Photovoltaic arc fault detection method based on transformer voltage signal

Because DC electric arc signal has not zero-crossing point, arc is not easy to extinguish naturally. In this paper, aiming at the arc faults of photovoltaic system, a series arc fault detection method based on transformer voltage signal is proposed. Firstly, an arc generation test platform is built to obtain the voltage signals at both ends of the current transformer timely. After removing the interference or noise from the original signals by the wavelet threshold shrinkage denoising method, wavelet analysis and singular value decomposition(SVD) are carried out to compare the difference between the normal and arc signals. Wavelet analysis is used to obtain the modulus maximum value of the wavelet coefficients of each node and the node power spectrum. At the same time, Toeplitz matrix is constructed to decompose SVD to obtain the range of eigenvectors. Then the training set and test set data are selected for BP neural network training, and the output results are normalized. Finally, the arc fault detection is realized by threshold method. The test results show that the accuracy of the detection algorithm is more than 99% in both laboratory environment and photovoltaic field. © 2021, Solar Energy Periodical Office Co., Ltd. All right reserved.