Answers are grounded in this paper and cite their sources.
Time-of-use (TOU) pricing is a pivotal strategy for optimizing the service quality and revenue of Electric Vehicle Charging Stations (EVCSs), yet its successful implementation relies heavily on accurate period-specific load forecasting. However, existing studies predominantly focus on coarse-grained long-term predictions and often overlook the dynamic spatial interdependencies among EVCSs that vary across different time periods, thereby limiting prediction accuracy. To address this gap, this paper proposes GLOPE, a novel G raph neural network based LO ad P redictor for accurately predicting the time-of-use load for E VCSs. The core innovation of GLOPE lies in capturing the shifting user charging behaviors by constructing dynamic correlation graphs for distinct time periods. Specifically, GLOPE utilizes a graph attention mechanism to aggregate features from neighboring stations based on historical price and load sequences, thereby effectively integrating spatial dependencies with temporal dynamics. These graph features are further combined with station-specific attributes and period-specific to predict the load of specific periods. Extensive experiments on two real-world datasets from a leading EV company demonstrate that GLOPE significantly outperforms state-of-the-art techniques, achieving performance improvements ranging from 15.01% to 44.69%.
No takes yet. Share an insight, caveat, or question.
Zhou et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: