A node-charge, graph-based online carshare rebalancing policy with capacitated electric charging

As the constitution of car-share fleets bespeak the planetary displacement to electrical vehicles (EV) operators volition request to code unsocial challenges to EV fleet scheduling. These see idiosyncratic clip and region requirements, clip needed to recharge vehicles, and organisation of charging facilities—including constricted availability of accelerated charging infrastructure (as of 2019 determination are 7 accelerated DC nationalist charging stations successful Manhattan including Tesla stations). Because of specified factors, the viability of electrical car-sharing operations depends connected fleet rebalancing algorithms.

The stakes are precocious due to the fact that potential customers whitethorn extremity up waiting oregon accessing a farther location, oregon adjacent balk from utilizing the work altogether if determination is nary disposable vehicle wrong a tenable proximity (which whitethorn impact important access, e.g. taking a subway from downtown Manhattan to midtown to prime up a car) oregon nary parking oregon instrumentality determination disposable adjacent the destination.

In a caller study, published successful the diary Transportation Science, the authors contiguous an algorithmic method based connected graph mentation that allows electrical mobility services similar carshares to trim operating expenses, successful portion due to the fact that the algorithm operates successful existent time, and anticipates aboriginal costs, which could marque it easier for fleets to power to EV operations successful the future.

The communal signifier for carshare scheduling is for users to publication circumstantial clip slots and reserve a conveyance from a circumstantial location. The instrumentality determination is required to beryllium the aforesaid for "two-way" systems but is relaxed for "one-way" systems. Examples of free-floating systems were the BMW ReachNow car sharing strategy successful Brooklyn (until 2018) and Car2Go successful New York City. These 2 systems precocious merged to go ShareNow, which is nary longer successful the North American market.

Rebalancing involves having either the strategy unit oregon users (through incentives) periodically driblet disconnected vehicles astatine locations that would amended lucifer proviso to demand. While determination is an abundant lit connected methods to grip carshare rebalancing, probe connected rebalancing EVs to optimize entree to charging stations is limited: determination is simply a deficiency of models formulated for one-way EV carsharing rebalancing that captures each the following: 1) the stochastic dynamic quality of rebalancing with stochastic demand; 2) incorporating users' entree outgo to vehicles; and 3) capacities astatine EV charging stations.

The researchers connection an innovative rebalancing argumentation based connected outgo relation approximation (CFA) that uses a caller graph operation that allows the 3 challenges to beryllium addressed. The team's rebalancing argumentation uses outgo relation approximation successful which the outgo relation is modeled arsenic a relocation occupation connected a node-charge graph structure.

The researchers validated the algorithm successful a lawsuit survey of electrical carshare successful Brooklyn, New York, with request information shared from BMW ReachNow operations successful September 2017 (262 conveyance fleet, 231 pickups per day, 303 postulation investigation zones) and charging presumption determination information (18 charging stations with 4 larboard capacities). The projected non-myopic rebalancing heuristic reduces the outgo summation compared to myopic rebalancing by 38%. Other managerial insights are further discussed.

The researchers reported that their formulation allowed them to explicitly see a customer's charging request illustration and optimize rebalancing operations of idle vehicles accordingly successful an online system. They besides reported that their attack solved the relocation occupation successful 15%–89% of the computational clip of commercialized solvers, with lone 7–35% optimality gaps successful a azygous rebalancing determination clip period.

The study's authors accidental aboriginal probe directions see dynamic request (function of time, terms and different factors), data-driven (machine learning) algorithms for updating, much realistic/ commercialized simulation situation utilizing information from larger operations, and elaborate cost-benefit investigation connected the tradeoffs of EV's and regular vehicles.

---

---

More information: Theodoros P. Pantelidis et al, A Node-Charge Graph-Based Online Carshare Rebalancing Policy with Capacitated Electric Charging, Transportation Science (2021). DOI: 10.1287/trsc.2021.1058