Research on Transmission Line Defect Detection Based on Adaptive Federated Learning

Abstract

The existing deep learning transmission line detection technology with cloud computing is faced with problems such as slow response speed, high communication cost, and difficult to obtain data scattered, as well as the huge amount of data, which causes huge pressure on cloud storage capacity and processing capacity. This paper proposes a transmission line defect detection technology based on adaptive federated learning (FL). Its advantage is that data does not need to be uploaded and shared, which not only reduces communication costs, but also improves data security. In this paper, an adaptive algorithm is added to the original FL algorithm, which can adaptively change the data volume of the next round of training according to the training effect of each round and the local training energy consumption, so as to achieve the optimal number of communication between the two, which greatly reduces the Improve training speed and reduce communication costs. Through experimental analysis, the model training efficiency of the adaptive FL proposed in this paper is 70% higher than that of the centralized cloud computing, and the computing cost is saved by about 40%.