Frequency Assessment Method Based on Integrating SFR Model

This paper proposes a frequency assessment method that integrates physical model and data knowledge to solve the problem of frequency response calculation after power system disturbances. Identifying key parameters in the SFR model based on LSTM network to predict frequency response quickly and accurately. The sensitivity of various parameters affecting frequency response was derived. The frequency response data after disturbance was used as the input feature for machine learning, and the key parameters of the SFR model were used as the output. The data-driven method was used to modify the key coefficients of the physical model, improving the adaptability and accuracy of the SFR model. Finally, the proposed prediction method for the frequency stability is verified by simulations on IEEE 39-bus system. It is demonstrated that the proposed method can improve the accuracy of frequency response prediction by comparing it with time-domain simulation results.