Research on a LSTM Based Method of Forecasting Primary Frequency Modulation of Grid

Abstract

The increasing scale of power grid and the integration of a large number of new energy sources, the prediction accuracy of FM frequency is limited. In order to quickly and accurately predict the frequency change curve of power system after disturbance. In this paper, a prediction method of primary frequency modulation based on long short-term memory network (LSTM) is proposed. Firstly, this method uses correlation analysis method to analyze various factors affecting frequency fluctuation, and selects strong correlation quantity; then constructs the power grid frequency prediction model based on LSTM recurrent neural network; finally, uses the historical data of a power grid company from 2016 to 2019 as a simulation example to verify the validity of the model. This method can realize the real-time analysis of power grid frequency and provide decision support for power grid to make frequency control strategy after disturbance.