基于改进LightGBM的电力通信数据流量异常检测方法  

Improve the Abnormal Detection Method of Power Communication Data Traffic Based on LightGBM

作　　者：李丹 张子杨 LI Dan;ZHANG Ziyang(State Grid Ezhou Power Supply Company Information and Communication Branch,Ezhou 436000,China)

机构地区：[1]国网鄂州供电公司信息通信分公司,湖北鄂州436000

出　　处：《通信电源技术》2024年第20期152-154,共3页Telecom Power Technology

摘　　要：针对电力通信网络攻击模式多变、检测泛化能力不足的问题,提出基于改进轻量级梯度提升机(Light Gradient Boosting Machine,LightGBM)的电力通信数据流量异常检测方法。结合最近邻规则(Edited Nearest Neighbor,ENN)算法、小波包分解技术和信息熵分析方法,提取电力通信数据流量异常特征,引入直方图算法和带深度限制的Leaf-wise生长策略,通过改进LightGBM算法建立电力通信数据流量异常检测模型,在模型中找到最优超参数配置,提高电力通信数据流量异常检测的准确率和效率。实验结果表明,设计方法在准确识别不同网络攻击类型和增强检测泛化能力方面具有显著优势,能够更好地应对电力通信网络中复杂多变的网络威胁,为电力通信系统的安全稳定运行提供有力保障。Addressing the issue of variable attack modes and insufficient detection generalization ability in power communication networks.the abnormal detection method of power communication data traffic based on improved Light Gradient Boosting Machine(LightGBM)is proposed.Combined with Edited Nearest Neighbor(ENN)algorithm,wavelet packet decomposition technology and information entropy analysis method,extract the power communication data traffic abnormal features,introduce the histogram algorithm and with depth limit Leaf-wise growth strategy,improve the LightGBM algorithm for power communication data traffic anomaly detection model,find the optimal super parameter configuration in the model,improve the accuracy and efficiency of power communication data traffic abnormal detection.The experimental results show that the design method can significantly improve the gray detection rate and improve the detection effect,so as to effectively deal with the changing network threat.

关 键 词：改进轻量级梯度提升机(LightGBM) 电力通信 数据流量 异常检测 超参数优化 

分 类 号：TM73[电气工程—电力系统及自动化]