An Enhanced iTransformer-Based Model for Electricity Load Forecasting

With the rapid rise of new energy technologies and electric vehicles, the volatility and uncertainty of electricity load have increasingly intensified, making electricity load forecasting more important than ever. However, existing load forecasting models often face certain limitations when dealing with complex time series data. To overcome this bottleneck, this paper proposes a novel electricity load forecasting model called FTFformer. This model effectively combines a meticulously designed Frequency Feature Extract Block (FFEB) with the encoder of iTransformer, enabling the model to excellently capture both frequency and temporal features of the time series, thereby improving prediction accuracy. Experimental results demonstrate that FTFformer performs exceptionally well on actual electricity load datasets, significantly surpassing the predictive accuracy of mainstream existing models. This achievement not only substantially enhances the accuracy and reliability of electricity load forecasting but also provides strong support for the efficiency and stability of power systems.