracy than the fixed threshold method, and it can effectively realize the cloud space identification of the ground-based sky image.
【作者单位】：云南电网有限责任公司电力科学研究院;新能源电力系统国家重点实验室;美国伊利诺伊大学厄巴纳-香槟分校;澳大利亚国立大学环境与社会学院;
【基金】：国家自然科学基金资助项目(51577067,51277075) 北京市自然科学基金资助项目(3162033) 河北省自然科学基金资助项目(E2015502060) 河北省科技支撑计划重点项目(12213913D) 云南省新能源重大科技专项(2013ZB005) 新能源电力系统国家重点实验室开放课题(LAPS15009,LAPS16007) 中央高校基本科研业务费重点项目(2014ZD29) 云南电网有限责任公司科技项目(YNKJQQ00000280)
【分类号】：TM615

【相似文献】