| With the development of modern electric power industry,the electrical fire poses serious potential risks to the issue of the electrical safety.Accordingly,arcing fault is considered to be one of the most probable causes.However,the most frequently used fault protection principles such as over current protection,short circuit protection and residual current protection cannot work well.Besides,the Arc-Fault Circuit-Interrupter(AFCI)cannot fulfill the protection function correctly due to the influence of the large amounts of nonlinear loads.To solve these issues,two arcing fault detection methods are presented based on both time domain and frequency domain characteristics of arc current.Firstly,arcing faults are simulated by the test setup and the database of arc current and voltage are built in order to capture the main features of arcing fault.Secondly,the flat shoulder percentage,the maximum of current variation rate and the mean value are taken as the characteristics of arc current in time domain,while using the similarity value as a synthesis to describe the changes when there is an arcing fault.In addition,the BP neural network is adopted to improve the accuracy of fault diagnosis with the help of self-learning and self-adaptive ability of the BP neural network.Finally,the performance of high-frequency components of arc current with typical load in time domain is quite different from that in frequency domain,based on which,a detection method is proposed for series AC arc fault.The ratio of arc current variation rate to its RMS and the current amplitude of 6 kHz ~12 kHz band are used as the characteristic parameters to identify the series AC arc fault.Since the duration of load startup current is far shorter than that of arc current,the time threshold of arc fault is set accordingly to reduce the influence of load startup process on the arc fault detection.Test results show that the proposed method can easily realize the rapid detection of series AC arc fault with simple hardware. |
PDF Full Text Request