Plug-in electric vehicle load modeling for charging scheduling strategies in microgrids


SUSTAINABLE ENERGY GRIDS & NETWORKS, vol.32, 2022 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 32
  • Publication Date: 2022
  • Doi Number: 10.1016/j.segan.2022.100819
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Compendex, INSPEC
  • Keywords: Plug-in electric vehicles, Load modeling, Kernel Density Estimation, Smart charging, EV charging scheduling, NETWORKS
  • Middle East Technical University Affiliated: Yes


Utilization of plug-in electric vehicle (PEV) load models can improve the performance of smart charging strategies, which increase the reliability of the grid by harnessing the flexibility of PEV loads. This paper presents a method for utilizing personal PEV load models in real-time stochastic charging control with single and finite system-time horizons. First, the drivers' load models are found with Kernel Density Estimation (KDE). Second, a single system-time horizon coordinated charging control algorithm is devised to ensure each PEV is charged at least a critical amount given a feasible set of optimization constraints. The coordinated charging algorithm tackles the NP-hardness of single-deadline charging scheduling problems efficiently with a sorting algorithm utilizing the stochastic PEV load models. Third, we extend the single system-time horizon coordinated charging control algorithm to a scheduling algorithm considering a finite system-time horizon. This approach utilizes the stochastic PEV load models in a model predictive control based approach to decrease the complexity of stochastic online charging scheduling problem into a deterministic case. The scheduling algorithm makes assumptions about the future arrivals to the charging station, unlike the classical online EV charging scheduling algorithms, which optimize the load demand revealed at the current time but underestimate the load demand revealed in the future. Our findings suggest the individual load models complement smart charging algorithms' decision process by improving the fairness of charging time allocation and extending the degree of knowledge of future random data for the scheduling algorithm.