Study On Detection Technology For Electrical Fire Of Arc And Development Of Arc Fault Detection Device

In China, electrical fire has become the primary cause and has an intensified trend. At present, the traditional circuit breaker mainly plays a preventive role in preventing both metallic short-circuit and leakage. When arc faults, such as insulation damage and aging, occur, the traditional circuit breaker is not good enough to carry out a comprehensive and effective prevention in electrical fires because the fault current is much lower than action thresholds. Therefore, how to effectively prevent the electrical fire caused by the arc fault is a very important and meaningful topic. In view of the load inhibitory, this paper focused on experimental investigation of arc fault characteristic parameters under different working conditions. Some feasible arc fault detection methods which could improve the accuracy of the arc fault detection was put forward in this paper, and a corresponding detection device was developed.According to the GB14287.4-2014 "electric fire monitoring system Part 4: arc fault detector ", an experiment platform for producing different types of arc faults was set up. Based on the national standard, the collection of current waveform and high frequency coupling signal was carried out and followed by characteristic analysis. Experimental results show that it is difficult to accurately distinguish arc fault only by current waveform or high frequency characteristics and there existing upper misjudgment rate and misdetection rate.In view of the limitations on single criterion, this paper adopted the mathematical statistics method, sought out three characteristics for arc faults, and then made a neural network fusion processing for them. This work greatly improves the recognition rate of arc fault. Moreover, the author put forward three quantitative formulas for microprocessor transplantation convenient.According to the above mentioned recognition algorithms, an arc fault detection device was designed and developed. The hardware design mainly includes the signal conditioning module, power module, micro processingmodule and communication module.Among them, the logarithmic detector amplifier, as the core of the high frequency circuit module, was designed for effectively extracting fault high-frequency arc contour feature;To facilitate the current signal processing on microprocessor, the author designed a single power precision rectifier circuit.As for the software part, the fusion algorithm of BP neural network is relatively complex, which causes it too difficult to finish fusion of eigenvalues in a short time. Hence, a simplified algorithm was used to calculate sum of characteristic values in accordance with the fusion of the weight sum, and then be treated as output parameters.In accordance with the writing ideas of the top-down modeling, modularization,structuration, the author completed the arc fault detection program design.According to the national standard, the performance test of the prototype was carried out, and the results show that the new developed arc fault detection device can effectively detect the arc fault in circuit and can avoid other branch high-frequency crosstalk.