Pith. sign in

REVIEW 3 major objections 4 minor 55 references

Deep reinforcement learning-based joint real-time energy scheduling for green buildings with heterogeneous battery energy storage devices

T0 review · 3 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper claims a deep reinforcement learning scheduler that coordinates stationary building batteries and parked electric-vehicle batteries in real time, and reports 37.94-40.01% lower operating cost than a MILP baseline in a…

desk verdict Competent engineering, but the headline cost reduction rests on an unspecified MILP baseline and a sign error in the objective; both need fixing before the claim can be evaluated. read the letter →

arxiv 2506.06824 v2 pith:VPVYQTAQ submitted 2025-06-07 eess.SY cs.SY

Chi Liu , Zhezhuang Xu , Jiawei Zhou , Yazhou Yuan , Kai Ma , Meng Yuan This is my paper
classification eess.SYcs.SY
keywords greenbuildingdeepreinforcementlearningjointenergyschedulingbatterydegradationelectricvehiclesmanagementloadforecastingvehicle-to-grid
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to show that a model-free deep reinforcement learning controller can schedule, hour by hour, the combined battery system of a green commercial building—stationary storage plus parked electric vehicles with vehicle-to-grid capability—so that operating cost falls well below what a mixed-integer linear programming schedule achieves. The authors report a 37.94-40.01% cost reduction over MILP while enforcing two constraints that often make EV storage unattractive: the battery must not degrade faster than its chemistry warrants, and each EV must leave with at least the state of charge it arrived with. If true, the result matters because real-time building energy management could then learn arbitrage and load-priority policies without a physical model, and EV owners could lend their batteries without sacrificing range. The paper's central mechanism is a reward function that prices battery chemistry-specific degradation separately for LFP stationary storage and NMC EV cells, together with forecast-guided charging and discharging priorities.

What carries the argument

The central object is the combined battery system (CBS): a controllable pool made of stationary LFP storage and an aggregated fleet of parked NMC electric vehicles, each with its own charging and discharging constraints, scheduling window, and degradation cost. The argument is carried by a degradation-aware reward function in which per-step cycle-aging coefficients, computed with a rainflow cycle-counting algorithm and semi-empirical LFP/NMC aging models, are turned into costs; battery-specific scaling coefficients make the agent treat EV capacity as more expensive than ESS capacity; and a sell-back price ratio together with a net-load ratio shape when charging or discharging is rewarded. A ranking-based dynamic ensemble deep random vector functional link forecaster supplies 24-hour net-load predictions to the state, and the action is chosen from a 25-option discrete ESS-by-EV action set by a double dueling DQN with prioritized experience replay.

What would settle it

Re-run the summer and winter test sets with a written-down MILP whose objective is the one in Section 3.8 and whose constraints include the ESS and EV operating rules, using the same degradation costs, the same 24-hour forecasts, and the same EV state-of-charge departure rule; if the MILP produces a feasible schedule with cost at or below the DRL policy's reported cost, the central cost-reduction claim is falsified.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that the RDedRVFL-D3QNPER algorithm—an ensemble deep-learning forecaster feeding a double-dueling deep Q-network with prioritized replay—learns a joint real-time scheduling policy for the combined battery system that reduces system operating costs by 37.94-40.01% relative to a MILP approach on summer and winter commercial-building test sets. The learned policy charges both storage types at valley tariffs, discharges at peak tariffs and high net load, prioritizes the building's load before selling to the grid, and uses the stationary ESS more often than the EV fleet because EV cycle-aging cost is priced higher. It satisfies the travel-dependent constraint that EV state of charge on departure not be lower than on arrival, and it keeps EV state of health above 99% in both seasonal tests. The paper also reports ablations showing that removing the forecast module raises costs by roughly 31%, removing the energy-allocation mechanism raises costs in proportion to the sell-back price gap, and removing the combined battery system raises costs by 25.65% in summer and 16.24% in winter.

