IEEE Transactions on Industrial Informatics · 2024 · 22 citations · 20 references
EngineeringElectrical EngineeringPower ElectronicsSemiconductor DevicesTime Series AnalysisDiagnosticsElectrical System DiagnosticsFault DetectionPhotovoltaic DevicesPower QualitySolar Photovoltaic SystemsArc Fault DetectionDiscrete Wavelet TransformElectrical Arc DetectionInverter NoiseInverter SystemsPhotovoltaic Fault DiagnosisThreshold Level SettingWavelet Transform
A series arc fault detector (AFD) is a significant device for preventing fire hazards in photovoltaic (PV) systems. The AFD should detect a series arc quickly and accurately. However, system noise due to the components of a PV system can cause false detection of the AFD. Furthermore, as the inverter types vary according to PV systems and the irradiation changes during one day, it is difficult to develop a universal arc fault detection algorithm with an adequate arc detection criterion, and these difficulties limit the commercialization of AFDs. This study proposes an arc detection algorithm based on differential discrete wavelet transform (DWT) analysis and variable threshold method. Differential DWT analysis increases the distinction performance of inverter noise and arc noise and detects series arcs quickly and effectively by acquiring the amplified arc noise and attenuated inverter noise. Moreover, the variable threshold method updates the adequate threshold level for arc detection in real time and does not require manual tuning for systems and irradiation change. This study also proposes an AFD based on the TMS320F28335 digital signal processor to analyze noise and detect the series arc fault in real time using the proposed algorithm. The performance of the series AFD was verified via a series arc-detection test under the UL1699B test facility. Furthermore, the reliability of the arc detection algorithm and AFD was verified via the unwanted tripping test and arc detection under a real PV system.
20
Communication in the Presence of Noise
Claude E. Shannon · Proceedings of the IRE · 1949 · 6.1K citations
Series DC Arc Fault Detection Based on Ensemble Machine Learning
Vu Le, Xiu Yao, Chad Miller et al. · IEEE Transactions on Power Electronics · 2020 · 149 citations