pith. sign in

arxiv: 2605.24543 · v1 · pith:6R2Z3PHLnew · submitted 2026-05-23 · 💻 cs.AI · cs.SY· eess.SY

Emission-Aware Reinforcement Learning for Sustainable Electric Vehicle Charging and Carbon Dioxide Reduction Under Varying Renewable Penetration

Ninglin Ou , Mohammad A. Razzaque , Iftekher Islam Shovon , Shafkat Khan Siam , Shafiuzzaman K Khadem , Krishnendu Guha , Mayeen U Khandaker , Md. Noor-A-Rahim This is my paper

Pith reviewed 2026-06-30 13:06 UTC · model grok-4.3

classification 💻 cs.AI cs.SYeess.SY
keywords emission-aware reinforcement learningelectric vehicle chargingcarbon dioxide reductionrenewable penetrationsoft actor-criticpower distribution networkEV2Gym simulatorworkplace charging
0
0 comments X

The pith

An emission-aware reinforcement learning agent using Soft Actor-Critic reduces EV charging carbon intensity to 23.96 gCO2/kWh under 50% wind penetration.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes that a reinforcement learning controller can schedule workplace electric vehicle charging to minimize grid carbon emissions while meeting user demand and avoiding grid overloads. It trains a Soft Actor-Critic agent on a multi-objective reward that penalizes high carbon intensity, wasted on-site renewables, and unmet charging requests inside the EV2Gym simulator with real EirGrid carbon data and behind-the-meter wind and solar profiles. The agent is tested against nine other strategies across five renewable penetration levels from 0% to 50%. If the results hold, fleets could shift charging to low-emission windows without new hardware, cutting emissions substantially while preserving transformer health and renewable self-consumption.

Core claim

The SAC agent achieves a carbon intensity as low as 23.96 grams of carbon dioxide per kilowatt-hour under 50% wind penetration, representing up to 87% emission reduction versus the uncontrolled baseline, and outperforms the external graph-based Power Distribution Network benchmark. Transformer overload stays below 7 kWh across scenarios compared with up to 1093 kWh for the As Fast As Possible heuristic, and renewable self-consumption reaches 52% under combined wind and solar supply.

What carries the argument

The multi-objective reward in the Soft Actor-Critic algorithm that penalizes carbon emissions, curtailed renewables, and unmet demand inside the EV2Gym environment with time-varying carbon intensity in the state.

If this is right

  • Charging schedules align with low-emission grid periods while keeping overloads low.
  • Renewable self-consumption improves to 52% under combined wind and solar supply.
  • The agent outperforms both heuristic and model-predictive baselines across all tested renewable shares.
  • Grid compliance and user satisfaction are preserved alongside the emission reductions.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same reward structure could be adapted to include vehicle-to-grid discharge during high-carbon periods.
  • Scaling the method to residential chargers would require adjusting the state for different arrival and departure patterns.
  • Embedding live carbon-price signals in the reward might further increase alignment with market incentives.

Load-bearing premise

The EV2Gym simulator with behind-the-meter solar and wind profiles and EirGrid carbon intensity data sufficiently represents real-world EV user behavior, grid constraints, and renewable variability.

What would settle it

A controlled deployment on live EVSE units connected to a real distribution feeder that records actual carbon intensity and measures whether emission reductions reach the simulated 87% level versus an uncontrolled baseline.

Figures

Figures reproduced from arXiv: 2605.24543 by Iftekher Islam Shovon, Krishnendu Guha, Mayeen U Khandaker, Md. Noor-A-Rahim, Mohammad A. Razzaque, Ninglin Ou, Shafiuzzaman K Khadem, Shafkat Khan Siam.

