基于地基云图的三维体素积云建模方法研究

陈雨璇; 王诗怡; 廖轶鹏; 黎东洲; 高瑾瑾

doi:10.13928/j.cnki.wrahe.2025.10.012

水利水电技术（中英文） ›› 2025, Vol. 56 ›› Issue (10) : 152 -164. DOI: 10.13928/j.cnki.wrahe.2025.10.012

碳中和与清洁能源开发利用专栏

基于地基云图的三维体素积云建模方法研究

陈雨璇 ¹^,² ,
王诗怡 ¹^,² ,
廖轶鹏 ¹^,² ,
黎东洲 ¹^,² ,
高瑾瑾 ¹^,²

作者信息 +

Research on 3D voxel-based cumulus cloud modeling method using ground-based cloud images

Yuxuan CHEN ¹^,² ,
Shiyi WANG ¹^,² ,
Yipeng LIAO ¹^,² ,
Dongzhou LI ¹^,² ,
Jinjin GAO ¹^,²

Author information +

文章历史 +

PDF

摘要

【目的】积云的快速生成与移动是导致光伏发电输出功率波动的主要气象因素，其对太阳辐照度的遮蔽效应显著影响光伏系统的运行稳定性和功率预测精度。然而，传统基于二维图像的积云建模方法难以反映云体在垂直方向的结构变化，缺乏对云基高度、云团厚度及内部粒子分布等关键物理参数的有效表达，难以满足光伏应用场景下对高分辨率、实时性和物理一致性的建模需求。【方法】为此，提出一种基于地基云图的三维体素积云建模方法，通过图像预处理与空间配准实现积云区域的地理定位，提出面向积云对象的云基高度和云团厚度自动提取方法，并引入云粒子密度参数，构建具有物理约束特征的体素表达模型，结合GPU渲染管线实现积云外形与内部特征的高效建模与可视化。【结果】试验结果表明，所提方法计算的云基高度与遥感影像反演结果的相对误差在5%以内，云粒子密度分布与CloudSat剖面观测一致，GPU方法在建模效率和渲染性能方面优于传统方法，三维场景交互帧率稳定。【结论】研究表明，该方法可实现积云的快速重建与高效可视化，验证了基于地基云图提取关键参数的可行性和准确性，为光伏功率预测提供了有效的模型与数据支撑。

Abstract

[Objective] The rapid formation and movement of cumulus clouds are the main meteorological factors causing fluctuations in photovoltaic(PV) power output, and their shading effects on solar irradiance significantly affect the operational stability of PV systems and the accuracy of power prediction. However, traditional cumulus cloud modeling method based on 2D images fail to reflect the structural variation of clouds in the vertical direction and lack effective expression of key physical parameters such as cloud base height, cloud thickness, and internal particle distribution, making it difficult to meet the modeling requirements for high resolution, real-time performance, and physical consistency in PV applications. [Methods] To address this, a 3D voxel-based cumulus cloud modeling method using ground-based cloud images was proposed. The geographic positioning of cumulus regions was achieved through image preprocessing and spatial registration. An automatic extraction method for cloud base height and cloud thickness for cumulus clouds was proposed, and a cloud particle density parameter was introduced to construct a voxel expression model with physical constraints. Combined with a GPU rendering pipeline, efficient modeling and visualization of cumulus cloud morphology and internal characteristics were realized. [Results] The experimental result showed that the calculated cloud base height had a relative error within 5% compared to remote sensing inversion result. The cloud particle density distribution was consistent with CloudSat profile observations. The GPU-based method outperformed traditional method in modeling efficiency and rendering performance, maintaining stable frame rates in 3D scene interaction. [Conclusion] The findings demonstrate that this method can achieve rapid reconstruction and efficient visualization of cumulus clouds, verifying the feasibility and accuracy of extracting key parameters based on ground-based cloud images and providing an effective model and data support for PV power prediction.