箱式变电站外围恒温网络入侵的动态检测研究  

Research on Dynamic Detection of Peripheral Thermostatic Network Intrusion in Box-Type Substation

作　　者：陈黎明 杜海红 王冬冬 李建泽 朱明星 CHEN Liming;DU Haihong;WANG Dongdong;LI Jianze;ZHU Mingxing(State Grid Anhui Electric Power Co.,Ltd.,Hefei 230022,China;State Grid Fuyang Supply Company,Fuyang 236000,China;State Grid Bengbu Supply Company,Bengbu 233000,China)

机构地区：[1]国网安徽省电力有限公司,安徽合肥230022 [2]国网阜阳供电公司,安徽阜阳236000 [3]国网蚌埠供电公司,安徽蚌埠233000

出　　处：《自动化仪表》2024年第5期94-97,102,共5页Process Automation Instrumentation

基　　金：国家自然科学基金资助项目(61894556)。

摘　　要：变电站外围恒温网络节点关联较弱,易受外界入侵。为此,提出基于深度信念网络(DBN)的箱式变电站外围恒温网络入侵动态检测方法。首先,对恒温网络数据实施类型转换和归一化处理,将Nominal类型转换为Numeric类型,以避免转换数据维度变高。然后,利用非线性迭代偏最小二乘法对恒温网络数据实施特征提取和分类处理,以增强入侵信号的聚类。最后,结合DBN模型并利用核极限学习机(KELM)分类算法,对恒温网络的入侵数据实施检测。试验结果表明,该方法的模型训练准确率最高为90%;当迭代次数达到200次时,训练时间为10.35 s;网络入侵检测效率高且检测结果误差分数小,平均误差分数为1%。该方法对各样本的网络入侵检测率稳定,有利于网络入侵检测的实际应用。Substation peripheral thermostatic network nodes are weakly associated and vulnerable to external intrusion.For this reason,a dynamic detection method of peripheral thermostatic network intrusion in box-type substation based on deep belief network(DBN)is proposed.Firstly,type conversion and normalization are implemented on the thermostatic network data,and the Nominal type is converted to Numeric type to avoid the converted data dimension from increasing.Then,non-linear iterative partial least squares method is utilized to implement feature extraction and classification processing on thermostatic network data to enhance the clustering of intrusion signals.Finally,combining the DBN model and utilizing the kernel extreme learning machine(KELM)classification algorithm,the intrusion data of the thermostatic network is implemented for detection.The experimental results show that the model training accuracy of the method is up to 90%;the training time is 10.35 s when the number of iterations reaches 200;the network intrusion detection efficiency is high and the error score of the detection results is small,with an average error score of 1%.The method has a stable network intrusion detection rate for each sample,which is conducive to the practical application of network intrusion detection.

关 键 词：箱式变电站 恒温网络 网络入侵 深度信念网络 数据类型转换 数据维度 

分 类 号：TH70[机械工程—仪器科学与技术]