Source-linked AI summary
Electric Vehicle Scheduling and Carbon-Aware Energy Dispatch for Charging Stations
Fulong Yao, Yiming Xu, Maurizio Albano, Liana Cipcigan, Naeima Hamed, Shangyin Weng, Omer Rana
TL;DR
EV charging stations must coordinate limited charger access with energy supply while considering both cost and carbon emissions. This paper combines SFGH vehicle scheduling with TD3 carbon-aware energy dispatch, and year-long simulations report the lowest service delay, energy cost, and emissions among evaluated configurations.
Problem
Coordinating access to limited chargers with energy supply while accounting for cost and carbon emissions remains a challenge for EV charging stations.
Source: §I · p. 1Method
The paper combines SFGH scheduling based on service-delay reductions with TD3 battery dispatch balancing financial and carbon objectives to meet the resulting demand.
Source: §I · p. 1Results
SFGH–TD3 achieves the lowest mean net cost and net emissions in all test sets, and the lowest service delay among the evaluated configurations.
Source: B. Simulation Evaluation · p. 4Takeaways & Limitations
The simulations support combining SFGH for scheduling with TD3 for dispatch to improve service, financial, and carbon outcomes in the studied charging station.
Source: B. Simulation Evaluation · p. 4, §V · p. 5The study’s future work will incorporate battery degradation and grid capacity constraints to assess their effects on scheduling and dispatch performance.
Source: §V · p. 5
Abstract
from arXiv · showhide
Electric vehicle (EV) charging stations integrating photovoltaic (PV) generation, battery storage, and bidirectional grid exchange must coordinate access to limited chargers with energy supply while considering both energy cost and carbon emissions. This paper proposes a solver-free greedy heuristic (SFGH) for dynamic vehicle scheduling (VS) and a twin delayed deep deterministic policy gradient (TD3) controller for carbon-aware energy dispatch (ED). SFGH assigns EVs according to reductions in service delays, and TD3 controls battery operation to balance energy costs and carbon emissions while meeting the resulting charging demand. The model allows both PV and battery exports, accounting for export revenue and carbon credits. Simulations on year-long case studies show that SFGH--TD3 achieves the lowest accumulated service delay, energy cost, and carbon emissions among the evaluated heuristic and multi-agent deep reinforcement learning (DRL) configurations.
I. INTRODUCTION
EV charging stations must coordinate vehicle scheduling with energy dispatch as arrivals, energy needs, PV output, prices, and grid carbon intensity vary. The paper frames this coordination around limited chargers, service delays, and both financial and carbon consequences.
I. INTRODUCTION
- The paper identifies limited integration of vehicle scheduling with carbon-aware dispatch as a gap in managing charger access while balancing electricity cost and emissions.Prior work addresses economic operation or carbon-aware energy management, but the cited studies provide limited insight into combining scheduling and carbon-aware dispatch.Source: §I · p. 1
- The proposed approach combines a solver-free greedy heuristic for vehicle scheduling with TD3 energy dispatch and explicitly models charger occupancy, exports, and separate financial and carbon accounts.SFGH assigns vehicles based on service-delay reductions; TD3 balances financial and carbon objectives to meet the resulting demand.Source: §I · p. 1
II. SYSTEM MODELLING AND PROBLEM FORMULATION
- The station model passes aggregate charging demand from vehicle scheduling to dispatch, where PV, a battery, and a bidirectional grid connection supply demand or export energy.PV and stored energy can be exported, and the battery can store electricity from PV or the grid for later use.Source: §II · p. 2
A. Vehicle Scheduling
- Vehicle scheduling accounts for waiting and overflow beyond a threshold, while chargers serve at most one EV at a time and charging continues until its energy demand is met.A charger released before an interval ends may serve another waiting EV during the remainder.Source: A. Vehicle Scheduling · p. 2
B. Energy Dispatch
Energy dispatch meets the charging demand remaining after PV supply using battery operation and grid exchange. Its objectives include net electricity cost and grid-related carbon emissions under an export-credit accounting convention.
