CloudSwinNetLite：轻量级地基云图细粒度分割网络

石超君; 谢雄彬; 张珂; 苏子博; 张筱筠; 李星宽

doi:doi:10.19912/j.0254-0096.tynxb.2024-2425

English Version Reader LoginJournal Login

您当前的位置：

首页 >

文章列表页 >

CloudSwinNetLite：轻量级地基云图细粒度分割网络

更新时间：2025-12-10

- CloudSwinNetLite：轻量级地基云图细粒度分割网络
- Issue 11, Pages: 374-382(2025)
- 作者机构：
- 作者简介：
- 基金信息：
- DOI：doi:10.19912/j.0254-0096.tynxb.2024-2425
  CLC： P183
- Published Online：04 December 2025，
  
  Published：04 December 2025
- 稿件说明：
移动端阅览
石超君, 谢雄彬, 张珂, et al. CloudSwinNetLite：轻量级地基云图细粒度分割网络[J]. 2025, (11): 374-382.
DOI：

石超君, 谢雄彬, 张珂, et al. CloudSwinNetLite：轻量级地基云图细粒度分割网络[J]. 2025, (11): 374-382. DOI： doi:10.19912/j.0254-0096.tynxb.2024-2425.

摘要

在现有的云图分割研究中

大多数方法仅简单将天空和云层进行分割

忽视了不同类型云层之间的差异。且现有模型计算开销较大

运行效率较低

难以满足实时分割云图的需求

无法有效支持超短期光伏功率预测。针对此问题

提出一种轻量级云图细粒度分割模型CloudSwinNetLite。该模型采用编码-解码结构

以Swin Transformer作为基础模块

使用轻量级多头自注意力机制（W-GLWMSA）和基于Ghost的反向残差前馈网络（GIRFFN）

在保持网络分割性能的同时

实现计算轻量化。实验结果表明

CloudSwinNetLite与其他分割模型相比

能够在较低计算开销下实现对不同类型云层更加有效的分割。

Abstract

In existing cloud image segmentation research

most methods only distinguish between the sky and clouds

neglecting the inherent differences among various cloud types. In addition

many current models are computationally intensive and inefficient

making them unsuitable for the real-time segmentation of cloud images and thus limiting their application in ultra-short-term photovoltaic power forecasting. To overcome these challenges

we propose CloudSwinNetLite

a lightweight fine-grained cloud image segmentation network. The model adopts an encoder-decoder structure with Swin Transformer as the base module

using a lightweight multi-head self-attention mechanism （W-GLWMSA） and a ghost-based inverted residual feed-forward network （GIRFFN）. These designs jointly enhance computational efficiency while preserving high segmentation accuracy. Experimental results demonstrate that CloudSwinNetLite achieves effective fine-grained cloud segmentation with substantially reduced computational cost compared with existing models.

关键词

Keywords

references

王育飞, 郝德扬, 薛花, 等. 计及云图和混沌特性的光伏功率组合预测方法[J]. 太阳能学报, 2023, 44(12): 74-81.

刘源延, 孔小兵, 马乐乐, 等. 基于小波包变换与深度学习的超短期光伏功率预测[J]. 太阳能学报, 2024, 45(5): 537-546.

王东风, 刘婧, 黄宇, 等. 结合太阳辐射量计算与CNN-LSTM组合的光伏功率预测方法研究[J]. 太阳能学报, 2024, 45(2): 443-450.

魏亮, 朱婷婷, 过奕任, 等. 基于DAR-CapsNet的地基云图云分类[J]. 太阳能学报, 2023, 44(11): 189-195.

SHI C J, ZHOU Y T, QIU B, et al.CloudU-Net: a deep convolutional neural network architecture for daytime and nighttime cloud images’ segmentation[J]. IEEE Geoscience and remote sensing letters, 2021, 18(10): 1688-1692.

SHI C J, ZHOU Y T, QIU B.CloudU-Netv2: a cloud segmentation method for ground-based cloud images based on deep learning[J]. Neural processing letters, 2021, 53(4): 2715-2728.

SHI C J, ZHOU Y T, QIU B.CloudRaedNet: residual attention-based encoder-decoder network for ground-based cloud images segmentation in nychthemeron[J]. International journal of remote sensing, 2022, 43(6): 2059-2075.

ZHANG Z, YANG S Z, LIU S, et al.Ground-based cloud detection using multiscale attention convolutional neural network[J]. IEEE geoscience and remote sensing letters, 2022, 19: 8019605.

方明, 张利箭. 一种基于决策融合策略的全天空地基云图云量估计方法[J]. 太阳能学报, 2023, 44(10): 245-254.

LIU S, ZHANG J F, ZHANG Z, et al.Integration transformer for ground-based cloud image segmentation[J]. IEEE transactions on geoscience and remote sensing, 2023, 61: 5606712.

SHI C J, SU Z B, ZHANG K, et al.CloudFU-Net: a fine-grained segmentation method for ground-based cloud images based on an improved encoder-decoder structure[J]. IEEE transactions on geoscience and remote sensing, 2024, 62: 5619913.

SHI C J, SU Z B, ZHANG K, et al.CloudSwinNet: a hybrid CNN-transformer framework for ground-based cloud images fine-grained segmentation[J]. Energy, 2024, 309: 133128.

YE L, CAO Z G, XIAO Y, et al.Supervised fine-grained cloud detection and recognition in whole-sky images[J]. IEEE transactions on geoscience and remote sensing, 2019, 57(10): 7972-7985.

CAO H, WANG Y Y, CHEN J, et al.Swin-Unet: unet-like pure transformer for medical image segmentation[C]//Computer Vision-ECCV 2022 Workshops. Cham, Switzed and, Springer, 2023: 205-218.

张臻, 陈天鹏, 王磊, 等. 基于地基云图的超短期太阳辐照预测方法与装置研究[J]. 太阳能学报, 2023, 44(1): 133-140.

石超君, 李星宽, 张珂, 等. 地基云图分割方法研究进展[J]. 计算机工程与应用, 2023, 59(13): 1-16.

朱婷婷. 基于地基云图的太阳直接辐照度超短期预测研究[D]. 南京: 东南大学, 2019.

ZHOU Z W, SIDDIQUEE M M R, TAJBAKHSH N, et al. UNet: redesigning skip connections to exploit multiscale features in image segmentation[J]. IEEE transactions on medical imaging, 2020, 39(6): 1856-1867.

RONNEBERGER O, FISCER P, BROX T.Attention U-Net: learning where to look for the pancreas[C]// 2018 Medical Image Computing and Computer-Assisted Intervention. Madrid, Spain, 2018.

SUN K, XIAO B, LIU D, et al.Deep high-resolution representation learning for human pose estimation[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Long Beach, CA, USA, 2020: 5686-5696.

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

基于BO-SVM和ISO改进的ENN光伏功率超短期预测模型

K-ISSA-LSTM PHOTOVOLTAIC POWER PREDICTION BASED ON POWER CHARACTERISTIC

PHOTOVOLTAIC POWER PROBABILITY PREDICTION BASED ON ATTENTION TIME CONVOLUTIONAL NEURAL NETWORK

一种高分辨率的光伏短期功率预测模型

Related Author

王育飞

郝德扬

薛花

米阳

汪弦哲

余光正

杨秀

Jin Weiyong

Related Institution

Electrical Engineering College, Zhejiang University

Information Engineering College, Hangzhou Dianzi University

State Grid Wenzhou Electric Power Supply Company

School of Control and Computer Engineering, North China Electric Power University

College of Electrical Engineering, Xinjiang University

AI问答

⁰