Transient Stability Assessment Using Deep Transfer Learning

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