Load-bearing premise

The load-bearing premise is that the MILP baseline is a fair and correctly specified solver for the same optimization problem, with the same degradation costs, forecast information, and EV constraints; the paper gives no MILP formulation, so this premise is unverified.

Editorial extensions

If this is right

  • If the central claim is correct, commercial-building operators can replace a re-solved MILP with a learned policy that produces decisions in about 0.12 seconds per step, far below the one-hour control interval.
  • The learned ESS-first, EV-second dispatch pattern implies that heterogeneous storage can be coordinated without a detailed system model, as long as each battery type's degradation is priced into the reward.
  • The ablation results imply that forecast quality and the energy-allocation priority are the main drivers of value: removing the forecast module raises costs by roughly 31%, and removing the allocation mechanism costs more as the sell-back price ratio falls.
  • The small EV-only benefit reported in Section 6.4.2 means the practical near-term value of EV participation is flexibility and avoided ESS capacity investment, not large arbitrage profit.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Beyond the paper, the stated 37.94-40.01% gap should be treated as a property of the implemented MILP comparison rather than of the algorithm, because no MILP formulation is given; a reader could test this by re-running the comparison with an explicit MILP that includes the Section 3.8 objective and the degradation costs.
  • Beyond the paper, the degradation parameters are calibrated to lab-scale LFP and NMC cells, so applying the method to other chemistries, temperatures, or battery management systems would require recalibrating the aging models before the cost split between ESS and EV is trusted.
  • Beyond the paper, the fixed parking window and the arrival-and-departure state-of-charge rule assume predictable EV behavior; adding stochastic departure times, which the paper lists as future work, would probably narrow the reported gap.
  • Beyond the paper, the hand-designed exponential scaling in the reward function does much of the work of deciding when an EV is worth scheduling; replacing it with a learned reward or an explicit constraint would show whether D3QNPER alone, or the engineered reward, carries the performance.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper proposes a model-free deep reinforcement learning method (RDedRVFL-D3QNPER) for joint real-time scheduling of stationary ESS and EV batteries in a green building, incorporating differentiated battery degradation models, a load/PV prediction module, and an energy allocation mechanism. The central claim is a 37.94%-40.01% reduction in system operating costs compared to a mixed-integer linear programming (MILP) approach, demonstrated on summer and winter commercial-building datasets.

Significance. The problem is well motivated, and the paper includes several strong elements: a differentiated LFP/NMC degradation model, explicit handling of EV range constraints, an ablation study of the prediction module and energy allocation mechanism, and a comparison against DQN and D3QN baselines. The reported inference time of 0.12 s is a practical strength. However, the headline result depends entirely on the MILP comparison, which is not credibly specified, and the objective function in Eq. (53) contains a sign error that would reward degradation. If these issues are corrected, the work could be a valuable contribution to BEMS scheduling with heterogeneous storage; in its current form, the central claim is unsupported.

