Short-Term Power Load Forecasting Model Design Based on EMD-PSO-GRU

Aiming at the nonlinear, nonstationary, and time series characteristics of power load, this study proposes a load forecasting method based on empirical mode decomposition particle swarm optimization gated recurrent unit neural network. First, original data are decomposed into limited number modal components residual component by using to reduce nonstationarity complexity sequence decrease association between different IMFs. The subsequences build prediction models network, respectively, use algorithm optimize network-related hyperparameters increase parameter accuracy model; finally, superimpose results each subsequence obtain final value. case show that compared with traditional algorithm, proposed EMD-PSO-GRU model can better dig trend information forecasting, fit curve better, have higher accuracy.