基于时间序列关联分析的稳定运行规则提取方法  

A Time Series Associative Classification Method for the Operation Rule Extracting of Transient Stability

作　　者：于之虹[1] 黄彦浩[1] 鲁广明[1] 史东宇[1] 周孝信[1] 严剑峰[1] 卜广全[1] 田芳[1] 李亚楼[1] 

机构地区：[1]中国电力科学研究院,北京市海淀区100192

出　　处：《中国电机工程学报》2015年第3期519-526,共8页Proceedings of the CSEE

基　　金：国家863高技术基金项目(2011AA05A119);国家电网公司大电网重大专项资助项目(SGCC-MPLG001-2012)~~

摘　　要：提出一种考虑运行方式变化时序的稳定规则提取方法。利用主成分分析进行敏感特征提取。对传统方法在数据标准化中存在的问题,通过数据均值化处理,保留原数据中各变量变异程度的差异信息。Considering the time-varying characteristics of power system, an operation rule extracting method was proposed. Firstly, features describing the system state and network configuration before/after fault happening of each area appeared in their abstract forms. Then, a novel principal component analysis method was proposed to extract the stability sensitive features and reduce the input dimension. With mean normalization, the variance difference of each feature can be preserved. Meanwhile, in order to adapt the operation changes in time and improve the computation efficiency, the recursive principal component analysis（PCA） formulas were derived. Its main process is to make use of the old model to calculate the new one according to the rolling data. Based on the ensemble associative classification method, some associative classification rules reflect the relationship between power system operating state and transient stability were generated. Additively, time dimension was introduced, and mainly used for those successive relationships. The proposed scheme was illustrated on the IEEE 39-bus test system. The results show that this method can be used to reveal the actual pace of operation change and reflect the dynamic changes. Rules generated by the proposed method can provide more insightful information, which is valuable for practical operation analysis and decision.

关 键 词：大数据 数据挖掘 安全评估 运行规则 时间序列关联分类分析 电力系统 

分 类 号：TM71[电气工程—电力系统及自动化]