Figure 1
Figure 1. Figure 1: An EV2Gym simulation consists of three phases: configuration, where models are initialised; simulation, where system states evolve over 𝑇 steps under the chosen control algorithm; and evaluation, where performance metrics, replay files, and optional visualisations are generated [11] [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The integrated system architecture incorporates CO2 intensity data, wind generation, and overall energy supply profiles. During simulation, the states of models such as EVs, charging stations, and emissions are updated according to the emission-aware RL algorithm. 4.1. EV2Gym Simulator Overview EV2Gym is an open-source simulation platform designed for comprehensive EV smart charging studies with V2G suppor… view at source ↗
Figure 3
Figure 3. Figure 3: Reward function flowchart. 6.2.1. Emission-Aware State Function Design In this work, we propose an emission-aware state representation for smart EV charging control. The state at timestep 𝑡 is defined as: First Author et al.: Preprint submitted to Elsevier Page 12 of 31 [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of baseline strategies. AFAP yields the highest CO2 emissions, whereas ALAP and RR align better with renewable generation and achieve lower emissions. for comparing strategy performance under varying renewable availability and load conditions [11]. This allows us to directly assess trade-offs (e.g. between emissions reduction and user satisfaction) across the Baseline, Smart Charging, and Smart+… view at source ↗
Figure 5
Figure 5. Figure 5: Energy Levels Over Time Across Charging Stations. In fact, overloads were effectively 0 kWh under 50% Solar and 50% Wind penetration levels, compared to hundreds of kWh with AFAP. However, this improvement comes at the expense of user satisfaction: ALAP achieved only 84-89% satisfaction, reflecting its sensitivity to accurately known departure times. Another rule-based baseline, RR, distributes available p… view at source ↗
Figure 6
Figure 6. Figure 6: Flexibility of Baseline Charging Strategies Across Scenarios. no-RE, 14 kg in 25% Hybrid, 30 kg in 50% Solar). For example, in the 50% Solar scenario, AFAP emits 66 kg CO2 versus 30 kg for ALAP. In the 50% Hybrid and 50% wind scenarios, emissions drop sharply across all strategies; under 50% wind they are effectively zero (0–0.03 kg). Among the AFAP family of strategies, the FSB consistently produced the h… view at source ↗
Figure 7
Figure 7. Figure 7: MPC Strategy Performance Across Scenarios. 7.3. Reinforcement Learning Strategy Evaluation To evaluate the effectiveness of our emission-aware RL agent, we conducted experiments under five distinct scenarios: No RE, 50% Wind, 50% Solar, 25% Hybrid, and 50% Hybrid. Each scenario was simulated for 10 independent runs, and the results reported below are averages across these runs to ensure robustness. Perform… view at source ↗
Figure 8
Figure 8. Figure 8: CO2 Emissions of RL Strategies Across Scenarios [PITH_FULL_IMAGE:figures/full_fig_p022_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Energy Charged and Discharged by RL Strategies Across Scenarios [PITH_FULL_IMAGE:figures/full_fig_p022_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Evaluation reward trajectories of the SAC agent over 4 million training steps across five independent runs [PITH_FULL_IMAGE:figures/full_fig_p024_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Actor and critic loss curves from a representative SAC training run over 4 million steps. evaluation reward improves substantially during the early and intermediate phases of training, then increases more slowly in the final stage. This suggests practical convergence rather than full saturation: the policy continues to improve, but the marginal gains become smaller toward the end of training. The relative… view at source ↗
Figure 12
Figure 12. Figure 12: Reward component ablation under different renewable settings and penetration levels. carbon weights, especially 𝑊𝐶𝑂2 = 1, generally increase emissions in the Hybrid and Wind settings at low to medium penetration, with 49.38 kgCO2 at 25% Hybrid and 50.22 kgCO2 at 25% Wind. This indicates that a weak carbon signal is not sufficient to shift charging reliably away from high-emission periods. Increasing the w… view at source ↗
Figure 13
Figure 13. Figure 13: Sensitivity of policy performance to the carbon reward weight relative to the 𝑊 = 5 baseline. 7.5.3. Robustness Across Renewable Penetration Levels Fig.14 evaluates the selected RL policy across a wider renewable-penetration sweep under Solar, Wind, and Hybrid supply. In all three cases, carbon emission decreases as renewable penetration rises, but the rate of improvement depends strongly on the renewable… view at source ↗
Figure 14
Figure 14. Figure 14: Effect of renewable penetration on renewable utilisation and carbon emission for the selected RL policy [PITH_FULL_IMAGE:figures/full_fig_p028_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Distribution of carbon emission under different charging demand levels. at 22.14%. The shaded bands also narrow at higher penetration levels, especially under Wind and Hybrid supply, indicating that the learned policy becomes more stable as more clean energy is available. 7.5.4. Sensitivity to Charging Demand Variability Figs. 15 and 16 examine the response of the selected RL policy under low-, medium-, a… view at source ↗
Figure 16
Figure 16. Figure 16: Operational response of the selected RL policy under different charging demand levels. 8. Conclusion This paper employed the EV2Gym platform to evaluate heuristic, MPC, and RL controllers in realistic V2G smart charging scenarios. The RL-based strategy demonstrated strong adaptability across diverse renewable scenarios (Wind, Solar, and Hybrid), effectively balancing competing objectives such as minimisin… view at source ↗
read the original abstract