major comments (3)
  1. [Section 6.2.4] The MILP baseline is not described beyond stating that it was implemented with CPLEX in Python. No formulation is given for the objective function, decision variables, constraints, treatment of degradation costs, or use of forecast information. For a finite-horizon discrete-action problem (25 actions per hour) with deterministic test data, a correctly specified MILP should match or beat a learned policy. The reported result that MILP is 13.69%-21.60% more expensive than DQN is therefore implausible and strongly suggests the MILP was solving a different or mis-specified problem. Without the MILP formulation, the claimed 37.94%-40.01% cost reduction is not credible.
  2. [Eq. (53)] The objective in Eq. (53) subtracts C_build, defined in Eq. (51) as a positive battery degradation cost borne by the building. This sign error would reward degradation costs, making the optimization problem incorrect. If the MILP was implemented with this sign, its poor performance is fully explained. The sign must be corrected and all cost comparisons in Fig. 13 re-run before the results can be interpreted.
  3. [Section 6.1, Section 6.2.4] Each season is evaluated on a single training run; the paper reports no random seeds, multiple independent runs, confidence intervals, or statistical significance tests. DRL training is stochastic, so the reported cost figures (including the headline reduction and the differences among DQN, D3QN, and the proposed method) could be within run-to-run noise. The authors should report mean and standard deviation over at least several seeds, and ideally provide the code or data to enable reproducibility.
minor comments (4)
  1. [Section 5.1.2] The section title contains a duplicated word: "Double dueling DQN with with prioritized experience replay" should be "Double dueling DQN with prioritized experience replay."
  2. [Section 3.3.2] There is a typo in the sentence "integrating EVs into the into the BEMS"; the duplicated phrase should be removed.
  3. [Fig. 13] The y-axis label in Fig. 13 says "Cumulative profits" while the text and captions refer to "cumulative costs"; these labels should be made consistent.
  4. [References] References [26] and [34] are duplicates, both being A. Buonomano, "Building to vehicle to building concept...", Applied Energy 261 (2020) 114077; one should be removed or the two references should be differentiated.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the cost reduction is an empirical simulation result with disclosed reward shaping, and the self-cited predictor is re-benchmarked in this paper.

full rationale

The central claim of a 37.94%-40.01% cost reduction is obtained by simulating the learned policy on held-out 2024 summer/winter data and comparing cumulative operating costs (Figs. 13, Eq. 53) against DQN, D3QN, MILP, and uncontrolled baselines; it is not a quantity fitted from the degradation coefficients or forecast errors. The reward function (Eqs. 61-69) does contain hand-designed terms that explicitly reward valley charging and peak discharging and penalize low-price EV cycling, and Section 6.2.3 partly describes behavior the reward was designed to produce; however, the paper discloses this design choice, and the headline cost saving is still measured under the actual cost objective rather than being definitionally equal to the reward. Degradation-model constants are taken from external semi-empirical studies [31][46], and the RDedRVFL predictor comes from the authors' prior work [2]; the predictor is not merely cited but re-implemented and benchmarked against ARIMA and LSTM on this dataset (Tables 5-6), so the self-citation is independent empirical support. The main weaknesses are correctness and reproducibility concerns rather than circularity: the MILP baseline is not specified with an objective, decision variables, and constraint set, and Eq. (53) appears to have a sign error in the treatment of C_build (the minimization subtracts the degradation cost instead of adding it). These issues mean the MILP comparison may be unfair or the cost metric mis-specified, but they do not make the derivation equivalent to its own inputs. No load-bearing step reduces to a fitted parameter renamed as a prediction, and no uniqueness claim is imported from the authors' prior work.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The paper leans on many imported degradation constants and hand-designed reward coefficients. The central scheduling result is therefore not derived from first principles; it is a policy learned inside a simulation whose cost structure is partly prescribed by the authors. The MILP comparison, which is the empirical anchor of the paper, is not described in enough detail to count as an independent benchmark.

