Wavelet-based protection strategy for series arc faults interfered by multicomponent noise signals in grid-connected photovoltaic systems. (1st May 2019)

• Record Type: Journal Article
• Authors: Chen, Silei; Li, Xingwen; Meng, Yu; Xie, Zhimin
• Is Part Of: Solar energy. Volume 183(2019)
• Page Start: 327
• Page End: 336
• Publication Date: 2019-05-01
• Subjects: Grid-connected PV system -- Series arc fault -- Noise interferences -- Feature improvement -- Effectiveness sensitivity analysis -- Protection strategy; Solar energy -- Periodicals; Solar engines -- Periodicals; 621.47
• DOI: 10.1016/j.solener.2019.03.008
• Languages: English
• ISSNs: 0038-092X

Abstract

Graphical abstract: Highlights: In grid-connected PV systems, the synthetic current signal is seriously affected by extra multicomponent noises. Due to the noise signal synthesis influence, the existing Db9-based features would become invalid. The established Rbio3.1-based features directly focus on the arc fault instead of synthetic signal. The proposed protection strategy could adapt to different PV systems even without adjusting parameters. The proposed protection strategy could achieve the effective arc fault detection. Abstract: The arc fault protection strategy should be applied in grid-connected photovoltaic (PV) systems to guarantee the human and equipment safety. In this paper, PV series arc faults are conducted in different grid-connected PV systems through the designed experimental platform firstly. Having analyzed these synthetic arc fault current signals through the existing Db9-based features used in the resistive system, the amplitude is found no longer increasing during the arc fault period in some frequency bands. Inverters would inject extra multicomponent noises to influence electrical characterization of arc faults in grid-connected PV systems, causing that Db9-based features show undesirable constant even decreasing amplitude patterns. Next, the Rbio3.1-based features have the better symmetric property to achieve the singularity detection of arc faults. The biorthogonality property of Rbio3.1 could further improve the reconstruction accuracy cooperating with the symmetric property. Besides, its higher vanishing moment is able to display the arc salient disturbances more clearly. Then the Rbio3.1 is able to give more arc fault occurrence time discovery pulses and better arc fault separation degree for different switching frequencies, inverter types, and inverter stages. Ideally, its feature improvement ability may not appear under fault-like conditions. Finally, the random forest (RF) outputs the final state recognition result through combining multiple classifications without adjusting parameters. For the desired Rbio3.1-based features, different feature combinations are able to significantly improve the detection accuracy at the acceptable speed for complicated interference scenes in multiple grid-connected PV systems based on the MATLAB platform.