基于KLDA-INFLO的继电保护整定数据异常识别方法

A detection method for anomalies in protection relay setting based on the KLDA-INFLO

当前电力系统存在规模不断扩大、功率输入来源不断增多以及用电需求不断上升等现状,电网中出现电力运行扰动的频率不断增加,对继电保护稳定性提出了更高的要求。为实现对继电保护系统在运行过程中潜在扰动的及时应对,构建运行数据异常检测方法实施预警和分析。首先,采用基于核函数的线性判别分析(KLDA)模型,实现原始数据的降维处理从而达到降低运算负担、加快响应时间的效果;其次,结合基于被动式异常因子检测(INFLO)模型,依据运行整定参数正常数值范围,能够及时发掘异常节点,从而对异常运行状况做出快速反应;最后,以某地区配电网继保设施监测数据为例进行仿真分析,结果表明:该方法具有较高的异常检测性能,能够实现针对安全风险的自动校核与管控。

Nowadays,the scales of power systems are enlarging,the types of input power sources are increasing,and the energy demands are also raising.Hence,the disturbance in grids become more frequent,which request a more reliable protection relay system.To achieve the timely response for the potential disturbances in protection relay systems,this paper establishes anomaly detection method for warning and analyzing such disturbances.Firstly,the Kernel Linear Discriminant Analysis(KLDA)model is utilized to reduce the dimensionality of input data,thus to decrease the computation burden and accelerate the response.Then,the Influenced Outlierness(INFLO)anomaly detection is designed.This model can find the outliers in time according to the common range of operation setting parameters,and thus to swiftly response to anomaly conditions.Finally,an empirical study which is based on the protection relay system in one operating distribution network is conducted.The results show that the performance of the proposed method is satisfying,and can be deployed to monitor or manage the countermeasures for potential risks.