Deep Learning-Based Models for Predicting Poorly Damped Low-Frequency Modes of Oscillations

This work proposes a real-time deep learning-based model for predicting the small-signal stability of an electrical network. The trained models equip power system operators with accurate and fast monitoring tool which can be used during online operation. To achieve this objective, three different architectures are employed in research; stacked long short-term memory (LSTM), convolutional neural network (CNN)-LSTM Convectional LSTM (Conv-LSTM). These using datasets contain oscillatory parameters (frequency damping ratio) both local inter-area modes oscillations. In addition, voltage phasors at buses taken as input where output comprises patterns modes. Furthermore, overall performance proposed is shown New-England 10-machine, 39-bus, IEEE 16-machine, 68-bus, 5-area, 50-machine, 145-bus benchmark test cases. main findings show that training CNN-LSTM Conv-LSTM provide better compared stacked-LSTM model. former have less number thus shorter time. CNN_LSTM prone to overfitting problems ability capturing spatial temporal features inherent data.