Study Of Low Voltage Series Arc Fault Diagnosis Technology

With the development of the electric power industry, the issue of the electrical safety has become a broadly debated topic and has attracted many attentions of the experts and the society. Among all the factors, electrical fire is the greatest potential threat that can result in personal injury and property losses. In order to prevent electrical fire effectively, reduce the accident rate and losses of life and property caused by fire has become a hot topic. Some faults, such as short circuit and overload, which cause fire easily, can be detected and treated rapidly and effectively by the conventional circuit breakers or the switches. Because the current of the series arc fault is usually lower than the action threshold of traditional protection devices and its dispersion, the series arc fault has become the main cause of the electrical fire. Therefore, it has become one of the key research problems at home and abroad.Based on the further research of mechanism of the arc and the conventional arc generating device, an integrated arc fault experimental platform consists of arc generating device, data collection unit and system control unit has been designed and built. And the platform is controlled by the microcontroller and the stepping motor, which realize the simulation of the arc fault in the circuit. During the test, the moving electrode is stable. In addition, its precise linear motion not only makes the whole process of arc generation and development in a small gap inμm level and even the arc blowout in a large gap, but also makes the whole arc burning process completely controllable. In order to obtain the judgment and identification of similarity between the series arc fault and the load used in common life, a great number of experiments have been done in two ways that one is the arc happened in single load circuit and the other is the arc happened in one branch of the parallel circuit with two loads. These experiments are based on the device and used eight typical loads as research objects. During the experiments, current and voltage data in both normal condition and arc fault condition are caught successfully and a foundation for later arc fault diagnosis and analysis has been made. By using the wavelet packet transform, frequency bands subdivision of experimentdata and energy statistics under different frequency band were accomplished, and thecurrent energy distributions for different loads under normal and arc fault conditions havebeen obtained and figured out. Additionally, the proposed energy criterion was applied asthe input and the relax-model wavelet neural network (WNN) was trained to identify thearc fault exactly and fast. The proposed arc detection network was verified for thedemonstrated cases. The conception of arc fault level graded by the intensity of arc burninghas been presented in the research of single resistive load. And the three-level arc has beendetected by WNN exactly, which provide data to the study of intelligent arc fault circuitinterrupter.