DC fault arc detection method based on adaptive EMD and deep learning

In view of the low accuracy and poor real-time performance of DC fault arc detection in DC power systems in the field of photovoltaic and transformer power, which is difficult to adapt to complex scenarios, especially for the problem of misjudgment and missed judgment in micro-arc, a DC fault arc detection method based on adaptive EMD and deep learning was proposed. According to the characteristics of DC fault arcs, the empirical mode decomposition is used to distribute the feature information in each mode, the feature information in the original current signal is increased. The mode is adaptively selected to extract the largest component of fault information contained in each layer of modal, and the mode with the largest information component is used to extract the features and bring it into the neural network for judgment, so as to further improve the accuracy of DC fault arc detection.