基于改进DBSCAN算法的变压器不良漏抗参数辨识

发布时间：2019-02-16 11:29

【摘要】：PSD-BPA在中国电力系统仿真计算中被广泛应用,但由于其数据格式的特殊性,往往容易出现许多人为原因的数据错误,这给仿真计算结果的准确性与可靠性带来了极大的隐患。首先,在给出变压器不良漏抗参数辨识步骤的基础上,结合PSD-BPA潮流数据中变压器参数数据的特点,提出了考虑特征相似度的具有噪声的基于密度的聚类(DBSCAN)改进算法。其次,基于各类参数向量簇的各属性最大相似系数,计算获得各类参数向量簇的典型特征向量。然后,基于各类的典型特征向量,针对聚类结果中的噪声簇,提出了基于离群系数的可疑不良数据分布模型;在此基础上,结合分布规律,提出了基于可疑度的不良参数判别方法。最后,通过实际算例验证了所述模型与方法的有效性。
[Abstract]:PSD-BPA is widely used in power system simulation calculation in China, but because of the particularity of its data format, it is prone to many man-made data errors, which brings great hidden trouble to the accuracy and reliability of simulation results. Firstly, based on the identification steps of transformer bad leakage reactance parameters and considering the characteristics of transformer parameter data in PSD-BPA power flow data, an improved (DBSCAN) clustering algorithm with noise based on density is proposed, which takes into account the similarity of features. Secondly, based on the maximum similarity coefficient of each attribute of all kinds of parameter vector clusters, the typical characteristic vectors of various parameter vector clusters are obtained. Then, based on the typical feature vectors, the distribution model of suspicious bad data based on outlier coefficients is proposed for the noise clusters in the clustering results. On this basis, combined with the distribution law, a method of identifying bad parameters based on the degree of doubt is proposed. Finally, the effectiveness of the proposed model and method is verified by a practical example.
【作者单位】： 华北电力大学电气与电子工程学院;国网江苏省电力公司;国网北京经济技术研究院输电网规划中心;
【分类号】：TM41
，

本文编号：2424401