The rapid growth of Electric Vehicle (EV) adoption challenges power distribution networks through peak load spikes, voltage instability, and transformer overloads from uncoordinated charging. While Model Predictive Control (MPC) and standard Reinforcement Learning (RL) methods have addressed these issues, existing approaches rarely treat real-time carbon intensity or fluctuating renewable energy (RE) availability as primary scheduling objectives, leaving substantial decarbonisation potential unrealised. This paper proposes an emission-aware RL strategy based on the Soft Actor Critic (SAC) algorithm, with a multi-objective reward that penalises carbon emissions, curtailed on-site renewables, and unmet user demand. The agent is trained within a unified benchmarking framework on the EV2Gym platform, incorporating behind-the-meter solar and wind profiles, time-varying EirGrid carbon intensity data, and realistic workplace EV behaviour across 25 Electric Vehicle Supply Equipment (EVSE) units. Nine control strategies, including heuristics, emission-aware MPC variants, and the proposed RL agent, are compared under five renewable penetration scenarios (0%-50%) over ten independent runs each. The RL agent achieves a carbon intensity as low as 23.96 grams of carbon dioxide per kilowatt-hour under 50% wind penetration, representing up to 87% emission reduction versus the uncontrolled baseline, and outperforms the external graph-based Power Distribution Network (PDN) benchmark. Transformer overload remains below 7 kWh across scenarios, against up to 1093 kWh for the As Fast As Possible (AFAP) heuristic, and renewable self-consumption reaches 52% under combined wind and solar supply. Embedding carbon intensity forecasts into the RL state and reward aligns charging with low-emission periods while preserving grid compliance and user satisfaction.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The manuscript proposes an emission-aware Soft Actor-Critic (SAC) reinforcement learning agent for EV charging scheduling. It uses a multi-objective reward penalizing carbon emissions, renewable curtailment, and unmet demand, trained in the EV2Gym simulator incorporating behind-the-meter solar/wind profiles, EirGrid carbon intensity data, and realistic workplace EV behavior across 25 EVSE units. Nine strategies (heuristics, MPC variants, RL) are compared over five renewable penetration scenarios (0-50%) in ten runs each. The central claims are that the RL agent reaches a carbon intensity of 23.96 gCO2/kWh at 50% wind penetration (up to 87% reduction vs. uncontrolled baseline), outperforms the graph-based PDN benchmark, keeps transformer overload below 7 kWh (vs. 1093 kWh for AFAP), and achieves 52% renewable self-consumption.

Significance. If the simulation results are robust and transferable, the work demonstrates that embedding carbon forecasts into RL state and reward can yield substantial emission reductions in EV charging while maintaining grid compliance and user satisfaction, outperforming both simple heuristics and MPC under varying renewable levels. The multi-scenario evaluation and use of real carbon intensity data provide a useful benchmark for sustainable control strategies.

major comments (2)
  1. [Abstract/Methods] Abstract and Methods: The reported performance numbers (23.96 gCO2/kWh, 87% reduction, outperformance vs. PDN) are obtained from ten independent runs but supply no reward-function weights for the multi-objective SAC, no SAC training hyperparameters, and no statistical significance tests or sensitivity analysis on user-arrival stochasticity; these omissions make the central quantitative claims impossible to reproduce or assess for robustness.
  2. [Methods] Methods: The 87% emission-reduction claim and outperformance results rest entirely on the EV2Gym simulator's internal models of EV user behavior, behind-the-meter renewable profiles, and grid constraints, yet no validation against empirical charging-session logs, grid telemetry, or sensitivity tests on renewable variability/curtailment dynamics is reported; this is load-bearing because any systematic mismatch would render the metrics simulation artifacts rather than transferable findings.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the constructive comments on reproducibility and robustness. We address each major point below and will revise the manuscript to improve transparency and add requested analyses where feasible.

