基于XGB算法及全天空成像仪图像的超短期DNI预测

ULTRA-SHORT-TERM DNI PREDICTION BASED ON XGB ALGORITHM AND ALL-SKY IMAGER IMAGES

吸热器是塔式太阳能热发电站的核心设备,但云层短时间的遮挡和离开会使吸热器接收的太阳直接辐射(DNI)发生随机波动,造成吸热器表面受热发生剧烈变化,会影响塔式太阳能热发电站的安全运行。本文提出了一种基于极值梯度提升(eXtreme Gradient Boosting,XGB)算法及全天空成像仪图像的超短期DNI预测方法。通过使用基于太阳位置的全天空成像仪标定方法,完成全天空成像仪的姿态校正和畸变矫正;对晴空背景参数进行拟合并建立晴空库,为云层的检测提供基础;通过对太阳及其临近区域进行插值修复,使1 min内的DNI预测成为可能;根据云的运动矢量,利用外推法得到未来云层可能会遮挡太阳的图像点,并提取图像特征,用于基于XGB算法的超短期DNI预测模型的训练,最后利用训练后的模型完成超短期DNI预测。

The heat absorber is the core equipment of tower CSP station.The direct solar radiation(DNI)received by the heat absorber will fluctuate randomly due to the short-time occlusion and departure of clouds,resulting in drastic changes in the surface of the heat absorber,which affects the safe operation of the CSP station.This paper proposes an ultra-short-term DNI prediction method based on eXtreme Gradient Boosting(XGB)and all-sky imager images.The attitude correction and distortion correction of the all-sky imager are completed by using sun-position-based all-sky imager calibration method.Accurately fit background parameters of clear sky and establish a clear sky library,which provides a basis for cloud layer detection.Enables DNI predictions within 1 min by interpolating repairs to the sun and its immediate vicinity.According to the motion vector of the cloud,the image point where the cloud may block the sun in the future is obtained by extrapolation,and the image features are extracted for the training of the ultra-short-term DNI prediction model based on the XGB algorithm.Finally,the ultra-short-term DNI prediction is completed with the trained model.