Hyperopia Optimization of Electric Vehicle Charging and Discharging Considering Discontinuous Parking Patterns

Electric vehicles (EVs) have been widely recognized as demand-side controllable loads, exhibiting substantial flexibility potential for demand response. However, challenges arise when controlling EVs to deliver ancillary services to the grid, primarily due to the uncertain temporal and energy consumption requirements imposed by the stochastic and discontinuous EV parking patterns. To address this issue, this paper proposes a deep reinforcement learning (DRL)-based optimization method for EV charging and discharging control, taking into consideration of the inherent fluences of flexibility dispatchment between current and adjacent parking epochs. Firstly, the paper proposes a stochastic parking model through simulating daily parking sequences with time and energy requirements. Secondly, a flexibility feasible region model is formulated to quantitatively describe the time availability and energy demand constraints for each parking epochs. Finally, a DRL-based framework is established for hyperopia optimization of EV flexibility dispatchment under real-time pricing mechanisms. The results demonstrate that the proposed approach not only fine-tunes EV scheduling within individual parking epochs, but also enhances the demand response performance in adjacent parking epoch by proactively deploying charging and discharging behaviors.