Electric load forecasting holds a pivotal role in reaching energy conservation,emission reductions,and global carbon neutrality.The urgency of accurate forecasting is escalating in light of intensifying global climate...Electric load forecasting holds a pivotal role in reaching energy conservation,emission reductions,and global carbon neutrality.The urgency of accurate forecasting is escalating in light of intensifying global climate change,acting as a linchpin for optimizing urban energy systems,minimizing energy consumption,and achieving low-carbon development.Addressing the prevalent challenges,especially the inability of current methods to effectively unearth latent load volume information resulting in diminished predictive accuracy,has become a focal point of contemporary research.This paper aims to tackle these issues by introducing a novel method that deconstructs electric load into seasonal and trend components,each forecasted through distinct models.Notably,for the seasonal components,a method incorporating both local and global information is utilized,and an innovative Expand intra-layerConvolution is introduced,facilitating effective forecasting through the use of residual blocks.When benchmarked against existing methodologies,this model demonstrates better performance in key metrics such as MAE and MSE.展开更多
文摘Electric load forecasting holds a pivotal role in reaching energy conservation,emission reductions,and global carbon neutrality.The urgency of accurate forecasting is escalating in light of intensifying global climate change,acting as a linchpin for optimizing urban energy systems,minimizing energy consumption,and achieving low-carbon development.Addressing the prevalent challenges,especially the inability of current methods to effectively unearth latent load volume information resulting in diminished predictive accuracy,has become a focal point of contemporary research.This paper aims to tackle these issues by introducing a novel method that deconstructs electric load into seasonal and trend components,each forecasted through distinct models.Notably,for the seasonal components,a method incorporating both local and global information is utilized,and an innovative Expand intra-layerConvolution is introduced,facilitating effective forecasting through the use of residual blocks.When benchmarked against existing methodologies,this model demonstrates better performance in key metrics such as MAE and MSE.