B. Energy Dispatch
- Battery charging and discharging are constrained by its energy capacity, power rating, efficiencies, and allowable state-of-charge range.Dispatch balances net cost and emissions over the operating horizon under the charging demand produced by scheduling.Source: B. Energy Dispatch · p. 2
- PV power is prioritized for EV charging, leaving positive net demand or surplus PV for battery charging or grid export.Grid imports and exports are mutually exclusive in the stated formulation.Source: B. Energy Dispatch · p. 2
- Dispatch accounts for net electricity cost and grid-related emissions, with exports receiving revenue and carbon credits at the same grid intensity used for imports.The resulting carbon measure is net emissions under the stated accounting convention.Source: B. Energy Dispatch · p. 2
III. APPROACH
The proposed control approach runs vehicle scheduling and energy dispatch sequentially at each interval. Scheduling determines assignments and demand, then TD3 dispatches battery operation using that demand, PV availability, prices, and carbon intensity.
III. APPROACH
- At each interval, SFGH uses EV waiting times, energy demands, and charger availability to assign vehicles and determine charging demand.Source: §III · p. 3
- Given charging demand and available PV generation, TD3 controls battery charging and discharging using electricity prices and grid carbon intensity.Source: §III · p. 3
- After dispatch, the station advances to the next interval and supplies updated observations for the next scheduling decision.Source: §III · p. 3
A. SFGH for Vehicle Scheduling
SFGH evaluates feasible vehicle–charger assignments by their reductions in waiting and overflow costs, then iteratively selects assignments to form the interval schedule and charging demand.
A. SFGH for Vehicle Scheduling
- SFGH limits evaluation to at most M candidate vehicles, prioritizing greater overflow, then longer waiting time, then smaller remaining energy demand.The selection order limits assignments evaluated and thereby reduces scheduling computational cost.Source: A. SFGH for Vehicle Scheduling · p. 3
- A vehicle–charger pair is feasible only when the charger becomes available before the interval ends, and SFGH compares waiting through the interval with starting at charger availability.Source: A. SFGH for Vehicle Scheduling · p. 3
- For each feasible assignment, SFGH scores the reductions in waiting and overflow costs, with κ weighting the overflow reduction.Source: A. SFGH for Vehicle Scheduling · p. 3
- SFGH selects the feasible pair with the largest score, updates the assigned EV’s waiting and charging duration, and repeats until candidates are exhausted or no charger becomes available.The resulting schedule determines the interval charging demand passed to energy dispatch; delay evaluation includes all waiting EVs, including those excluded from candidate selection.Source: A. SFGH for Vehicle Scheduling · p. 3
B. TD3 for Energy Dispatch
TD3 dispatches battery power using financial and carbon benefits relative to an idle-battery baseline. Its state combines current operating conditions with recent history, and its training updates twin critics and a less-frequently updated actor.
B. TD3 for Energy Dispatch
- Exploration noise is added to actor outputs during training but omitted during evaluation.TD3 is an off-policy continuous-control algorithm with a deterministic actor and two critics estimating discounted returns.Source: B. TD3 for Energy Dispatch · pp. 3–4
- The ED state includes net demand, purchase and sale prices, carbon intensity, battery state of charge, remaining EV energy, and time features, plus six historical time steps.The history augments the eight current features with past net demand, prices, and carbon intensity.Source: B. TD3 for Energy Dispatch · p. 3
- The actor’s normalized action determines battery charge or discharge power, which is clipped to satisfy SOC-dependent limits before setting grid exchange and the SOC update.Positive action values indicate discharge and negative values indicate charge.Source: B. TD3 for Energy Dispatch · p. 3
- The reward compares dispatch with an idle-battery baseline, combining electricity-cost savings and net grid-emissions reductions with carbon weight ω_ce.The baseline uses the same EV demand and PV generation, with the battery idle; surplus PV is exported in both cases.Source: B. TD3 for Energy Dispatch · p. 3
- TD3 trains twin critics on replay-buffer mini-batches using a smoothed target action and target value; terminal transitions omit the future-return term.The actor updates less frequently than the critics, and target-network parameters are softly updated after actor updates.Source: B. TD3 for Energy Dispatch · pp. 3–4
A. Simulation Settings
The simulations use year-long South Wales case-study datasets at 15-minute resolution, combining real electricity prices, synthetic transport demand, modeled PV generation, and grid-carbon data. The station and TD3 configuration follows specified charging, battery, and training settings.
