Detection of Hydrophobicity Grade of Composite Insulators Based on MDC‐YOLO Algorithm DOI Creative Commons
Shaotong Pei,

Weiqi Wang,

Chenlong Hu

et al.

Energy Science & Engineering, Journal Year: 2025, Volume and Issue: unknown

Published: April 24, 2025

ABSTRACT In the field of power equipment inspection, aging condition composite insulators is often determined by detection water repellency. However, existing methods are difficult to effectively extract repellency level features in complex background, and it meet real‐time requirements. Therefore, this paper proposes a MDC‐YOLO algorithm for classification insulators. The combines efficient convolutional network (EfficientNetV2), deformable attention mechanism (DAttention), lightweight convolution (CSPPC), which significantly realizes network, at same time improves accuracy insulator umbrella skirt identification, class classification. Experimentally verified, multilayer YOLO proposed paper, i.e., algorithm, 5.19% reduces GFLOPs 3.5, Top‐1 grade 4.654% 1.1. results present research can be widely applied study used fields smart grid construction state assessment, provides strong support technical development related fields. It meets requirements hydrophobicity insulators, proves effectiveness superiority through ablation comparison tests.

Language: Английский

Detection of Hydrophobicity Grade of Composite Insulators Based on MDC‐YOLO Algorithm DOI Creative Commons
Shaotong Pei,

Weiqi Wang,

Chenlong Hu

et al.

Energy Science & Engineering, Journal Year: 2025, Volume and Issue: unknown

Published: April 24, 2025

ABSTRACT In the field of power equipment inspection, aging condition composite insulators is often determined by detection water repellency. However, existing methods are difficult to effectively extract repellency level features in complex background, and it meet real‐time requirements. Therefore, this paper proposes a MDC‐YOLO algorithm for classification insulators. The combines efficient convolutional network (EfficientNetV2), deformable attention mechanism (DAttention), lightweight convolution (CSPPC), which significantly realizes network, at same time improves accuracy insulator umbrella skirt identification, class classification. Experimentally verified, multilayer YOLO proposed paper, i.e., algorithm, 5.19% reduces GFLOPs 3.5, Top‐1 grade 4.654% 1.1. results present research can be widely applied study used fields smart grid construction state assessment, provides strong support technical development related fields. It meets requirements hydrophobicity insulators, proves effectiveness superiority through ablation comparison tests.

Language: Английский

Citations

0