Source-linked AI summary
Electric Power Allocation in a Network of Fast Charging Stations
I. Safak Bayram, George Michailidis, Michael Devetsikiotis, Fabrizio Granelli
TL;DR
The paper addresses how to operate a network of fast-charging stations that provides EV-driver QoS without imposing excessive strain on the power grid. It models stations with local storage and evaluates increasingly coordinated power-allocation and customer-routing schemes. The framework serves more customers with the same grid power, can reduce power requirements for a QoS target through routing, and considers communication effects as a realism issue.
Problem
Fast-charging networks must provide EV drivers with QoS while managing grid strain and the mobility of charging loads.
Method
The paper develops a stochastic fast-charging-station model with local storage and evaluates resource allocation across three cases, including power allocation and customer rerouting.
Results
10% power savings can provide ε=0.05 QoS when customer routing partially affects customer choices.
Takeaways & Limitations
Customer routing enables blocking-probability targets to be achieved with less grid power, while the framework compares network operation and profitability across allocation schemes.
Abstract
from arXiv · showhide
In order to increase the penetration of electric vehicles, a network of fast charging stations that can provide drivers with a certain level of quality of service (QoS) is needed. However, given the strain that such a network can exert on the power grid, and the mobility of loads represented by electric vehicles, operating it efficiently is a challenging problem. In this paper, we examine a network of charging stations equipped with an energy storage device and propose a scheme that allocates power to them from the grid, as well as routes customers. We examine three scenarios, gradually increasing their complexity. In the first one, all stations have identical charging capabilities and energy storage devices, draw constant power from the grid and no routing decisions of customers are considered. It represents the current state of affairs and serves as a baseline for evaluating the performance of the proposed scheme. In the second scenario, power to the stations is allocated in an optimal manner from the grid and in addition a certain percentage of customers can be routed to nearby stations. In the final scenario, optimal allocation of both power from the grid and customers to stations is considered. The three scenarios are evaluated using real traffic traces corresponding to weekday rush hour from a large metropolitan area in the US. The results indicate that the proposed scheme offers substantial improvements of performance compared to the current mode of operation; namely, more customers can be served with the same amount of power, thus enabling the station operators to increase their profitability. Further, the scheme provides guarantees to customers in terms of the probability of being blocked by the closest charging station. Overall, the paper addresses key issues related to the efficient operation of a network of charging stations.
I. INTRODUCTION
The paper frames fast-charging infrastructure as necessary for metropolitan EV use while seeking operating regimes that limit grid strain and preserve driver QoS. It proposes a stochastic station model, local storage, and increasingly coordinated power-allocation and customer-routing schemes.
- Motivation: 40% of surveyed California EV drivers travel daily beyond their fully charged battery range, creating daytime recharging needs that fast-charging networks can address.The need is especially pronounced in metropolitan areas and among residents without convenient nighttime charging.
- Motivation: Large-scale EV charging can strain the power grid and potentially contribute to grid instability, while station-level energy storage can reduce that impact.The impact depends on EV penetration, charging requirements, location, and time of day.
- Research objective: The study targets operating regimes that minimize grid strain while maintaining good quality of service for EV drivers.This is presented as the central operational challenge for charging-station networks.
- Approach: The proposed architecture models station operation stochastically, measures performance as the percentage of served customers, and uses local storage to smooth stochastic demand.The architecture concerns EV fast DC charging stations.
- Evaluation design: The resource-allocation framework is evaluated through three increasingly complex cases: no allocation, power allocation, and power allocation combined with customer rerouting.The scenarios are motivated by non-uniform vehicle-trip distributions in actual Seattle-area traffic traces.
- Coordination: A two-way communications protocol coordinates EV assignments and reroutes customers when necessary to support customer allocation.The paper identifies rerouting as an essential element for increasing the number of charging EVs without increasing grid power.
- Related work: Prior scheduling studies largely assume stationary vehicles at customer premises or large parking lots, whereas this study addresses spatially distributed charging demand.The paper uses centralized decision making for a subset of EVs.
- Related work: Related charging-station designs use storage and examine storage technologies with different efficiencies and power ratings, which this work also considers in its station architecture.The paper notes that storage suitability can depend on station-specific constraints such as space and cost.
III. CHARGING STATION ARCHITECTURE
The proposed charging-station architecture combines constant grid power, local energy storage, multiple charging classes, and blocking probability as its QoS metric. A finite-state stochastic model supports power partitioning across customer classes and evaluates how storage characteristics affect blocking and service capacity.
- The architecture supports fast and slow charging requests to accommodate different customer needs and vehicle constraints.
- Blocking probability is the QoS metric because the bufferless model assumes arriving customers do not wait for service.Customers are blocked when concurrent demand exceeds the combined station and storage capacity.
- The station draws constant grid power, uses local storage for excess demand, and recharges storage with otherwise idle grid power.This design smooths stochastic demand and insulates the grid from peak charging demand.
- A. Stochastic Model for Station Dynamics: A finite-state continuous-time Markov chain models station dynamics and yields steady-state probabilities from its transition-rate matrix.The model uses a two-dimensional finite state space, with Q containing transition rates and π containing steady-state probabilities.
- A. Stochastic Model for Station Dynamics: The operator partitions grid power among customer classes by solving an optimization problem based on their service-rate demands and population composition.The model explicitly represents multiple customer classes and class proportions.
- A. Stochastic Model for Station Dynamics: 95%-efficient fast storage outperforms 85%-efficient slow storage in blocking probability, while optimized power assignments serve more customers when the fast-charging share is larger.For class compositions of (75%,25%), (50%,50%), and (25%,75%), the corresponding optimal allocations are,, and.
B. Profit Model
The profit model connects charging-state probabilities with revenues, blocking penalties, and storage costs to evaluate station profitability across customer classes and arrival rates.