A. Simulation Settings
- The year-long training set and three test sets use South Wales case-study data at 15-minute resolution, with real purchase prices, stochastic vehicle demand, modeled PV output, simulated sale prices, and NESO grid-carbon intensity.Vehicle arrivals follow non-homogeneous Poisson processes; sale prices are simulated as half the purchase price multiplied by a stochastic factor.Source: A. Simulation Settings · p. 4
- The station configuration uses five 90 kW chargers and a 900 kWh battery rated at 450 kW, with SOC limits of 0.10–0.95 and scheduling and carbon weights of 1 and 0.5.TD3 uses two 256-unit hidden layers and is trained for 15,000 epochs, each with up to 768 decision intervals.Source: A. Simulation Settings · p. 4
B. Simulation Evaluation
Across the test sets, SFGH–TD3 achieved the lowest mean net cost and emissions, while SFGH produced the lowest service delays. Annual outcome decomposition shows how exports contribute to both financial and carbon results.
B. Simulation Evaluation
- Table I compares seven VS–ED combinations; learned configurations use five policies trained with different seeds, while deterministic baselines are evaluated once per test set.Annual performance uses cumulative service delays over all intervals in each year-long test set.Source: B. Simulation Evaluation · p. 4, Table I · p. 5
- SFGH had the lowest cumulative service delays in all three test sets, reducing delays by 2.62%, 2.88%, and 2.85% versus FIFO.With ED fixed to TD3, it also improved on PPO–TD3 by 4.10%, 5.85%, and 5.21%, and on TD3–TD3 by 4.24%, 5.61%, and 4.87%.Source: B. Simulation Evaluation · p. 4, Table I · p. 5
- SFGH–TD3 achieved the lowest mean net cost and net emissions in all test sets, cutting both versus No ESS.With VS fixed to SFGH, TD3 reduced cost by 8.51–8.82% and emissions by 5.67–5.80% compared with No ESS.Source: B. Simulation Evaluation · p. 4, Table I · p. 5
- Exports offset 11.22–11.89 kGBP in annual grid-purchase costs and provide 47.22–47.59 tCO2 in carbon credits.Net costs were 211.87–214.81 kGBP and net emissions were 361.82–365.21 tCO2 under the adopted accounting convention.Source: B. Simulation Evaluation · p. 4, Figure 3 · p. 4
C. Operating Performance
In a representative 72-hour period, queues cleared as chargers became available, while PV and battery discharge partly met charging demand. PV allocation and storage timing illustrate cost and emissions management, bounded by storage capacity.
C. Operating Performance
- Busy-period queues cleared as chargers became available, with no persistent backlog during the representative 72-hour period.PV generation and battery discharge partially supplied EV charging demand, reducing reliance on direct grid purchases.Source: C. Operating Performance · p. 5
- PV primarily supplied EV charging, while surplus was stored or exported to address mismatches between generation and demand.This operating allocation is shown in Fig. 5(a).Source: C. Operating Performance · p. 5
- Charging before purchase-price peaks and discharging near them illustrates the energy-arbitrage mechanism for reducing net cost.Charging at relatively low carbon intensity and discharging as intensity rises provides an additional opportunity to reduce net emissions.Source: C. Operating Performance · p. 5
- Storage capacity limits how much energy can be shifted across intervals, as indicated by state-of-charge plateaus.Source: C. Operating Performance · p. 5
V. CONCLUSION
The paper concludes that SFGH–TD3 is an approach for vehicle scheduling and carbon-aware energy dispatch at a PV–battery charging station with bidirectional grid exchange. Simulations support its effectiveness; future work will examine battery degradation and grid capacity constraints.
V. CONCLUSION
- SFGH–TD3 combines vehicle scheduling with carbon-aware energy dispatch at a PV–battery station with bidirectional grid exchange.SFGH assigns chargers based on waiting and overflow reductions; TD3 controls battery operation within feasible limits using electricity prices and carbon intensity.Source: §V · p. 5
- Simulation studies support the effectiveness of the proposed SFGH–TD3 approach.Source: §V · p. 5
- Future work will incorporate battery degradation and grid capacity constraints to assess their effects on scheduling and dispatch performance.Source: §V · p. 5