首页|基于pix2pixHD图像修复的光伏电站秒级功率预测方法

基于pix2pixHD图像修复的光伏电站秒级功率预测方法

扫码查看
云团遮挡导致地面辐照度发生瞬变是光伏电站出力剧烈波动的根本原因,为提高在云团遮挡情况下光伏功率预测的精度,提出一种基于pix2pixHD图像修复的光伏电站秒级功率预测方法.首先,依据光伏电站光伏组件的参数、内部光伏阵列的排列结构和布局,推导光伏电站精细化模型;其次,深入挖掘逆变器集群输出的光伏功率数据特征,剖析光伏功率与辐照度的映射关系,构建能够描述云团形状、厚度和运动方向的虚拟云图用以表征云团遮挡(功率缺失)情况;随后,提出生成对抗网络的 pix2pixHD 图像修复算法对缺损的虚拟云图进行修复,融合最近 5 s的修复云图,提高对云团性质的精确表达;最后,依据光伏功率、辐照度、虚拟云图像素值三者之间的线性关系,实现高精度的光伏电站秒级功率预测;以山东某地市的实际光伏电站为例,仿真结果表明所提pix2pixHD图像修复的预测模型能够有效提高秒级光伏功率预测精度.
Second-level Photovoltaic Power Forecasting Method Based on pix2pixHD Image Restoration
The fundamental reason for the drastic fluctuations in the output of photovoltaic power plants is the obstruction of cloud clusters leading to transient changes in ground irradiance.In order to improve the accuracy of photovoltaic power forecasting under cloud cover,this paper proposes a photovoltaic power forecasting method at the second level based on pix2pixHD image restoration.Firstly,a refined model of the photovoltaic power plant is derived based on the parameters of photovoltaic components,the arrangement structure,and layout of internal photovoltaic arrays.Secondly,the data characteristics of photovoltaic power output from the inverter cluster are deeply explored,and the mapping relationship between photovoltaic power and irradiance is analyzed.A virtual cloud image that describes the shape,thickness,and movement direction of cloud clusters is constructed to represent the situation of cloud obstruction(power loss).Subse-quently,a pix2pixHD image restoration algorithm of the generative adversarial network is proposed to restore the defective virtual cloud image.The algorithm integrates the most recent 5 seconds of repaired cloud images to improve the accurate representation of cloud cluster properties.Finally,based on the linear relationship between photovoltaic power,irradiance,and pixel values of virtual cloud images,high-precision photovoltaic power forecasting at the second level is achieved.An actual photovoltaic power station in a city in Shandong is taken as an example,and simulation results indi-cate that the proposed pix2pixHD image restoration forecasting model effectively enhances the accuracy of second-level photovoltaic power forecasting.

second level photovoltaic power forecastingvirtual cloud imagepix2pixHDimage inpaintingcloud oc-clusiondeep learning

孟祥剑、石欣羽、张承慧、张玉敏、杨明

展开 >

山东科技大学电气与自动化工程学院,青岛 266590

山东大学控制科学与工程学院,济南 250000

山东大学电气工程学院,济南 250000

秒级光伏功率预测 虚拟云图 pix2pixHD 图像修复 云团遮挡 深度学习

国家自然科学基金山东省自然科学基金

52107111ZR2023QE259

2024

高电压技术
中国电力科学研究院 中国电机工程学会

高电压技术

CSTPCD北大核心
影响因子:2.32
ISSN:1003-6520
年,卷(期):2024.50(9)
  • 16