基于二维离散余弦S变换的电能质量扰动类型识别

Power quality disturbance type identification based on a two-dimensional discrete cosine S-transform

为获得可靠的高质量电能,提高电能质量扰动(Power Quality Distrubances, PQD)类型识别准确率,提出了一种基于二维离散余弦S变换(2D-DCST)的PQD类型识别方法。首先在数学模型的基础上,生成包括7种复合扰动在内的17类不同的电能质量事件。然后将一维的PQD信号转换成行列相等的二维信号,利用2D-DCST方法从二维信号中得到其振幅矩阵,对振幅矩阵提取基于统计、能量和图像的特征。再使用第二代非支配排序遗传算法(NSGA-Ⅱ)将提取的大量特征降维成少量有用的特征组。最后对所选特征使用支持向量机(SVM)分类器,构建一个分类准确率高、特征数目少的类型识别模型。实验结果表明,该方法能够准确高效地识别17类电能质量事件,并且有较好的抗噪性。同时对复合扰动也有较高的识别准确率,为电能质量扰动类型识别问题提供了新的方法。

To obtain reliable high-quality power and improve the accuracy of Power Quality Disturbance(PQD) type identification, a new PQD type based on a Two-Dimensional Discrete Cosine S-Transform(2 D-DCST) is proposed. First, based on mathematical models, 17 kinds of power quality events including 7 kinds of complex disturbances are generated. Then, one-dimensional PQD signals are upgraded into two-dimensional signals with equal rows and columns. An amplitude matrix is obtained from the two-dimensional signals using the 2 D-DCST method. The statistics, energy and image features of the amplitude matrix are first extracted, and then reduced into a small number of useful feature groups using the Non-Dominated Sorting Genetic Algorithm-Ⅱ(NSGA-Ⅱ). Finally, a Support Vector Machine(SVM) classifier is used to construct a type identification model with high identification accuracy and few features. Experimental results show that the method can identify 17 kinds of power quality events accurately and efficiently and also exhibits a good noiseproof feature. At the same time, the method also has a high identification accuracy for complex disturbances. This provides a new method for power quality disturbance type identification.