Demand-Aware Charger Planning for Electric Vehicle Sharing

Deployment Optimization for Shared e-Mobility Systems With Multi-Agent Deep Neural Search

“…In addition, our approach does not need to work with a predefined budget, which is often difficult to estimate in practice, but just finds the self-sustaining deployment plans that can cover the cost. On the other hand, the work in [1] also studies EV charger planning for private EV sharing platforms, which shares the similar problem with our work. However, it assumes the deployment of chargers is an one-off task, and does not consider dynamic deployment cases as studied in this paper.…”

“…The simulation environment for taxi in [13] assumes the taxis need to move randomly on the road network, while in [12] the potential users have to wait until they are picked up by an arriving taxi, both of which are often not true in practice. In the context of simulating facility/infrastructure deployment, most of the existing work only considers onetime [1], or multiple-stages [2] of infrastructure deployment or expansion. Our simulation environment, on the other hand, simulates the system operation and the actual interactions between users and the system with much finer details, and supports dynamic/continuous infrastructure deployment, i.e., stations can be opened or shut down at arbitrary times.…”

“…The reason why we adopt this two-phase approach rather than sampling independently from the distributions of origins and destinations, is that in practice the origins and destinations of user demand are highly correlated, and the correlations often depend on different times. In addition, unlike existing work [1] which directly uses historical data as demand, we use GPs to i) fuse different sources of information (taxi and the e-mobility data), and ii) generate distributions of user demand rather than scalar values for robustness. As shown later in Sec.…”

“…Specifically, this particular problem falls into a broader class of optimization tasks in the context of e-mobility systems, which have attracted attentions from various communities. For instance, the work in [1] addresses a similar problem of charger planning for electric vehicle (EV) sharing services. It proposes a demandaware planning approach and uses heuristics to approximate the optimal plan which ensures both pervasive coverage (in terms of reaching as many POIs as possible) and satisfies sufficient user charging demand.…”

Optimizing the Locations and Sizes of Solar Assisted Electric Vehicle Charging Stations in an Urban Area

“…To meet different charging requirements, Bai et al [19] proposed a hybrid evolutionary algorithm that combines the non-dominated sorting genetic algorithm-II (NSGA-II) with linear programming and neighborhood search to solve the location, size, and type of charging station. Du et al [20] showed that the location and sizing problem is NP-hard, and they designed a charger-based greedy solution with theoretical guarantees to maximize the satisfied charging demand. Li et al [21] analyzed the real EV users' charging behavior data, and proposed a Bayesian-inference-based algorithm to solve the mixed-integer programming EVCS problem with a flexible objective function.…”