Data-Driven Trajectory Prediction of Grid Power Frequency Based on Neural Models

Frequency in power systems is a real-time information that shows the balance between generation and demand. Good system frequency observation vital for security protection. This paper analyses response following disturbances proposes data-driven approach predicting it by using machine learning techniques like Nonlinear Auto-regressive (NAR) Neural Networks (NN) Long Short Term Memory (LSTM) networks from simulated measured Phasor Measurement Unit (PMU) data. The proposed method uses horizon-window reconstructs input time-series data order to predict features such as Nadir. Simulated scenarios are based on gradual inertia reduction including non-synchronous into Nordic 32 test system, whereas PMU collected taken different locations Power System (NPS). Several horizon-windows experimented observe an adequate margin of prediction. Scenarios considering noisy signals also evaluated provide robustness index predictability. Results show proper performance level prediction Root Mean Squared Error (RMSE) index.