一种基于深度学习的充电桩网络异常检测模型

An Anomaly Detection Model of Charging Pile Network Based on Deep Learning

针对分布式充电桩网络安全评估时存在评估效果差、计算量大、执行效率低等问题,提出一种基于深度学习的命令级异常检测模型。基于改进长短时记忆网络训练样本数据,提升模型泛化能力和鲁棒性。以国家电网公司车路协同充电桩网络平台中的公共用户订阅者/发布者模型为例,对所提模型进行验证。仿真结果表明,所提模型可基于更小的学习样本实现更高的学习效果,从而为分布式配电网安全、可靠运行提供借鉴。

A command level anomaly detection model based on deep learning is proposed to solve the problems of poor evaluation effect,large amount of calculation and low execution efficiency in the network security evaluation of distributed charging piles.The sample data is trained based on the improved long short-time memory network,so as to improve the generalization ability and robustness of the model.The public subscriber/publisher model in the vehicle road collaborative charging pile network platform of State Grid Corporation is taken as an example to verify the proposed model.The simulation results show that the proposed model can achieve higher learning effect based on smaller learning samples,which can provide reference for the safe and reliable operation of distributed distribution network.