Direct current series fault arc recognition based on adaptive noise and sparrow optimization

Affected by new energy transformation and the continuous growth of electricity demand, direct current series fault arc is the main factor causing electrical fires, which has attracted widespread attention on how to accurately identify it. However, current methods for identifying direct current series fault arcs often suffer from issues such as low identification quality. To optimize the quality of direct current series fault arc recognition, a recognition method combining extreme learning machine and sparrow search algorithm is proposed. By combining a set of empirical mode decomposition, permutation entropy based on adaptive white noise, reconstruction component algorithm, and wavelet threshold denoising, the characteristics of DC series fault arcs are extracted. Finally, the recognition method is applied to identify direct current series fault arc characteristics. SVM algorithm, BP algorithm and CNN algorithm are used as the benchmark algorithms the performance comparison analysis showed that the average running time and average recognition accuracy of this method were 96.8% and 0.89 s, respectively, which was significantly better than the comparison algorithm. Moreover, the application effect of the direct current series fault arc recognition model was analyzed. The model was superior to comparison models in terms of visual classification effect and recognition precision. Therefore, the direct current series fault arc recognition model has good performance and practical value, which can help improve the quality of direct current series fault arc recognition. The novelty of this study lies in the combination of Sparrow search algorithm and extreme learning machine algorithm, and the introduction of adaptive white noise complete set empirical mode decomposition algorithm, permutation entropy reconstruction component algorithm and wavelet denoising method. The combination of these methods not only improves the accuracy and stability of feature extraction, but also provides new ideas and methods for research in related fields.