Demand Response Baseline Estimation for Public Fast Charging Stations Based on DSW-Informer

With the rapid growth of electric vehicles (EVs), public fast charging stations (PFCSs) are emerging as vital demand-side resources in urban environments. While PFCSs offer substantial potential for demand response (DR) initiatives, their baseline load estimation is challenged by the heterogeneity of physical charging characteristics and the randomness of user charging behaviors. These factors contribute to the complexity and volatility of overall load patterns in PFCSs, which can result in significant inaccuracies when applying conventional baseline estimation methods, hindering the effective participation of PFCSs in DR programs. In this paper, considering the complicated temporal dependencies between adjacent time points, we design a novel framework for DR baseline estimation of PFCSs. Firstly, we analyze the individual charging behavior patterns for EVs with different models and the aggregated load characteristics within existing PFCSs. Subsequently, we construct a “model pool” that encompasses various charging behaviors and battery performance parameters, and then develop a new temporal charging feature that integrates the charging characteristics of individual EVs at a granular level into the aggregated level of PFCSs. Finally, we propose the Dimension-Segment-Wise -Informer baseline estimation model, which embeds input features in the form of segments and operates these 2D vectors using a sparse attention mechanism. Numerical experimental results based on a real-world dataset demonstrate that our proposed method outperforms other methods in terms of baseline estimation accuracy and robustness under uncertain DR signals.