Freight EVCS Optimization

The problem of locating and allocating charging equipment (PLACE) is essential to be solved to provide a charging network into POLARIS. The PLACE is solved using an optimization model that is inspired by the stochastic location modeling literature. In this stream of the literature, Berman and Krass (2019) developed a model called stochastic location model with congestion and immobile servers, balanced objective (SLCIS-BO). Their model is applicable to siting problems abiding by several assumptions, e.g., demand follows a Poisson process and supply follows an exponential distribution. The model to solve the PLACE aims to minimize the strategic charger station and charger costs, as well as the operational-level travel to charging station, charging time, and queue waiting costs. To solve the problem, we first obtain potential locations where the charging demand arises and compute the frequency of charging demand for each location. Then, we gather potential charging station location, and this is dependent on the application area. Currently, we implement this framework for e-commerce delivery vehicles and business to business trips. In the former, we use traffic analysis zones and e-commerce depots of four agencies: Amazon, FedEx, UPS, and USPS as the candidate charging stations. In the latter, we obtain the candidate stations from various publicly available sources. Once the candidate stations and demand are gathered, we follow a clustering approach for both. For instance, in one of the e-commerce delivery applications of the Chicago metropolitan region 450,000 households were clustered into 200 groups, and nearly 2,000 candidate stations are clustered into 50 groups. Solving the problem at the cluster level, we obtain the number and type of chargers, as well as the clusters to house these chargers. Then, for each selected cluster, we obtain the households assigned to this cluster and solve a subproblem by using actual POLARIS locations as candidate stations. In this phase, we distribute the number and type of chargers to multiple locations while limiting the numbers to the ones obtained in the cluster level. Eventually, we obtain the final charging network to be embedded into POLARIS.

Important

Please refer to the the following paper for more detail:

Berman, O., Krass, D., 2019. Stochastic location models with congestion, in: Laporte, G., Nickel, S., Saldanha da Gama, F. (Eds.), Location Science. 2nd ed.. Springer, Switzerland. chapter 17, pp. 477–531. doi:10.1007/978-3-319-13111-5_17.