基于改进长短期记忆网络的新能源场站网络安全评估方法研究  

Research on new energy station network security assessment method based on improved LSTM network

作　　者：刘珊 李瑞 王尧 LIU Shan;LI Rui;WANG Yao(State Grid Shanxi Electric Power Research Institute,Taiyuan 030001,China;State Grid Shanxi Electric Power Company,Taiyuan 030021,China)

机构地区：[1]国网山西省电力公司电力科学研究院,山西太原030001 [2]国网山西省电力公司,山西太原030021

出　　处：《电信科学》2024年第10期124-133,共10页Telecommunications Science

基　　金：国网山西省电力公司科技项目(No.520530230009)。

摘　　要：为了解决新能源大规模并网造成现有新能源场站网络安全防护体系无法满足网络异常监测和告警需求的问题,提出一种基于改进长短期记忆网络的新能源场站网络安全评估方法。首先,根据新能源场站网络系统架构,分析网络安全发生原因;其次,基于随机森林算法求解新能源场站网络流量的基尼系数,进而求出网络流量所有特征的重要系数,选出重要特征;最后,将重要特征输入长短期记忆网络中,利用注意力机制自适应分配数据的时间和特征,加强对网络流量中重要时间和特征的重视,进而提高模型对网络安全评估的准确性。试验结果表明,该方法能够准确评估新能源场站网络安全状态,与支持向量机、卷积神经网络、传统长短期记忆网络相比,评估准确率分别提升了12.65%、9.34%、8.79%,提升了新能源电力系统的网络安全状态感知、评价和告警能力。In order to solve the problem of the inability of the existing network security protection system for new energy stations to meet the needs of network anomaly monitoring and alarm caused by the large-scale integration of new energy,a new energy station network security assessment method based on an improved long short-term memory network was proposed.Firstly,based on the architecture of the new energy station network system,the reasons for network security incidents were analyzed.Secondly,based on the random forest algorithm,the Gini coefficient of new energy station network traffic was solved,and then the important coefficients of all network traffic features were calculated to select important features.Finally,important features were input into the long short-term memory network,and attention mechanisms were used to adaptively allocate data time and features,strengthening the emphasis on important time and features in network traffic,thereby improving the accuracy of the model for network security assessment.The experimental results show that this method can accurately evaluate the network security status of new energy power stations.Compared with support vector machines,convolutional neural networks,and traditional long short-term memory networks,the evaluation accuracy has been improved by 12.65%,9.34%and 8.79%,respectively,enhancing the perception,evaluation,and alarm capabilities of network security status in new energy power systems.

关 键 词：新能源场站 网络安全 长短期记忆网络 随机森林算法 注意力机制 

分 类 号：TM73[电气工程—电力系统及自动化]