A DC arc test platform was established to address DC series arc faults in photovoltaic systems.Using Hall current sensors to obtain a large amount of fault current data,four characteristic quantities that can reflect the relationship between current and arc faults were extracted from the time-frequency domain of the data;a method for calculating the variation of feature quantities was proposed,effectively avoiding the problem of traditional fixed threshold comparison methods being difficult to apply to complex and diverse working conditions;by using machine learning logistic regression methods,the detection parameters for arc faults were obtained.Through extensive data validation,it was found that there were almost no arc false alarms.The proposed algorithm can be applied to embedded devices and deployed in photovoltaic systems for arc detection.
关键词
直流串联电弧故障/逻辑回归/霍尔电流传感器/光伏系统/快速傅里叶变换
Key words
DC series arc fault/logistic regression/Hall current sensor/photovoltaic system/fast fourier transform
维普
万方数据