read point-by-point responses

  1. Referee: [Abstract/Methods] Abstract and Methods: The reported performance numbers (23.96 gCO2/kWh, 87% reduction, outperformance vs. PDN) are obtained from ten independent runs but supply no reward-function weights for the multi-objective SAC, no SAC training hyperparameters, and no statistical significance tests or sensitivity analysis on user-arrival stochasticity; these omissions make the central quantitative claims impossible to reproduce or assess for robustness.

    Authors: We agree the manuscript omitted these details. In the revision we will explicitly report the reward-function weights for emissions, curtailment and unmet demand; the complete SAC hyperparameters (learning rate, discount factor, batch size, network architecture, entropy coefficient schedule); results of statistical significance tests (e.g., paired t-tests or Wilcoxon tests across the ten runs); and a sensitivity analysis that perturbs user-arrival distributions while keeping other factors fixed. revision: yes

  2. Referee: [Methods] Methods: The 87% emission-reduction claim and outperformance results rest entirely on the EV2Gym simulator's internal models of EV user behavior, behind-the-meter renewable profiles, and grid constraints, yet no validation against empirical charging-session logs, grid telemetry, or sensitivity tests on renewable variability/curtailment dynamics is reported; this is load-bearing because any systematic mismatch would render the metrics simulation artifacts rather than transferable findings.

    Authors: The study is a controlled simulation benchmark using real EirGrid carbon-intensity traces and the EV2Gym environment. We will add sensitivity experiments that systematically vary renewable generation profiles and curtailment parameters to quantify robustness. Full empirical validation against proprietary charging-session logs and grid telemetry is not possible in the current work because such datasets were unavailable to the authors. revision: partial

standing simulated objections not resolved
  • Empirical validation of simulator models against real-world charging-session logs and grid telemetry, as such data were not accessible.

Circularity Check

0 steps flagged

No significant circularity; results are forward simulation comparisons

full rationale

The paper's central claims consist of empirical performance metrics (carbon intensity of 23.96 gCO2/kWh, up to 87% reduction vs baseline, outperformance vs heuristics/MPC/PDN) obtained by training an SAC agent with a multi-objective reward in the EV2Gym simulator and running forward simulations under five renewable scenarios. No equations, fitted parameters, or self-citations are shown that reduce these metrics to the inputs by construction. The reward design and state embedding are explicit choices, not tautological reductions. Comparisons to external baselines and ten independent runs make the evaluation self-contained and falsifiable outside the paper's fitted values.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract supplies no explicit free parameters, axioms, or invented entities; the central claim rests on the unstated modeling assumptions inside the EV2Gym simulator and the choice of reward weights that are never quantified.

pith-pipeline@v0.9.1-grok · 5896 in / 1223 out tokens · 34572 ms · 2026-06-30T13:06:53.083738+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

45 extracted references · 2 canonical work pages · 1 internal anchor

  1. [1]

    Licence: CC BY 4.0

    IEA, Global ev outlook 2024,https://www.iea.org/reports/global-ev-outlook-2024, 2024. Licence: CC BY 4.0

    2024

  2. [2]

    Rossi, C

    F. Rossi, C. Diaz-Londono, Y. Li, C. Zou, G. Gruosso, Smart electric vehicle charging algorithm to reduce the impact on power grids: A reinforcement learning based methodology, IEEE Open Journal of Vehicular Technology (2025)

    2025

  3. [3]

    Panda, S

    S. Panda, S. Ganguly, A multi-objective optimization model for smart ev charging scheduling considering the benefits of ev owners and distribution network operators, in: 2024 23rd National Power Systems Conference (NPSC), IEEE, 2024, pp. 1–6

    2024

  4. [4]

