DSpace About DSpace Software

Institutional Repository at Tsinghua University >
本科生优秀毕业论文 >
本科生优秀毕业论文(2012) >

Please use this identifier to cite or link to this item: http://hdl.handle.net/123456789/2433

Title: 湖北电网自动化基础数据智能检测的研究
Other Titles: Research on Hubei Power System Automation Data Intelligent Detection
Authors: 杨韵
Keywords: 异常数据
intelligent detection
knowledge discovery
Issue Date: 20120611
Abstract: 电力系统中有海量数据,其中也包含了一些异常数据。异常数据主要分为两种,一种是噪声异常数据,会恶化数据分析的结果,应剔除;另一种是真实异常数据,可能蕴含着有意义的信息,应利用数据挖掘方法进一步研究和分析。目前很多电力系统业务缺乏数据智能检测方法。 聚类能使相异度大的对象分为不同的簇,从而检测出于其他数据不同的异常数据。自组织特征映射神经网络(SOM)在聚类中有广泛的应用。和其他聚类算法相比,聚类性能较佳。针对电力系统数据高维、非线性、海量的特点,对SOM进行改进,引入加权欧氏距离、核函数并与K-means算法结合。 由于SOM的噪声敏感性大,且聚类后部分异常数据可能会划分到正常数据类别,本文设计了一种基于密度的一维异常数据检测方法,辅助SOM实现噪声异常数据的辨识和定位。此外,为了挖掘电力系统数据中蕴含的知识,本文结合领域知识和支持向量机,实现了数据的知识发现和利用。据此,本文设计了基础数据智能检测模型,实现了异常数据的辨识和知识发现两大功能。通过实例仿真给出了模型的数据智能检测过程。
There are mass amounts of data in power system, including outliers. There are two types of outlier. One is noise outlier, which is useless and may be adverse to data analysis. The other is real outlier which includes significant knowledge and therefore deserves a further analysis using data mining techniques. However, large numbers of power system business now lack intelligent data detection techniques. Clustering can partition objects into clusters such that outlier can be partitioned into a different cluster from normal data. Self-organizing Feature Map (SOM) is a popular approach in clustering. On account of the multi-dimension, non-linearity and massive amount of power system data, SOM is improved by introducing weighted Euclidean distance, kernel function and K-means clustering algorithm. Considering that SOM is sensitive to noise, a one-dimensional outlier density-based detection method (ODDBCAN) is designed. With specific domain knowledge and Support Vector Machine (SVM), the underlying knowledge of data can be further tapped. Accordingly an intelligent data detection model is designed with two basic functions, outlier detection and knowledge discovery.
Description: 26
URI: http://hdl.handle.net/123456789/2433
Appears in Collections:本科生优秀毕业论文(2012)

Files in This Item:

File SizeFormat
022杨韵2008010997.pdf1.96 MBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.


Valid XHTML 1.0! DSpace Software Copyright © 2002-2010  Duraspace - Feedback