Power Equipment Fault Diagnosis Model Based on Deep Transfer Learning with Balanced Distribution Adaptation

摘要

In recent years, an increasing popularity of deep learning models has been widely used in the field of electricity. However, in previous studies, it is always assumed that the training data is sufficient, the training and the testing data are taken from the same feature distribution, which limits their performance on the imbalanced tasks. So, in order to tackle the imbalanced data distribution problem, this paper presents a new model of deep transfer network with balanced distribution adaptation, aiming to adaptively balance the importance of the marginal and conditional distribution discrepancies. By conducting comparative experiments, this model is proved to be effective and have achieved a better performance in both classification accuracy and domain adaptation effectiveness.