基于二阶聚类和鲁棒性随机分割森林算法的低压台区线损异常辨识

Line Loss Anomaly Identification of Low-voltage-station Based on Second-order Clustering and Robust Random Cut Forest Algorithm

为精准识别台区的线损异常,保证配电网经济、稳定运行,针对台区线损的异常情况,提出一种基于二阶聚类和鲁棒性随机分割森林(robust random cut forest,RRCF)算法的台区线损异常检测方法。首先,运用二阶聚类将台区不同的运行工况进行聚类,将相同工况的线损节点归并,然后将各类工况的节点线损数据导入RRCF算法中分析,通过删除和插入样本节点,并对插入节点后评判模型的复杂度进行计算,得到线损异常节点的评分值,进一步找出线损异常的节点。最终,通过有关实例验证所提方法的准确性与有效性。

To accurately identify the abnormal line loss in the substation area and ensure the economic and stable operation of the distribution network,in allusion to the abnormal line loss in the substation area,based on second-order clustering and robust random cut forest(abbr.RRCF)algorithm a method to detect the abnormal line loss in the substation area was proposed.Firstly,by means of second order clustering the different operating conditions of the substation area were clustered and the line loss nodes under the same operating conditions were merged.Secondly,the nodal line loss data of all kinds of operating conditions was led into RRCF algorithm to conduct the analysis.By means of deleting and inserting sample nodes and computing the complexity of the evaluation model after inserting nodes,the score values of abnormal line loss nodes could be obtained,and further the nodes with abnormal line loss could be found out.Finally,the effectiveness and accuracy of the proposed method are verified by related examples.