摘要
针对传统电能质量分级预警的等级区间划分及阈值设置僵硬,难以反映和体现不同预警对象对电能质量的敏感程度和差异化诉求的问题,提出一种考虑数据聚类特性的电能质量分级预警阈值设置方法。首先,在系统分析了国家电能质量标准及传统电能质量等级划分的基础上,引入聚类方法对各预警对象进行聚类划分,将同类预警对象的各电能质量指标值的算术平均值作为该类对象的个性化预警限值,以深度挖掘、体现和反映各类预警对象对电能质量的个性化要求;然后,依据各类预警对象的个性化预警限值与国标限值及传统电能质量各等级区间的关系,分6种情况重新设置"因情而异"的电能质量等级区间及阈值,有效避免因各级阈值设置不当造成漏报警或频繁报警现象;最后,以某供电局20个10 kV电压等级的公共连接点的电能质量分级预警为例,验证了本文方法的灵活可调性和广泛适用性。
As for such issue as inflexible graded interval classification and threshold setting of traditional power quality and difficulty in reflecting the appeal of different early warning objects to the sensitivity and difference of power quality, a kind of graded early warning threshold setting method of power quality considering the data clustering features is proposed. Firstly,the clustering method is introduced to make clustering classification of each early warning object and to take arithmetic mean value of each power quality index of the same type of early warning object as individual early limit value of the object on the basis of analyzing systematically China national power quality standard and grade classification of traditional power quality, so to exploit and reflect deeply the individual requirement of each early warning object to the power quality. Then, power quality grade interval and threshold value is set as 6 cases based on different situation in accordance with the relations between individual early warning value and international limit of each type of early warning object and each grade interval of traditional power quality, which avoids effectively such circumstance as missing and frequent alarm due to inappropriate setting of each threshold value;Finally,the power quality grade early warning of 20 pieces of 10 kV common connection points for certain power supply bureau is taken as an example to verify the flexible adjustability and wide applicability of the method.
出处
《电力电容器与无功补偿》
北大核心
2016年第3期88-94,共7页
Power Capacitor & Reactive Power Compensation
基金
国家自然科学基金(51377060)
国家自然科学基金青年基金(61104181)
关键词
电能质量
分级预警
预警阈值
模糊聚类
power quality
grade early warning
threshold
fuzzy clustering.
