Photovoltaic modules may experience direct current arcing due to loose joints

aging of devices

insulation rupture

and other reasons. However

direct current arcing

without zero crossing characteristics

is difficult to extinguish on its own and can easily cause fires

posing serious hazards. Therefore

it is crucial to quickly and accurately detect series arcing. we studied series arc from the perspective of time domain. The method of short-term local peak rate is used to analyze the current signal. The average value of local peak rate under fixed time window conditions and the improvement ratio of local peak rate before and after short-term conditions are used as time domain features. The model is trained in the random forest algorithm

and finally a DC series arc detection algorithm is constructed based on the random forest model. We established a DC series arc experimental platform with various operating conditions and line impedances

designed multiple circuits to simulate actual operating conditions

and verified the algorithm detection results by collecting series arc waveform data through the experimental platform. The verification results show that the algorithm can effectively identify series arcs and has the characteristics of clear physical meaning

low computational complexity

and fast running time.