基于分形特征的最小二乘支持向量机局部放电模式识别

摘要

为了提高局部放电模式识别的正确率,提出了一种基于最小二乘支持向量机(LS-SVM)的局部放电模式识别的方法.应用分形理论,同时结合小波包分析技术,计算各个频段信号的分形维数,把各频段局放信号的分形维数的例数输入到多分类最小二乘支持向量机中进行训练,实现对放电样本的分类.结果表明,分形特征浓缩了局部放电信号的信息,有效地解决了模型参数选择耗时巨大的问题.该方法在有限样本情况下能够达到较高的识别率,具有良好的应用价值.%In order to improve the correct rate of partial discharge (PD) pattern recognition, a method based on the least squares support vector machine ( LS-SVM ) is put forward to recognize the discharge models.Using the wavelet analysis technology and the fractal theory, the fractal dimension of signals in each frequency-band can be calculated, and the reciprocal of fractal dimensions of each frequency-band are input to multi-classified LS-SVMs for training to implement PD samples classification.The results show that by adopting fractal characteristics, the PD signal information is concentrated and the time-consuming problem in parameter determination is solved.Moreover, the method enables to detect a high recognition rate under condition of small samples, and has good value in PD pattern recognition.