Study On The Feature And Detection Method Of DC Arc Fault In The PV Systems

With the PV power generation systems widely used, the Europe, United States and other places suffered fire incidents happened in the PV system one after another, causing economic loss. Later, the accident investigation showed dc arc fault was responsible for most electrical fire in the PV system. With the PV power plant put into operation many years, electronic component aging, cable rupture, contacting point loosing or animal bites and other reasons could cause arc fault. Besides, most installation of the PV array used a long list of high voltage dc power supply, which increased the safety problems related to the arc. The current fuse and circuit breaker could only be used for protecting the system from over current and short circuit fault. It didn’t work for the arc fault. In order to guarantee the safe and reliable operation of the photovoltaic system, this article studies a kind of detection method aiming at this kind of dc arc fault.Such arc detection has not been studied further at home and abroad and it’s characteristics haven’t been known well. The arc itself is random and complex. So it’s hard to establish a mathematical model for arc. We must explore arc characteristics through experiments. Thus, arc generator was made and PV system arc fault test platform was also designed base on the PV power. A series of tests, such as under different working point of the photovoltaic panels, gap length of the electrodes, arc fault occurring at various locations etc, were conducted in order to study the arc fault’s behaviors and influence of different factors to the arc fault by recording arc voltage, current and voltage on the dc side. Considering the palaces where the arc may happen are uncertain. It’s not possible to detect the arc by monitoring voltage changing of arc. By contrast, no matter which branch circuit occurs arc fault, arc current characteristics can be reflected in the dc distributor roads current. Therefore, this article focused on the analysis of the arc current, both from the time domain and frequency domain, finding the special features in favour of arc identification. The method of using multiple criteria can reduce misjudgment rate. Finally, a kind of algorithm was carried out with help of BP neural network, which has the self-learning ability. Analysis and experimental results showed that this method was workable and superior to threshold method in arc detection.