An Arc Fault Detection Method Based on Current Amplitude Spectrum and Sparse Representation

When a series arc fault occurs, the current value of circuit is often less than the threshold of the circuit breaker. But the temperature of the arc combustion can be as high as thousands of degrees, which can lead to electrical fire. The current data of the normal work and arc fault are collected by using arc fault experiment. The arc fault is detected based on the current amplitude spectrum and the sparse representation algorithm. In traditional sparse representation algorithm based on ${L}_{p}$ norm, the regular order p selects the fixed value, and ${p}$ is usually 1 or 1/2. According to the previous studies on the characteristics of ${L}1$ , ${L}3$ /4, ${L}1$ /2, and ${L}1$ /4 norm, it is found that the accuracy and sparsity of classification can be improved by adopting different norms for different data. Then, the online adjustment method of regular order ${p}$ has been proposed. The regular order ${p}$ is adjusted by calculating the minimum value of residual, which can make ${p}$ be the optimal value for any test data.