Non-invasive electric arc fault detection based on sliding approximate entropy

Abstract

In recent years, electrical fires have occurred frequently and have become the first cause of various fires. Among them, the electric arc is one of the important causes of electrical fires. In this paper, the research on non-invasive electric arc fault detection is carried out for low-voltage users, which can discover and detect the electric arc only by analysing the aggregated current measurements from outdoors. In order to achieve this goal, firstly, the abrupt change of the current mode is discovered and detected with sliding window from the total load current signal based on the approximate entropy. Then the current signature samples are extracted around the abrupt change. Finally, according to the pre-set electric arc signature classifier, it is judged whether the abrupt change is caused by the electric arc fault. In addition, under laboratory conditions, an electric arc simulation experiment is carried out for common appliances of actual low-voltage users. The results show the effectiveness of the proposed non-invasive electric arc fault detection method based on approximate entropy.