Low-Voltage DC-Arc-Current Model by Time-Series Interval Prediction and Equivalent Circuit Model

The arc current is essential for arc fault detection in power generation systems such as residential direct current (DC) microgrids and rooftop photovoltaic (PV) systems. However, the DC arc is complex and it is difficult to develop an arc model to estimate the arc current precisely. In this paper, a hybrid DC arc-current model based on the time-series interval prediction and equivalent circuit model is developed. The relationship between the arc current and the electrode separation speed, arc gap length, voltage, and current levels is investigated. It is found that the interval prediction method can estimate the dynamic characteristics of arc current. The variation range of the arc current can be determined by the time-series interval prediction through the confidence interval. The static characteristics of arc current can be demonstrated by the equivalent circuit of an arc fault. Thus, a hybrid arc-current model can be developed by the time-series interval prediction and equivalent circuit model. The experimental results verify that the proposed hybrid arc-current model can effectively reflect the actual time-domain behavior of arc current with a similarity higher than 94.95%. The proposed model can predict the arc current accurately for arc fault detection in residential DC microgrids and rooftop PV systems.