    J.Chu,B.Gilmore,J.Hassol,A.Jenn,S.Lommele,L.Myers,H.Richardson,A.Schroeder,M.Shah,NationalElectricVehicleInfrastructure Formula Program Annual Report: Plan Year 2022-2023, Technical Report, National Renewable Energy Laboratory (NREL), Golden, CO (United States), 2023

    2022

  5. [5]

    Şengör, O

    İ. Şengör, O. Erdinç, B. Yener, A. Taşcıkaraoğlu, J. P. Catalao, Optimal energy management of ev parking lots under peak load reduction based dr programs considering uncertainty, IEEE Transactions on Sustainable Energy 10 (2018) 1034–1043

    2018

  6. [6]

    K. S. Holkar, L. M. Waghmare, An overview of model predictive control, International Journal of control and automation 3 (2010) 47–63

    2010

  7. [7]

    Kempton, J

    W. Kempton, J. Tomić, Vehicle-to-grid power fundamentals: Calculating capacity and net revenue, Journal of power sources 144 (2005) 268–279

    2005

  8. [8]

    M. Dan, A. Easwaran, A cooperative bargaining game framework for vehicle-to-vehicle energy sharing and trading at charging stations, in: 2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), IEEE, 2024, pp. 96–103

    2024

  9. [9]

    Deep Reinforcement Learning: An Overview

    Y. Li, Deep reinforcement learning: An overview, arXiv preprint arXiv:1701.07274 (2017)

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  10. [10]

    Y.Yang,H.-G.Yeh,R.Nguyen, Arobustmodelpredictivecontrol-basedschedulingapproachforelectricvehiclechargingwithphotovoltaic systems, IEEE Systems Journal 17 (2022) 111–121

    2022

  11. [11]

    Orfanoudakis, C

    S. Orfanoudakis, C. Diaz-Londono, Y. Yılmaz, P. Palensky, P. Vergara, Ev2gym: A flexible v2g simulator for ev smart charging research and benchmarking, 2024

    2024

  12. [12]

    Y. E. Yılmaz, S. Orfanoudakis, P. P. Vergara, Reinforcement learning for optimized ev charging through power setpoint tracking, in: 2024 IEEE PES Innovative Smart Grid Technologies Europe (ISGT EUROPE), IEEE, 2024, pp. 1–5

    2024

  13. [13]

    A.-M.Koufakis,E.S.Rigas,N.Bassiliades,S.D.Ramchurn, Offlineandonlineelectricvehiclechargingschedulingwithv2venergytransfer, IEEE Transactions on Intelligent Transportation Systems 21 (2019) 2128–2138

    2019

  14. [14]

    U.Qureshi,A.Ghosh,B.K.Panigrahi, Multiobjectivepareto-optimalintelligentelectricvehiclechargingscheduleinacommercialcharging station: A stochastic convex optimization approach, IEEE Transactions on Industrial Informatics (2024)

    2024

  15. [15]

    Hosseini, A

    S. Hosseini, A. Yassine, M. S. Hossain, Optimizing electric vehicle charging through an artificial intelligence mechanism for smart transportation, IEEE Internet of Things Journal (2024)

    2024

  16. [16]

    W.Kanchana,J.G.Singh,W.Ongsakul,etal., Optimizingelectricvehiclechargingschedules:Minimizingchargingtimeandtraveldistance, in: 2024 International Conference on Sustainable Energy: Energy Transition and Net-Zero Climate Future (ICUE), IEEE, 2024, pp. 1–5

    2024

  17. [17]

    Diaz-Londono, G

    C. Diaz-Londono, G. Fambri, P. Maffezzoni, G. Gruosso, Enhanced ev charging algorithm considering data-driven workplace chargers categorization with multiple vehicle types, eTransportation 20 (2024) 100326

    2024

  18. [18]

    D.Qiu,Y.Wang,W.Hua,G.Strbac, Reinforcementlearningforelectricvehicleapplicationsinpowersystems:Acriticalreview, Renewable and Sustainable Energy Reviews 173 (2023) 113052

    2023

  19. [19]