free parameters (6)
  • Reward-shaping coefficient w_dis = Piecewise exponential based on phi_pri and phi_net (Eqs. 61-64)
    Hand-designed to amplify rewards for discharging at high price and high net load and to penalize otherwise; this encodes the strategy the paper later presents as learned.
  • Reward-shaping coefficient w_ch = Piecewise exponential based on phi_pri and phi_net (Eqs. 61-64)
    Hand-designed to amplify charging rewards at low price and low net load, penalizing otherwise; the agent's behavior is steered by this term.
  • EV degradation weight w_EV = theta_EV/theta_base with theta_scale = 0.916 (Eqs. 67-69)
    Hand-tuned to make EV scheduling costlier at low prices, injecting expert knowledge about when EVs should be used.
  • Initial degradation coefficients alpha_ESS_d and alpha_EV_d = 0.35 and 0.45 (Table 2)
    Initial per-kWh cycle aging costs; later updated from simulation history, so the first episodes depend on these arbitrary starting values.
  • Degradation model parameters for LFP and NMC = k_alpha, k_beta, k_gamma, k_delta1-3, k_T, k_t, alpha_SEI, beta_SEI, k_z (Table 3)
    Empirical constants imported from cited battery aging studies [31,46]; no degradation experiments are performed here, and the cost allocations depend on them.
  • Price coefficient beta_pri = 0.9 (Eq. 34, Table 2)
    Sets sell-back price as 90 percent of purchase price; directly affects arbitrage potential and the EAM ablation results.
assumptions (6)
  • domain assumption Day-ahead electricity price series is known 24 hours in advance
    State vector contains c_b(t) through c_b(t+23) (Section 4.2), and scheduling exploits this look-ahead.
  • domain assumption The load and PV forecasts from RDedRVFL are sufficiently accurate to guide the agent
    Predicted net load is included in the state (Section 4.2) and is essential for the claimed long-term strategy; if forecasts are poor, cost benefits shrink.
  • domain assumption The semi-empirical degradation models from [31,46] are valid for this scheduling task
    Eqs. (35)-(42) are imported without experimental validation here, yet they set the degradation costs in the reward and objective.
  • domain assumption EVs are treated as a single aggregate battery with a departure SoC at least as high as arrival SoC
    Eqs. (11)-(15) model the fleet as one entity and impose SoC(arrival) <= SoC(departure), which simplifies individual travel patterns.
  • ad hoc to paper The building manager pays EV cycle degradation costs and EV users bear only calendar aging
    Cost allocation in Eqs. (49)-(51) is a stakeholder assumption chosen by the authors; it directly changes which actions are economically rewarded.
  • domain assumption The environment is stationary enough for a policy trained on 2023 data to transfer to 2024 test periods
    Training and test splits use the same seasons in different years (Section 6.1), assuming similar load, PV, and price patterns.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Deep reinforcement learning-based joint real-time energy scheduling for green buildings with heterogeneous battery energy storage devices." pith.science (2026). https://pith.science/paper/VPVYQTAQ

@misc{pith2026250606824,
  author       = {Pith},
  title        = {Pith review of: Deep reinforcement learning-based joint real-time energy scheduling for green buildings with heterogeneous battery energy storage devices},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VPVYQTAQ}},
  note         = {Machine review of arXiv:2506.06824}
}
read the original abstract

Green buildings (GBs) with renewable energy and building energy management systems (BEMS) enable efficient energy use and support sustainable development. Electric vehicles (EVs), as flexible storage resources, enhance system flexibility when integrated with stationary energy storage systems (ESS) for real-time scheduling. However, differing degradation and operational characteristics of ESS and EVs complicate scheduling strategies. This paper proposes a model-free deep reinforcement learning (DRL) method for joint real-time scheduling based on a combined battery system (CBS) integrating ESS and EVs. We develop accurate degradation models and cost estimates, prioritize EV travel demands, and enable collaborative ESS-EV operation under varying conditions. A prediction model optimizes energy interaction between CBS and BEMS. To address heterogeneous states, action coupling, and learning efficiency, the DRL algorithm incorporates double networks, a dueling mechanism, and prioritized experience replay. Experiments show a 37.94 percent to 40.01 percent reduction in operating costs compared to a mixed-integer linear programming (MILP) approach.

Figures

Figures reproduced from arXiv: 2506.06824 by the authors.

