Research on Short-term Load Forecasting Based on GWO-BILSTM

Abstract In short-term power load forecasting, changes in external multi-dimensional factors will have a certain impact on the accuracy of forecasting. response to this problem, paper proposes prediction model GWO-BILSTM that combines Grey Wolf Optimizer (GWO) and bi-directional long memory (BILSTM) network. Taking real data as set, high correlation parameters are selected input through Pearson analysis, hyperparameters BILSTM optimized by GWO algorithm, finally is established based predict set. The experimental results show mean absolute percentage error, root square error index better than other comparison models, which effectively improves data.