    Zhang, Z

    F. Zhang, Z. Wang, Y. Li, C. Zhang, 2019 ieee international conference on communications, control, and computing technologies for smart grids (smartgridcomm) (2019)

    2019

  20. [20]

    C. Yeh, V. Li, R. Datta, J. Arroyo, N. Christianson, C. Zhang, Y. Chen, M. M. Hosseini, A. Golmohammadi, Y. Shi, et al., Sustaingym: Reinforcementlearningenvironmentsforsustainableenergysystems, AdvancesinNeuralInformationProcessingSystems36(2023)59464– 59476

    2023

  21. [21]

    D. Qiu, Y. Wang, Z. Ding, G. Strbac, Graph reinforcement learning for carbon-aware electric vehicles in power-transport networks, IEEE Transactions on Smart Grid 15 (2024) 3919–3935. First Author et al.:Preprint submitted to ElsevierPage 30 of 31 Short Title of the Article

    2024

  22. [22]

    J. Fan, A. Liebman, H. Wang, Safety-aware reinforcement learning for electric vehicle charging station management in distribution network, in: 2024 IEEE Power and Energy Society General Meeting (PESGM), IEEE, 2024. doi:10.1109/PESGM51994.2024.10688856

    work page doi:10.1109/pesgm51994.2024.10688856 2024

  23. [23]

    Q.Zhao,C.Xu,C.Sun,Y.Han, Smartresidentialelectricvehiclecharginganddischargingschedulingviamulti-agentasynchronous-updating deep reinforcement learning, Computers and Electrical Engineering 126 (2025) 110473

    2025

  24. [24]

    C.A.M.Silva,R.J.Bessa, Carbon-awaredynamictariffdesignforelectricvehiclechargingstationswithexplainablestochasticoptimization, Applied Energy 389 (2025) 125674

    2025

  25. [25]

    Meenakumar, M

    P. Meenakumar, M. Aunedi, G. Strbac, Optimal business case for provision of grid services through evs with v2g capabilities, in: 2020 Fifteenth International Conference on Ecological Vehicles and Renewable Energies (EVER), IEEE, 2020, pp. 1–10

    2020

  26. [26]

    C. F. Lee, K. Bjurek, V. Hagman, Y. Li, C. Zou, Vehicle-to-grid optimization considering battery aging, IFAC-PapersOnLine 56 (2023) 6624–6629

    2023

  27. [27]

    Saxena, J

    S. Saxena, J. MacDonald, S. Moura, Charging ahead on the transition to electric vehicles with standard 120v wall outlets, Applied Energy 157 (2015) 720–728

    2015

  28. [28]

    E. S. Rigas, S. Karapostolakis, N. Bassiliades, S. D. Ramchurn, Evlibsim: A tool for the simulation of electric vehicles’ charging stations using the evlib library, Simulation Modelling Practice and Theory 87 (2018) 99–119

    2018

  29. [29]

    T.Morstyn,K.A.Collett,A.Vijay,M.Deakin,S.Wheeler,S.M.Bhagavathy,F.Fele,M.D.McCulloch, Open:Anopen-sourceplatformfor developing smart local energy system applications, Applied Energy 275 (2020) 115397

    2020

  30. [30]

    G.Karatzinis,C.Korkas,M.Terzopoulos,C.Tsaknakis,A.Stefanopoulou,I.Michailidis,E.Kosmatopoulos,Chargym:Anevchargingstation modelforcontrollerbenchmarking, in:IFIPInternationalConferenceonArtificialIntelligenceApplicationsandInnovations,Springer,2022, pp. 241–252

    2022

  31. [31]

    EirGrid and SONI, Smart Grid Dashboard: CO2 Intensity,https://www.smartgriddashboard.com/roi/co2/, 2025

    2025

  32. [32]

    B. Liu, W. Ni, R. P. Liu, Y. J. Guo, H. Zhu, Optimal electric vehicle charging strategies for long-distance driving, IEEE Transactions on Vehicular Technology 73 (2023) 4949–4960

    2023

  33. [33]

    URL:https://www.ipcc-nggip.iges.or.jp/public/2006gl/, prepared by the National Greenhouse Gas Inventories Programme

