Wide time-frequency machine vision for DC series arc fault detection during photovoltaic inverter startup
Year: 2026
Type: article
Abstract: Accurate detection of series arc faults (SAFs) is crucial for ensuring the safe operation of photovoltaic (PV) systems and remains a notable research challenge. During PV inverter startup, the persistently rising current with a wide dynamic range often exhibits nonlinear hybrid interference comprising both transient and slow-changing components, which frequently causes nuisance tripping of arc fault detectors. Despite this known impact, research on SAF detection during PV inverter startup remains limited. To address this gap, this study analysed an RLC model and the two-stage characteristics of electrical arcing, ultimately establishing a historical spectral projection (HSP) method for SAF detection using wide time-frequency domain (WTFD) features and machine vision techniques. The proposed method converts multiple spectral sets into coloured WTFD images, applies the Otsu algorithm for binarisation, extracts wideband edges using the Prewitt differential operator, performs positive time dilation to obtain the HSP of the two-stage arcing current and adopts fixed thresholds to detect SAFs. The method was validated through simulations and experiments on an actual PV system, achieving a high accuracy of 98.93%, precision of 99.39% and detection time of 16–55 ms. These findings demonstrate that the proposed method can effectively detect SAFs while preventing nuisance tripping during PV inverter startup and in other scenarios.
Source: Solar Energy
Institution: Tianjin University of Science and Technology
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $3,900
Awards: G1088192680, G8886223600
OpenAlex ID: W7202365259