Abstract:

It proposes a method to identify Power Quality Disturbances（PQD） based on wavelet energy difference distribution and Support Vector Machine（SVM）.It uses wavelet transform to analyze PQD signals，extracts disturbance duration time and energy differences of every level between PQD signal and standard signal as feature vectors，forms the training samples and testing samples.It preprocesses the training set by using neighborhood rough set model to delete those abnormal samples and disturbances.It traines the PQD samples by using Binary Tree SVM（BT-SVM） to identify PQD signals.Testing results indicate that the proposed method can identify seven PQD signals and sinusoidal signal，has an excellent performance on correct ratio（the average ratio can reach 97 percent），has high identification speed and strong resistance to noise，and is very suitable for PQD identification system.

Key words: power quality, disturbance identification, wavelet energy difference, support vector machine, neighborhood rough set

摘要： 提出了小波能量差分布和支持向量机（Support Vector Machine，SVM）相结合的电能质量扰动（Power Quality Disturbance，PQD）识别方法。该方法用小波变换对PQD信号进行分析，提取信号各层暂态能量与标准信号的能量之差和扰动持续时间为特征向量，组成训练样本和测试样本;使用基于邻域粗糙集模型对训练样本集进行预处理，剔除噪声和异常样本;使用具有二元树结构的SVM对PQD样本进行训练，实现PQD的识别。测试结果表明，该方法可以实现 7种PQD的识别，准确率高（平均可达97%），抗噪声能力强，辨识速度快，适用于PQD识别系统。

关键词: 电能质量, 扰动识别, 小波能量差, 支持向量机, 邻域粗糙集