基于实时数据流特征提取的设备能耗异常识别算法研究  

Research on Equipment Energy Consumption Anomaly Recognition Algorithm Based on Real-time Data Stream Feature Extraction

作　　者：杨婧 辛明勇 宋强 YANG Jing;XIN Ming-yong;SONG Qiang(Guizhou Power Grid Co.,Ltd.,Guiyang 550002 China;Electric Power Research Institute of Guizhou Power Grid Co.,Ltd.,Guiyang 550002 China)

机构地区：[1]贵州电网有限责任公司,贵州贵阳550002 [2]贵州电网有限责任公司电力科学研究院,贵州贵阳550002

出　　处：《自动化技术与应用》2024年第3期74-77,共4页Techniques of Automation and Applications

摘　　要：电力设备局部放电能耗异常,会产生连锁反应导致出现重大故障情况,威胁设备运行安全,为此研究一种基于实时数据流特征提取的设备能耗异常识别算法。采集设备实时状态,利用双层可视窗口对多组设备数据流聚类分析,设备状态间满足马尔科夫链性关系,通过模型映射节点关系,计算转移概率提取放电特征,凭借核函数分解设备产生局部放电异常时的实际放电量、视在功率、放电均方率及平均放电电流,结合不同电磁波信号强弱的时间差,识别出能耗异常点位置。仿真实验证明所提算法在加压情况下也能精准识别异常点位置,适用性强、鲁棒性好。The abnormal partial discharge energy consumption of power equipment causes a chain reaction to cause major failures and threaten the safety of equipment operation.Therefore,an abnormality identification algorithm based on real-time data stream feature extraction is studied.It collects the real-time status of the equipment,and uses the double-layer visual window to cluster and analyzes the data streams of multiple groups of equipment.The equipment states satisfy the Markov chain relationship.Through the model mapping node relationship,the transition probability is calculated to extract the discharge characteristics,and the equipment is decomposed by the kernel function.The actual discharge amount,apparent power,electric mean square rate and average discharge current when abnormal partial discharge occurs,combined with the time difference of different electromagnetic wave signal strengths,it identifies the location of abnormal energy consumption.Simulation experiments show that the proposed algorithm can accurately identify the location of abnormal points even under pressure,and has strong applicability and robustness.

关 键 词：数据流挖掘 特征提取 设备能耗异常 局部放电 马尔科夫链 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TM591[自动化与计算机技术—计算机科学与技术]