基于小波轮廓描述符和动态时间扭曲的电压扰动分类

Identification of disturbance signals based on wavelet boundary descriptors and dynamic time warping

为更好地提高电压扰动信号的识别精度,提出了一种基于小波轮廓描述符和动态时间扭曲的新算法。首先将信号通过小波变换进行滤波处理,然后根据小波轮廓描述符的定义,在Mallat算法的基础上提取各扰动信号的小波轮廓描述符作为参考模板,再通过动态时间扭曲分类器与参考模板进行匹配,选取最优路径获得分类结果。本文提出的识别方法无需先对扰动信号进行多种特征提取再分类,识别过程较为简单;而且解决了对电压闪变和暂态脉冲等扰动不明显或扰动时间较短的信号易引起误判的问题。Matlab仿真分析的结果证明,此方法能够快速有效地识别出扰动信号,准确率高。

A novel algorithm based on the wavelet boundary descriptors and the dynamic time warping was proposed for improving the accuracy of voltage disturbance classification. In this algorithm, firstly, wavelet is used in feature extraction of those disturbances; and then, according to the definition of wavelet boundary descriptors, the wavelet boundary descriptors are constructed on the basis of Mallat algorithm. Secondly, it calculated distance matrix be- tween testing and seven kinds of reference disturbances, then DTW is used to search the optimum path which needs to be the shortest in every distance matrix to guarantee that the testing signals have the most resemblance with refer- ence signals. Finally, it selects the shortest path as classification result. Without extraction of the disturbance sig- nals, this method makes the recognition process simple. Moreover, this method has solved the problem, that caused the misjudgment in the case of obvious and short disturbance signal, like the voltage flicker and impulse transient. A simulation was done in Matlab. The simulation results showed that this method can recognize the dis- turbance signals quickly and effectively.