光伏电站串联型直流电弧的检测算法与保护实验研究

该文提出结合电弧频域特征和多路信号频谱的相关系数的直流电弧检测算法。通过信号调理和奇异值分解对采集的电流信号降噪,采用线性规划优化参数的分段频谱二次加权的方法提取了电弧频域特征T和多路信号频谱的相关系数R,以T和R作为输入量设计了基于逻辑回归的直流电弧检测算法。对2700组历史数据进行电弧检测,误判率0.3%,漏判率0.29%。同时,研制电弧发生器与在线监测设备,在光伏电站进行串联型直流电弧模拟试验。结果表明,检测系统能准确识别电弧故障,迅速动作熄灭电弧,并向后台发送报警信息,响应时间约为60 ms。

EXPERIMENTAL STUDY ON DETECTION ALGORITHM AND PROTECTION OF DC SERIES ARC IN PHOTOVOLTAIC STATION

This paper presents DC series arc detection algorithm based on the frequency domain characteristics of the arc and the correlation coefficient of multi-channel signal spectrum.Signal conditioning and singular value decomposition are used to reduce the noise of the acquired current signal.Then the method of linear programming is used to optimize the parameters of the segmented spectrum to extract the arc frequency domain feature T and the correlation coefficient R of the multi-channel signal spectrum.Both T and R are used as inputs of the arc detection algorithm based on logistic regression.When this detection algorithm is applied to distinguish DC arc faults from 2700 groups of historical data acquiring in field,the error rate and miss rate are as low as 0.3%and 0.29%respectively.A DC arc on-line detection and protection system was developed and experiments were done in the PV station.The results show that the monitoring system can accurately identify series arc faults,extinguish the arc quickly and send alarm information to the background.The response time is about 60 ms.