基于小波特征提取和筛选的爬坡事件分类  

Classification of Wind Power Ramps Based on Screened Wavelet Energy Characteristics

作　　者：欧阳庭辉 查晓明[1] 秦亮[1] 熊一[1] 夏添[1] 黄鹤鸣[1] 

机构地区：[1]武汉大学电气工程学院,湖北省武汉市430072

出　　处：《中国电机工程学报》2016年第9期2373-2380,共8页Proceedings of the CSEE

基　　金：国家重点基础研究发展计划项目(973计划)(2012CB215101)~~

摘　　要：为了便于对不同爬坡事件采取有针对性的控制调度策略,该文提出采用小波变换挖掘爬坡事件内部特征,并筛选出有效的能量特征用于指导爬坡分类。首先分析小波变换的原理和多分辨率特性,给出小波能量特征的提取方法。其次,结合类内稳定和类间差异的目标,以离散度矩阵为参考,使用判别分析法筛选出有效特征。最后文中以美国BPA区域2013年的风电数据为实例,按照风过程的类别生成3类爬坡样本,并根据文中所提方法训练得到爬坡分类器,然后对实例数据进行爬坡分类和验证。结果表明,小波能量特征对于区分不同类别爬坡事件具有可行性,同时筛选的小波特征对于进一步改善爬坡分类的性能具有重要意义。In order to adopt different and effective control strategies to reduce the harm of different ramp events, this paper put forward a method using wavelet transform to mining the internal characteristics of ramps, and screened some effective characteristics for ramp classification. Firstly, it introduced the principle and multi-resolution feature of wavelet transform, and proposed the methods of extracting wavelet energy feature. Then, taking the stability inside a class and differences between categories as the target and discrete degree matrix as indicator, it used the discriminant analysis method to screen effective features. Finally, taking the data of BPA in 2013 as an example, this paper generated three kinds of ramps according to wind process and trained the ramp classifier, then finished the classification of validation data. The results of performance evaluation on ramp classification show that wavelet features are valid to classify different ramp events well, and that the screened features based on the proposed method in this paper are useful to improve the performance of wind power ramp classification.

关 键 词：爬坡事件分类 小波变换 多分辨率特性 特征筛选 

分 类 号：TM614[电气工程—电力系统及自动化]