Physics-Informed Deep Learning for Electric Field Distortion Analysis in Power Transmission Line Foreign Object Recognition

Foreign object intrusion remains a persistent challenge in power transmission line monitoring, with traditional visual-based approaches suffering from environmental sensitivity and limited interpretability. This paper introduces a physics-informed deep learning framework that exploits electric field distortions for robust object recognition. We develop a comprehensive 49-dimensional feature space grounded in electromagnetic theory, capturing complex conductor-object interactions. A conditional autoencoder with spatial attention learns interpretable latent representations, while a residual network provides baseline field predictions. Simulation results demonstrate 98.5% classification accuracy across five object categories, maintaining robust performance under incomplete sensor observations.