Distributed Cloud Detection Method for 5G Charging Devices Based on Invariance Constraint
Xuling Li, Zhenlan Dou, Yanhua Ma, Xuan Zhang, Xuefeng He · 2022
In this paper, an invariance constraint-based federated learning method is proposed to implement the distributed cloud detection for 5G charging devices. In order to solve the problem that the charging devices have large numbers and scattered distribution, a distributed scenario modeling of the charging devices is proposed. Each charging device is regarded as a child node, and a central node is constructed on cloud end. In this modeling scenario, the 5G network is used to transmit only the model and gradient parameters between the central node and each child node, which avoids high cost of data transmission and high burden of centralized data storage. In addition, the invariance-based loss function is designed to solve the problem that the data distribution of each child node has significant statistical heterogeneity in the train stage. The proposed method makes the model generalize better, and improves the accuracy of the safety detection of charging devices.