Real-Time Short-Term Voltage Stability Assessment using Temporal Convolutional Neural Network

Abstract

Voltage stability assessment is crucial to maintain secure operation of the power system under disturbances. A deep learning model, Temporal Convolutional Neural Networks, is proposed in this paper for real-time short-term voltage stability assessment. It covers both the voltage instability and the fault-induced delayed voltage recovery phenomenon. In this work, the time series post-disturbance bus voltage trajectories is taken as input, which is sampled by phasor measurement units within a small observation time window. The model then predicts the stability state of the power system: stable or unstable or stable-with fault-induced delayed voltage recovery phenomenon, in a computationally efficient manner. Furthermore, this study also explores the performance of another deep learning-based time series classification algorithm, the Long Short-Term Memory Neural Network. Finally, the effectiveness of these models are evaluated on IEEE 9- Bus test system. The result shows that both models can predict the short-term voltage stability phenomenon with the accuracy of more than 97% with high computational efficiency and thus are suitable for online application.