A method for detecting DC fault arcs using a complete ensemble empirical mode decomposition

This paper proposes a method for quickly identifying DC fault arcs in DC power systems. The method focuses on decomposing DC fault arc signals using complete ensemble empirical mode decomposition (CEEMD) technology. By effectively utilizing this algorithm for signal decomposition, the data processing speed is increased, making it easier to obtain frequency band information from the original fault signals. By reducing noise interference, solving mode mixing problems, and improving decomposition efficiency, the method enhances fault identification efficiency and accuracy to a certain extent. Subsequently, the paper constructs feature vectors using permutation entropy and employs a BP neural network fault identification model for fault recognition, significantly improving the accuracy of DC fault arc discrimination.