Research on short-term power load forecasting method based on EMD-GRU

The volatility, non-stationarity, and non-linearity of power load data make it tricky for traditional prediction methods to accurately predict it, while precise forecasting of load data can raise the reliability and economy. Therefore, a combination prediction model based on EMD-GRU network is presented in this paper. Firstly, this paper collected the electric load data set of Cyprus for the whole year of 2018, with a sampling interval of 1 hour. Then, the data was decomposed into multiple sub-sequence modal components using the EMD method, and GRU was used to predict each sub-sequence component. Finally, superimposing the forecasting results of each sub-sequence component, the prediction of power load data was completed. The experimental results indicate that the EMD-GRU-based power load prediction model exhibits superior prediction precision and outperforms other neural network algorithms currently employed.