基于多模态语义理解与强化学习的绝缘子缺陷多场景自适应检测方法

戈睛天; 赖志超; 陆 峰; 王宝金; 王晓晶

doi:10.19783/j.cnki.pspc.251270

欢迎来到能源电力期刊网！

English Version 期刊登录

您当前的位置：

首页 >

文章列表页 >

基于多模态语义理解与强化学习的绝缘子缺陷多场景自适应检测方法

设计开发 | 更新时间：2026-09-23

- 基于多模态语义理解与强化学习的绝缘子缺陷多场景自适应检测方法
- A multi-scenario adaptive detection method for insulator defects based on multimodal semantic understanding and reinforcement learning
- 电力系统保护与控制 2026年54卷第18期
- 作者机构：
- 作者简介：
- 基金信息：
- DOI：10.19783/j.cnki.pspc.251270
  中图分类号：
- 纸质出版：2026
- 稿件说明：
移动端阅览
戈睛天, 赖志超, 陆峰, 等. 基于多模态语义理解与强化学习的绝缘子缺陷多场景自适应检测方法[J]. 电力系统保护与控制, 2026,54(18).

GE Jingtian, LAI Zhichao, LU Feng, et al. A multi-scenario adaptive detection method for insulator defects based on multimodal semantic understanding and reinforcement learning[J]. Power System Protection and Control, 2026, 54(18).
戈睛天, 赖志超, 陆峰, 等. 基于多模态语义理解与强化学习的绝缘子缺陷多场景自适应检测方法[J]. 电力系统保护与控制, 2026,54(18). DOI： 10.19783/j.cnki.pspc.251270.

GE Jingtian, LAI Zhichao, LU Feng, et al. A multi-scenario adaptive detection method for insulator defects based on multimodal semantic understanding and reinforcement learning[J]. Power System Protection and Control, 2026, 54(18). DOI： 10.19783/j.cnki.pspc.251270.

摘要

针对高压输电线路绝缘子缺陷检测中传统方法鲁棒性差，深度学习模型跨场景泛化性能不足的问题，提出一种基于多模态语义理解与强化学习的多场景自适应检测方法。首先，利用对比语言-图像预训练(contrastive language-image pre-training

CLIP)多模态语义编码器提取图像与场景标签的语义特征，通过余弦相似度匹配实现场景类型识别。然后，基于场景识别结果，采用多模态生成模型对训练数据进行针对性增强，并构建多尺度YOLO检测集群以适应不同场景下目标检测的需求。最后，引入强化学习调度机制，根据实时检测反馈动态选择最优检测模型与参数，形成闭环自适应优化。实验结果表明，该方法在雨天、雾霾、逆光、夜间等多种复杂场景下均能保持高检测精度与鲁棒性，优于现有单一模型方法，具有良好的工程应用前景。

Abstract

To address the poor robustness of traditional image processing methods and insufficient cross-scenario generalization of deep learning models for insulator defect detection on high-voltage transmission lines

a multi-scenario adaptive detection method based on multimodal semantic understanding and reinforcement learning is proposed. First

a contrastive language-image pre-training (CLIP) multimodal semantic encoder is employed to extract semantic features from both images and scene labels

enabling scene type identification via cosine similarity matching. Then

based on the identified scene type

a multimodal generative model is applied for targeted data augmentation

and a multi-scale YOLO detection cluster is constructed to accommodate object detection requirements under different scenarios. Finally

a reinforcement learning scheduling mechanism is introduced to dynamically select the optimal detection model and parameters according to real-time detection feedback

thereby forming a closed-loop adaptive optimization. Experimental results demonstrate that the proposed method maintains high detection accuracy and robustness in multiple complex scenarios

including rainy

foggy

backlit

and nighttime conditions

outperforming existing single-model approaches and exhibiting strong potential for engineering applications.

关键词

Keywords

references

XU Z, TANG X. Transmission line insulator defect detection algorithm based on MAP-YOLO8[J]. Scientific Reports, 2025, 15(1).

YAN K, WANG F, ZHANG Z, et al. Edge detection of composite insulators hydrophobic image based on improved canny operator[J]. Energy and Power Engineering, 2013, 5(4): 593-596.

PERNEBAYEVA D, BAGHERI M, JAMES A. High voltage insulator surface evaluation using image processing[C] // 2017 International Symposium on Electrical Insulating Materials (ISEIM), September 11-15, 2017, Toyohashi, Japan: 520-523.

