This study proposes a deep transfer learning method using a deep convolutional neural network pre-trained with ImageNet for transient stability assessment. The procedure of deep transfer learning, incorporating the role and considerations of the transient stability assessment system, is suggested. The transient assessment system learns the relationship between the severity of disturbances and transient stability in the power system through the proposed training method. The severity of a disturbance based on the physical causal relationship of the angle stability is proposed to train and implement the transient stability assessment model. The power system state variables obtained from the phasor measurement unit are converted into the feature map described by the severity of a disturbance, enabling the training of transient stability characteristics of the power system to the deep convolutional neural network. The training dataset is constructed using the time-domain simulation on IEEE 39 and IEEE 118-bus benchmark power system models configured in MATLAB/Simulink. As a result of deep transfer learning, which involves training with freezing some of the convolutional layers of the pre-trained deep network, the most suitable model for transient stability assessment is selected among the pre-trained deep networks. The effectiveness of the proposed method is compared with other approaches using the confusion matrix, and the robustness against noise interference is also investigated.

Accurate classification between magnetizing inrush currents and internal faults is essential for reliable transformer protection and stable power system operation. Because their transient waveforms are so similar, conventional differential protection and harmonic restraint techniques often fail under dynamic conditions. This study presents a two-stage classification model that combines a self-organizing map (SOM) and a convolutional neural network (CNN) to enhance robustness and accuracy in distinguishing between inrush currents and internal faults in power transformers. In the first stage, an unsupervised SOM identifies topologically structured event clusters without the need for labeled data or predefined thresholds. Seven features are extracted from differential current signals to form fixed-length input vectors. These vectors are projected onto a two-dimensional SOM grid to capture inrush and fault distributions. In the second stage, the SOM’s activation maps are converted to grayscale images and classified by a CNN, thereby merging the interpretability of clustering with the performance of deep learning. Simulation data from a 154 kV MATLAB/Simulink transformer model includes inrush, internal fault, and overlapping events. Results show that after one cycle following fault inception, the proposed method improves accuracy (AC), precision (PR), recall (RC), and F1-score (F1s) by up to 3% compared with a conventional CNN model, demonstrating its suitability for real-time transformer protection.