Enhancing Electric Vehicle Charging Demand Forecasting with Hybrid Markov Chain and Time Series Models

Abstract The rapid adoption of electric vehicles (EVs) necessitates robust and accurate forecasting of charging demand to ensure the reliability and efficiency of charging infrastructure. This paper proposes a novel hybrid framework that integrates Markov Chains and Transition Matrices with advanced time series prediction models to enhance the prediction accuracy of EV charging demand. Specifically, the study leverages LSTM, Seq2Seq-Transformer, and Time2Vec-Transformer models, augmented by Markov Chain-based system modeling, to capture the stochastic nature and temporal dependencies in EV charging patterns. Extensive experiments were conducted on a dataset comprising nine years of charging station usage data from Palo Alto, California. The performance of the proposed models was evaluated against traditional time series models. The results demonstrate significant improvements in prediction accuracy, with the hybrid models outperforming their counterparts regarding prediction accuracy metrics. Additionally, the proposed framework’s ability to model the overall reliability of charging stations was analyzed, revealing insights into the operational efficiency and reliability distribution of the EV charging network.