Multi branches dilated CNN federated learning for transmission line fault diagnosis
Wenhao Sun, Hongbo Ma, Wei Li, Yangyang Pan, Fei Hao, Tao Wang · International Conference on Mechanical Engineering, Measurement Control, and Instrumentation · 2021
Using convolutional neural network for transmission line fault diagnosis is an accurate and effective method, but it relies on a large amount of data with positive labels and is limited by kernel size that decides the receptive field. However, it is difficult to centralize data in reality, which causes low accuracy of the model. Federated learning has made great progress recently, and it is possible to train a model with high accuracy without centralizing data. In this paper, we propose a transmission line fault diagnosis method based on multi-branch convolutional neural network combined with federated learning. First, we design a novel three branches network with two dilated convolution kernels to increase the receptive field of the kernel. Then we integrate it into the federated learning framework to expand the amount of data used to train the model while preserving data security and privacy. The experimental results show that our method is feasible and can effectively improve the accuracy of the model, and provides a new idea for transmission line fault diagnosis.