Figure 1
Figure 1. The framework of the system model. 0 200 400 600 800 1000 Time (days) 99.4 99.5 99.6 99.7 99.8 99.9 100 Remaining capacity (%) LFP - SoC=40%, T=45°C LFP - SoC=50%, T=35°C LFP - SoC=60%, T=35°C LFP - SoC=60%, T=25°C NMC - SoC=40%, T=45°C NMC - SoC=50%, T=35°C NMC - SoC=60%, T=35°C NMC - SoC=60%, T=25°C (a) 0 1000 2000 3000 4000 5000 6000 Cycle number 60 65 70 75 80 85 90 95 100 Remaining capacity (%) LFP - DoD=40%, T… view at source ↗
Figure 2
Figure 2. The degradation characteristics of LFP and NMC lithium-ion batteries under various working conditions. (a) Comparison of calendar aging characteristics of LFP and NMC at different SoC and temperatures; (b) Comparison of the cycle aging characteristics of LFP and NMC at 50% SoC and different DoD and temperatures. First Author et al.: Preprint submitted to Elsevier Page 37 of 36 [PITH_FULL_IMAGE:figures/full_fig_p037… view at source ↗
Figure 3
Figure 3. Multi-battery degradation assessment framework in DRL-based real-time energy scheduling. -2 3 3 0 2 2 2 1 1 4 0 0 -1 0 1 2 3 (a) -3 3 3 -1 2 2 1 1 1 3 0 0 -2 -1 0 1 2 (b) [PITH_FULL_IMAGE:figures/full_fig_p038_3.png] view at source ↗
Figures from the paper (19 more)
Figure 4
Figure 4. Figure 4: The design principles of scaling coefficients for discharging and charging. (a) Discharging scaling coefficient; (b) Charging scaling coefficient. First Author et al.: Preprint submitted to Elsevier Page 38 of 36 [PITH_FULL_IMAGE:figures/full_fig_p038_4.png]
Figure 5
Figure 5. Figure 5: The design principles of scaling coefficients for the degradation cost of ESS and EV. DQN agent Scheduling decision max ( , ), 1 ,1 a arg Q s a X aX         max ( , ), 1 ,1 a arg Q s a X aX          a  a Control decision Environment X   1 …
Figure 6
Figure 6. Figure 6: Energy scheduling decision-making based on DQN. First Author et al.: Preprint submitted to Elsevier Page 39 of 36 [PITH_FULL_IMAGE:figures/full_fig_p039_6.png]
Figure 7
Figure 7. Figure 7: The framework of proposed energy scheduling method. First Author et al.: Preprint submitted to Elsevier Page 40 of 36 [PITH_FULL_IMAGE:figures/full_fig_p040_7.png]
Figure 8
Figure 8. Figure 8: Electricity prices in different seasons. (a) Summer; (b) Winter. First Author et al.: Preprint submitted to Elsevier Page 41 of 36 [PITH_FULL_IMAGE:figures/full_fig_p041_8.png]
Figure 9
Figure 9. Figure 9: The prediction results of building load and PV generation. (a) The building load prediction curve; (b) The PV generation prediction curve. 0 2000 4000 6000 8000 10000 Episode -5 0 5 10 15 20 Episode reward DQN D3QN Proposed (a) 0 2000 4000 6000 8000 10000 Episode -5 0 …
Figure 10
Figure 10. Figure 10: The episode reward during the training process for DRL-based models in different seasons. (a) Summer; (b) Winter. First Author et al.: Preprint submitted to Elsevier Page 42 of 36 [PITH_FULL_IMAGE:figures/full_fig_p042_10.png]
Figure 11
Figure 11. Figure 11: The charging/discharging control results of the proposed method in summer. (a) The blue bar and orange bar represent the charging(-)/discharging(+) control actions of EV and ESS, the green line and yellow line denote the electricity price and net load. (b) The blue ba…
