Overhead conductors are critical components of power transmission systems, and their corrosion-induced degradation poses severe threats to grid stability and operational safety. Traditional corrosion inspection methods rely on manual patrols or single-source data analysis, suffering from inefficiency, subjectivity, and limited accuracy in dynamic environments. To address these limitations, this paper proposes a real-time corrosion estimation framework for overhead conductors by fusing UAV-acquired RGB images and load current data. First, an improved YOLOv8 model integrated with a Coordinate Attention (CA) module is designed to segment corroded regions from RGB images, enabling quantitative extraction of corrosion features (e.g., rust layer fractal dimension, crack density). Second, a physics-informed corrosion rate model is established to characterize the synergistic effect of load current and environmental factors, based on Faraday’s electrolysis law and corrosion electrochemistry theory. Third, a Bayesian network-based data fusion strategy is developed to integrate image-derived features and current-based corrosion rates, realizing accurate estimation of corrosion severity (intact, mild, moderate, severe, failure risk). Experiments are conducted on a 110kV transmission line segment in an industrial polluted area, where 21026 UAV RGB images and 3 months of load current data are collected. The results demonstrate that the proposed framework achieves a corrosion severity classification accuracy of 92.3% outperforming single-source methods and state-of-the-art fusion models. The inference speed reaches 28FPS, satisfying real-time inspection requirements. This study provides a reliable technical solution for intelligent corrosion monitoring of overhead conductors, supporting data-driven power grid maintenance decisions.
Wan Niu (Sun,) studied this question.