    Intergovernmental Panel on Climate Change (IPCC), 2006 IPCC Guidelines for National Greenhouse Gas Inventories, Institute for Global Environmental Strategies (IGES), Hayama, Japan, 2006. URL:https://www.ipcc-nggip.iges.or.jp/public/2006gl/, prepared by the National Greenhouse Gas Inventories Programme

    2006

  34. [34]

    URL:https://ghgprotocol

    World Resources Institute (WRI), World Business Council for Sustainable Development (WBCSD), GHG Protocol Scope 2 Guidance: An Amendment to the GHG Protocol Corporate Standard, World Resources Institute, Washington, DC, 2015. URL:https://ghgprotocol. org/sites/default/files/2023-03/Scope%202%20Guidance.pdf

    2015

  35. [35]

    EuropeanCommissionJointResearchCentre,PhotovoltaicGeographicalInformationSystem(PVGIS),https://re.jrc.ec.europa.eu/ pvg_tools/en/, 2025

    2025

  36. [36]

    Holttinen, J

    H. Holttinen, J. Kiviluoma, D. Flynn, J. C. Smith, A. Orths, P. B. Eriksen, N. Cutululis, L. Söder, M. Korpås, A. Estanqueiro, et al., System impactstudiesfornear100%renewableenergysystemsdominatedbyinverterbasedvariablegeneration, IEEETransactionsonPowerSystems 37 (2020) 3249–3258

    2020

  37. [37]

    Dudurych, M

    I. Dudurych, M. Holly, M. Power, Integration of wind power generation in the irish grid, in: 2006 IEEE Power Engineering Society General Meeting, IEEE, 2006, pp. 8–pp

    2006

  38. [38]

    Hurtado, T

    M. Hurtado, T. Kërçi, S. Tweed, E. Kennedy, N. Kamaluddin, F. Milano, Analysis of wind energy curtailment in the ireland and northern ireland power systems, in: 2023 IEEE Power & Energy Society General Meeting (PESGM), IEEE, 2023, pp. 1–5

    2023

  39. [39]

    A. H. Ganesh, B. Xu, A review of reinforcement learning based energy management systems for electrified powertrains: Progress, challenge, and potential solution, Renewable and Sustainable Energy Reviews 154 (2022) 111833

    2022

  40. [40]

    Raffin, A

    A. Raffin, A. Hill, A. Gleave, A. Kanervisto, M. Ernestus, N. Dormann, Stable-baselines3: Reliable reinforcement learning implementations, Journal of machine learning research 22 (2021) 1–8

    2021

  41. [41]

    J. Fan, A. Liebman, H. Wang, Safety-aware reinforcement learning for electric vehicle charging station management in distribution network, in: 2024 IEEE Power & Energy Society General Meeting (PESGM), IEEE, 2024, pp. 1–5

    2024

  42. [42]

    U.Damodarin,G.C.Cardarilli,L.DiNunzio,M.Re,S.Spanò, Smartelectricvehiclechargingmanagementusingreinforcementlearningon FPGA platforms, Sensors 25 (2025) 2585

    2025

  43. [43]

    K. Park, I. Moon, Multi-agent deep reinforcement learning approach for ev charging scheduling in a smart grid, Applied Energy 328 (2022) 120111

    2022

  44. [44]

    URL:http://data

    European Parliament and the Council of the European Union, Regulation (eu) 2023/1804 of the european parliament and of the council of 13 september 2023 on the deployment of alternative fuels infrastructure, and repealing directive 2014/94/eu, 2023. URL:http://data. europa.eu/eli/reg/2023/1804/oj, text with EEA relevance

    2023

  45. [45]

    URL:http://data.europa.eu/eli/reg_del/2025/656/oj, published in the Official Journal of the European Union, 18 June 2025

    EuropeanCommission,Commissiondelegatedregulation(eu)2025/656of2april2025amendingregulation(eu)2023/1804oftheeuropean parliamentandofthecouncilasregardsstandardsforwirelessrecharging,electricroadsystem,vehicle-to-gridcommunicationandhydrogen supply for road transport vehicles, 2025. URL:http://data.europa.eu/eli/reg_del/2025/656/oj, published in the Offici...

    2025