Figure 12
Figure 12. Figure 12: The charging/discharging control results of the proposed method in winter. (a) The blue bar and orange bar represent the charging(-)/discharging(+) control actions of EV and ESS, the green line and yellow line denote the electricity price and net load. (b) The blue ba…
Figure 13
Figure 13. Figure 13: Comparison results of cumulative profits for all methods in different seasons. (a) Summer; (b) Winter. First Author et al.: Preprint submitted to Elsevier Page 44 of 36 [PITH_FULL_IMAGE:figures/full_fig_p044_13.png]
Figure 14
Figure 14. Figure 14: Comparison of operating costs without prediction mechanism in different seasons. (a) Summer; (b) Winter. 0 8 16 24 32 40 48 Time (hours) -1 -0.5 0 0.5 1 Price (¥) / Net load (kW) -1 -0.5 0 0.5 1 Charging/Discharging (kW) Price Net load ESS action EV action (a) 0 8 16 …
Figure 15
Figure 15. Figure 15: The charging/discharging control results of the proposed method without prediction in summer. (a) The blue bar and orange bar represent the charging(-)/discharging(+) control actions of EV and ESS, the green line and yellow line denote the electricity price and net lo…
Figure 16
Figure 16. Figure 16: Comparison of operating costs without EAM in different seasons. (a) Summer; (b) Winter. 0 50 100 150 200 250 300 Time (hours) 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 Cumulative cost (¥) 104 (a) 0 50 100 150 200 250 300 Time (hours) 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 Cumulative cost …
Figure 17
Figure 17. Figure 17: Comparison of operating costs without CBS in different seasons. (a) Summer; (b) Winter. First Author et al.: Preprint submitted to Elsevier Page 46 of 36 [PITH_FULL_IMAGE:figures/full_fig_p046_17.png]
Figure 18
Figure 18. Figure 18: Comparison of operating costs without EV in different seasons. (a) Summer; (b) Winter. 0 50 100 150 200 250 300 Time (hours) 0 0.5 1 1.5 2 2.5 3 3.5 Cumulative cost (¥) 104 Without degradation With degradation (a) 0 50 100 150 200 250 300 Time (hours) 0 0.5 1 1.5 2 2.…
Figure 19
Figure 19. Figure 19: Comparison of operating costs without considering battery degradation in different seasons. (a) Summer; (b) Winter. First Author et al.: Preprint submitted to Elsevier Page 47 of 36 [PITH_FULL_IMAGE:figures/full_fig_p047_19.png]
Figure 20
Figure 20. Figure 20: The cumulative scheduling power and cost of batteries in different seasons. (a) Summer; (b) Winter. 0 50 100 150 200 250 300 Time (hours) 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Cumulative cost (¥) (a) 0 50 100 150 200 250 300 Time (hours) 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.6…
Figure 21
Figure 21. Figure 21: The evolution of battery degradation coefficient in different seasons. (a) Summer; (b) Winter. First Author et al.: Preprint submitted to Elsevier Page 48 of 36 [PITH_FULL_IMAGE:figures/full_fig_p048_21.png]
Figure 22
Figure 22. Figure 22: The evolution curve of battery SoH in different seasons. (a) Summer; (b) Winter. First Author et al.: Preprint submitted to Elsevier Page 49 of 36 [PITH_FULL_IMAGE:figures/full_fig_p049_22.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

