State Perception Method for Overhead Transmission Lines Based on Multiple Sensing System Networking and Federated Learning

A state perception method for overhead transmission lines based on multiple sensing system networking and federated learning is proposed to address the problems of time-consuming, labor-intensive, and low accuracy in traditional transmission line state judgment. Firstly, multiple sets of monitoring devices for sensor system networking are designed to achieve unified access and management of multiple types of data. Then, the long short-term memory network (LSTM-PSO) improved by particle swarm optimization algorithm is used to extract features from sensor data, and a lightweight segmentation network is used for sensor image state recognition. Finally, based on the federated learning architecture, LSTM-PSO and lightweight segmentation network are deployed at edge nodes to achieve information analysis of various sensors, and the FedAdam aggregation algorithm is used to obtain the line state perception results. Based on the selected overhead line dataset, experimental analysis is conducted on the proposed method, and the results show that its state perception accuracy exceeds 95% in daytime environments, and the perception time is about 1.9s, demonstrating good comprehensive performance and engineering practice ability.