Research on DC arc fault detection in PV systems based on adjacent multi-segment spectral similarity and adaptive threshold model

Abstract

This paper proposes a practical adaptive detection method of series DC arc that could be more adaptable to PV systems’ intricate and complex environment than conventional detecting methods. Firstly, The discharge in a short air gap with unsymmetrical dielectric-covered electrodes will bring additional noise interferences to the current of the circuit where the arc is located and give rise to the alternation of the spectrum of the noise current. By analyzing the current noise spectrum before and after the occurrence of a DC arc fault, it was found that the adjacent multi-segment spectral similarity (AMSSS) characteristic could identify the event of the mark. Secondly, an algorithm architecture using principal component analysis(PCA) of AMSSS and integrated φ-statistic was used to determine an adaptive threshold model. Finally, to validate the fidelity and resilience of the pattern, a photovoltaic plant platform containing 20 PV modules and one three-phase inverter was constructed to acquire relevant data. Then, validation of the adaptive thresholding model was implemented by including several sets of tests generated under different conditions, such as normal, shading, inverter start-up, crosstalk, and arcing conditions. According to the comparative tests, We tentatively conclude that the adaptive model has a relatively strong arc detection capability and high environmental adaptability and could be applied to natural PV systems.