A Short-Term Load Forecasting Model Based on Crisscross Grey Wolf Optimizer and Dual-Stage Attention Mechanism

Accurate short-term load forecasting is of great significance to the safe and stable operation power systems development market. Most existing studies apply deep learning models make predictions considering only one feature or temporal relationship in time series. Therefore, obtain an accurate reliable prediction result, a hybrid model combining dual-stage attention mechanism (DA), crisscross grey wolf optimizer (CS-GWO) bidirectional gated recurrent unit (BiGRU) proposed this paper. DA introduced on input side improve sensitivity key features information at points simultaneously. CS-GWO formed by horizontal vertical crossover operators, enhance global search ability diversity population GWO. Meanwhile, BiGRU optimized accelerate convergence model. Finally, collected dataset, four evaluation metrics parametric non-parametric testing manners are used evaluate CS-GWO-DA-BiGRU The experimental results show that RMSE, MAE SMAPE reduced respectively 3.86%, 1.37% 0.30% those second-best performing CSO-DA-BiGRU model, which demonstrates can better fit data achieve results.