基于联邦学习的电力施工场景分类

Classification of Power Construction Scenes Based on Federated Learning

深度学习的快速发展促进了人工智能在各个领域的应用,包括智能安防。但是施工现场的图像往往涉及到公司的隐私,不方便进行长距离的传输。为了实现电网施工现场的智能管控,论文将联邦学习与深度学习图像分类算法结合,提出了一种基于联邦学习的电力施工场景分类方法,该方法保护数据隐私和数据安全的前提下,利用各节点的数据联合训练得到具有高准确率的电力场景分类模型。同时,论文依据客户节点的数据量提出了动态权重算法。实验证明,该方法可以加速联邦学习中心模型的拟合速度,并提高准确率。

The rapid development of deep learning has promoted the application of artificial intelligence in various fields,including intelligent security.However,the images of the construction site often involve the privacy of the company and are not convenient for long-distance transmission.In order to intelligently control the power grid construction site,this paper combines federated learning and deep learning algorithms to propose a classification method for power construction scenes.This method uses the data of each client node to jointly train a high-accuracy scene classification model,which can protect data privacy and data security.In addition,this paper proposes a dynamic weighting algorithm based on the data volume of client nodes.Experiments results show that this method can accelerate the fitting speed of the center model and improve the accuracy.