A Cross-Area Knowledge Learning Method for Status Monitoring of Transportation-Oriented Energy Interconnected System
Xuguang Hu, Dazhong Ma, Guanjun Meng, Rui Wang, Huaguang Zhang · IEEE Transactions on Instrumentation and Measurement · 2022
In the status monitoring study of transportation-oriented energy interconnected system, the labor-intensive nature of collecting enough labeled samples tremendously limits the actual applications of deep learning-based methods. To address it, a cross-area knowledge learning method with domain-invariant information of system status is proposed in this paper. First, considering the spatial-temporal correlation change of transportation-oriented energy interconnected system, convolutional neural network with global feature enhancement, as feature extraction module, is proposed to extract and separate features of different system status. Second, to improve the learning ability of the proposed model on the condition of unlabeled samples, sample classify module including two classifiers is proposed to make feature extraction module capture the intrinsic similarity features between different systems. Third, for improving the classification accuracy of status monitoring, virtual adversarial training item is added in the loss function of the proposed method to reduce the disturbance influence. The proposed method is evaluated by three different datasets, and ten typical methods are selected for comparison. The comprehensive results demonstrate effectiveness and superiority of the proposed method for status monitoring of transportation-oriented energy interconnected system with unlabeled samples.