LIU Y, HUANG X, LIU D. Weather-domain transfer- based attention YOLO for multi-domain insulator defect detection and classification in UAV images[J]. Entropy, 2024, 26(2): 1-18.

POURSAEED A H, NAMDARI F. Online transient stability assessment implementing the weighted least-square support vector machine with the consideration of protection relays[J]. Protection and Control of Modern Power Systems, 2025, 10(1): 1-17.

JIANG H, JIN L, YAN S. Recognition and fault diagnosis of insulator string in aerial images[J]. Journal of Mechanical Electrical Engineering, 2015, 32(2): 274-278.

DINH N, HOANG V T, LE V T, et al. YOLOv10- HQGNN: a hybrid quantum graph learning framework for real-time faulty insulator detection[J]. Computers, Materials Continua, 2026, 86(3): 1-23.

XU J, LIAO H, LI K, et al. Multi-scale feature fusion transformer with hybrid attention for insulator defect detection[J]. IEEE Transactions on Instrumentation and Measurement, 2025, 74.

XI Y, ZHOU K, MENG L W, et al. Transmission line insulator defect detection based on swin transformer and context[J]. Machine Intelligence Research, 2023, 20(5): 729-740.

ZHANG L, SONG M, GUO H, et al. Insulator defect detection via a residual denoising diffusion mechanism[J]. Materials, 2025, 18(8).

HIEN A T, TRAN A D, VIET D C, et al. Insulator defect detection based on feature pyramid network and diffusion model[J]. Signal, Image and Video Processing, 2025, 19(5): 1-11.

XU C, ZHENG C, ZHANG J. UAVs detect hazards with multi-directional Mamba on overhead transmission lines[J]. Scientific Reports, 2025, 15(1).

MENG F, QIN Y, WU Y, et al. A subtle defect recognition method for catenary fastener in high-speed railroad using destruction and reconstruction learning[J]. Advanced Engineering Informatics, 2024, 60.

LIU C, WU Y, LIU J, et al. Insulator faults detection in aerial images from high-voltage transmission lines based on deep learning model[J]. Applied Sciences, 2021, 11(10).

LI Y, NI M, LU Y. Insulator defect detection for power grid based on light correction enhancement and YOLOv5 model[J]. Energy Reports, 2022, 8: 807-814.

WANG H, YANG Q, ZHANG B, et al. Deep learning based insulator fault detection algorithm for power transmission lines[J]. Journal of Real-time Image Processing, 2024, 21(4).

ROMBACH R, BLATTMANN A, LORENZ D, et al. High-resolution image synthesis with latent diffusion models[C] // Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, June 18-24, 2022, New Orleans, LA, USA: 10684-10695.

HO J, JAIN A, ABBEEL P. Denoising diffusion probabilistic models[C] // Advances in Neural Information Processing Systems, December 6-12, 2020, Vancouver, Canada: 6840-6851.

RADFORD A, KIM J W, HALLACY C, et al. Learning transferable visual models from natural language supervision[C] // International Conference on Machine Learning, PMLR, July 18-24, 2021, Vienna, Austria: 8748-8763.

ZHENG Z, WANG P, LIU W, et al. Distance-IoU loss: faster and better learning for bounding box regression[C] // Proceedings of the AAAI Conference on Artificial Intelligence, February 7-12, 2020, New York, USA: 12993-13000.

VAN HASSELT H, GUEZ A, SILVER D. Deep reinforcement learning with double Q-learning[C] // Proceedings of the AAAI Conference on Artificial Intelligence, February 12-17, 2016, Phoenix, USA: 2094-2100.

MNIH V, KAVUKCUOGLU K, SILVER D, et al. Human- level control through deep reinforcement learning[J]. Nature, 2015, 518: 529-533.

EVERINGHAM M, ESLAMI S M A, VAN GOOL L, et al. The pascal visual object classes challenge: a retrospective[J]. International Journal of Computer Vision, 2015, 111(1): 98-136.

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

基于深度强化学习的区域综合能源系统主动调节灵活性规则提取及可解释优化调度

基于RLMTS的小样本风力发电机齿轮箱故障检测

基于行为克隆TD3强化学习的低碳园区柔性资源优化策略

数据孤岛下基于联邦学习的用户电价响应刻画及其应用

兼顾激励相容的水电富集电网中长期电力市场交易模型