Time Series Prediction of Power System Events Using LSTM, GRU, and Transformer Networks to Increase Computation Efficiency
Naga Lakshmi Thotakura, Zhihao Jiang, Yilu Liu · 2024
The growing complexity of interconnected power systems increases dynamic simulation time, creating computational challenges. To improve efficiency, researchers explore model reduction strategies, focusing on a study area while approximating external regions. Measurement-based techniques to represent these areas through dynamic loads. Recurrent Neural Networks (RNNs), such as long short-term memory (LSTM), gated recurrent unit (GRU), and Transformers, excel at predicting non-linear power system behavior. In this study, we apply RNN models to the Northeast Power Coordinating Council (NPCC) 140-bus system, with two equivalent loads representing external areas. We forecast time-series power system events, including generator and load trips. The results confirm the RNN models' accuracy in predicting power system behavior during events, showcasing their potential for event prediction.