- The model relates the stochastic charging-station model to cost parameters and guides grid-power choices for different arrival rates.
- Revenue varies by customer class, while blocked EVs incur penalties that represent dissatisfaction, reputation effects, and QoS control.
- Net profit is computed by classifying Markov-chain states into grid-charging, storage-charging, and blocking states.
- For low arrival rates, acquisition and installation costs outweigh charging revenue, producing negative profit.
- At high arrival rates, blocking costs dominate and reduce net profit; higher proportions of fast-charging customers increase profit because they lower blocking probability.
A. Overview
The network model considers spatially uneven urban demand and progressively expands from fixed, identical stations to optimized grid-power allocation and customer rerouting.
- Overview: Urban traffic varies spatially and temporally, making demand density important for charging-station utilization and power allocation.
- Case I: No Allocation: Case I uses identical stations with no allocation, so customers drive to the nearest station and arrival rates differ with traffic density.
- Case II: Optimal grid power-S allocation: Case II allocates grid power across stations while considering QoS targets, with subcases for selfish and partially reroutable drivers.
- Case III: Optimal S and λ allocation: Case III jointly optimizes station capacity and customer assignment within small geographic areas where inter-station driving costs are negligible.
- Network constraints: Additional distribution-network constraints can be incorporated by limiting maximum power allocation at individual stations.
A. Overview
The evaluation uses Seattle-area bus movements to model weekday rush-hour demand, fit its spatial distribution, and estimate traffic intensity at deployed charging stations.
- Traffic traces: The study uses publicly available Seattle bus movements as a proxy for urban traffic during weekday rush hours.The analyzed periods are 7am–9am and 5pm–7pm.
- Spatial distribution: The fitted weekday-rush-hour spatial distribution is piecewise beta, with 0.6% mean squared error.
- Spatial distribution: The analysis reports an x–y coordinate correlation coefficient of 0.06.
- Station deployment: Eight charging stations are deployed at locations adopted from the reference base-station layout, and simulation is used to calculate station traffic intensity.
2) Charging Station Placement:
The placement and evaluation setup uses a discrete-event simulation and station-level traffic intensities to expose demand imbalance and approximate blocking probabilities.
- Charging Station Placement: Table I reports the traffic intensity of each station.
- Charging Station Placement: At overall arrival rate λ = 50, stations three and four have blocking probabilities of 0.58, while the other six stations range from 0.014 to 0.053.
- Charging Station Placement: Response Surface Methodology approximates blocking probability B as a second-order polynomial function of storage size, power rating, service rate, and arrival rate.
1) Metamodeling of Blocking Probabilities:
The paper models station blocking probability with a regression metamodel and uses sensitivity analysis to identify how inputs affect QoS. The model reports strong fit statistics and indicates that grid power has the greatest effect on reducing blocking.
- Network implications: Some stations exhibit very high blocking probability while overprovisioned stations exhibit very low blocking probability.
- Regression metamodel: The regression model expresses blocking probability as a function of storage capacity, station capacity, arrival rate, and grid power after logit transformation.The model uses the response variable logit(B) and then applies the inverse-logit transformation to obtain blocking probabilities.
- Regression metamodel: 88.06% R-Square and 0.52% mean square root error quantify the regression model’s fit.
- Sensitivity analysis: Grid power S has the highest impact on decreasing blocking probability among the modeled inputs.The sensitivity analysis considers S, R, ν, and λ through the Jacobian and Hessian quantities.
- Sensitivity analysis: High arrival rates leave little spare capacity, causing the local storage device to be frequently empty.
C. Comparison of three cases
The paper compares progressively more complex allocation schemes, from no allocation to power allocation and customer routing. The proposed schemes use fixed total grid resources and storage types while evaluating QoS and financial performance across charging stations.
- Compared scenarios: The comparison covers an identical-station baseline, power allocation for selfish or mixed populations, and power allocation for EV fleets.
- Power allocation: Minimum grid power is calculated to meet station-level QoS targets, provided the required power does not exceed total available generation capacity.The optimization problems are solved with standard interior point methods; nonlinear integer programs are relaxed and rounded.
- Power allocation: Power resource allocation enables more vehicles to receive service with the same amount of grid power.
- Customer routing: Customer routing can produce 10% power savings while providing ε=0.05 QoS.The comparison concerns selfish EVs and a mixed population of selfish EVs and fleets.
- Evaluation setup: The evaluation compares baseline and allocation schemes for stations 2, 3, and 4 at arrival rates λ=20, 25, and 30.The comparison fixes total grid resources and uses the same type of energy storage devices.
- Evaluation outcomes: The proposed framework significantly improves system QoS and financial performance in the reported comparison.Average net profit per charging station is used to assess financial performance.
VI. TOWARDS A MORE REALISTIC MODEL: THE ROLE OF COMMUNICATIONS AND INCENTIVES
The proposed network requires communications infrastructure to support timely information dissemination, but realistic deployment must account for communication failures and driver noncompliance with routing decisions.
- Communications infrastructure: The network architecture allocates power and reroutes customers, relying on communications infrastructure and protocols for timely information dissemination.The framework combines optimal power allocation with customer routing.
- Communication effects: Communication delays and losses between EVs, stations, and the coordinator may affect routing decisions and QoS.The paper identifies quantifying these operational effects as important.
- Connectivity: Temporary network disconnections can cause vehicles to miss information, making communication-network connectivity an operational concern.Information dissemination is framed in terms of evaluating the communication network’s degree of connectivity.
- Driver compliance: The framework assumes EVs follow coordinator-issued routes, although drivers may instead choose the nearest station.Pricing incentives are proposed as a possible way to address deviations from assigned routes.