55 extracted references · 53 canonical work pages

  1. [1]

    Hwang, J

    J. Hwang, J. Kim, S. Yoon, Dt-bems: Digital twin-enabled building energy management system for information fusion and energy efficiency, Energy (2025) 136162

    work page 2025

  2. [2]

    C. Liu, Z. Xu, M. Yuan, J. Xie, Y. Yuan, K. Ma, Building electrical load forecasting with occupancy data based on wireless sensing, Applied Energy 380 (2025) 124960

    work page 2025

  3. [3]

    Z.Yadollahi,R.Gharibi,R.Dashti,A.T.Jahromi,Optimalenergymanagementofenergyhub:Areinforcementlearningapproach,Sustainable Cities and Society 102 (2024) 105179

    work page 2024

  4. [4]

    A. E. Onile, J. Belikov, Y. Levron, E. Petlenkov, Energy efficient behavior modeling for demand side recommender system in solar microgrid applications using multi-agent reinforcement learning model, Sustainable Cities and Society 90 (2023) 104392

    work page 2023

  5. [5]

    X. Zhai, Z. Li, Z. Li, Y. Xue, X. Chang, J. Su, X. Jin, P. Wang, H. Sun, Risk-averse energy management for integrated electricity and heat systems considering building heating vertical imbalance: An asynchronous decentralized approach, Applied Energy 383 (2025) 125271

    work page 2025

  6. [6]

    Tavana, M

    M. Tavana, M. Izadikhah, R. Farzipoor Saen, R. Zare, An integrated data envelopment analysis and life cycle assessment method for performance measurement in green construction management, Environmental Science and Pollution Research 28 (2021) 664–682

    work page 2021

  7. [7]

    S.-H.Yoon,S.-Y.Kim,G.-H.Park,Y.-K.Kim,C.-H.Cho,B.-H.Park,Multiplepower-basedbuildingenergymanagementsystemforefficient management of building energy, Sustainable Cities and Society 42 (2018) 462–470

    work page 2018

  8. [8]

    G. Zhao, H. Yang, C. Jiang, Direct seamless transfer strategy by limiting voltage drop of interface filter in energy storage equipment of microgrid, IEEE Transactions on Industrial Electronics 67 (12) (2020) 10421–10432.doi:10.1109/TIE.2019.2958296

    arXiv 2020

Show all 55 references
  1. [9]

    M. A. Ortega-Vazquez, Optimal scheduling of electric vehicle charging and vehicle-to-grid services at household level including battery degradation and price uncertainty, IET Generation, Transmission & Distribution 8 (6) (2014) 1007–1016. First Author et al.:Preprint submitted...

    2014

  2. [10]

    Z. Wan, H. Li, H. He, D. Prokhorov, Model-free real-time ev charging scheduling based on deep reinforcement learning, IEEE Transactions on Smart Grid 10 (5) (2018) 5246–5257

    2018

  3. [11]

    M. H. K. Tushar, A. W. Zeineddine, C. Assi, Demand-side management by regulating charging and discharging of the ev, ess, and utilizing renewable energy, IEEE Transactions on Industrial Informatics 14 (1) (2018) 117–126.doi:10.1109/TII.2017.2755465

    2018

  4. [12]

    O.Erdinc,N.G.Paterakis,T.D.P.Mendes,A.G.Bakirtzis,J.P.S.Catalão,Smarthouseholdoperationconsideringbi-directionalevandess utilization by real-time pricing-based dr, IEEE Transactions on Smart Grid 6 (3) (2015) 1281–1291.doi:10.1109/TSG.2014.2352650

    2015

  5. [13]

    Zhang, G

    H. Zhang, G. Zhang, M. Zhao, Y. Liu, Load forecasting-based learning system for energy management with battery degradation estimation: A deep reinforcement learning approach, IEEE Transactions on Consumer Electronics (2024)

    2024

  6. [14]

    J.Cao,D.Harrold,Z.Fan,T.Morstyn,D.Healey,K.Li,Deepreinforcementlearning-basedenergystoragearbitragewithaccuratelithium-ion battery degradation model, IEEE Transactions on Smart Grid 11 (5) (2020) 4513–4521

    2020

  7. [15]

    Thomas, O

    D. Thomas, O. Deblecker, C. S. Ioakimidis, Optimal operation of an energy management system for a grid-connected smart building considering photovoltaics’ uncertainty and stochastic electric vehicles’ driving schedule, Applied Energy 210 (2018) 1188–1206

    2018

  8. [16]

    Z. Shen, C. Wu, L. Wang, G. Zhang, Real-time energy management for microgrid with ev station and chp generation, IEEE Transactions on Network Science and Engineering 8 (2) (2021) 1492–1501

    2021

  9. [17]

    R.Li,L.Bao,L.Chen,C.Zha,J.Dong,N.Qi,R.Tang,Y.Lu,M.Wang,R.Huang,etal.,Acceleratedagingoflithium-ionbatteries:bridging battery aging analysis and operational lifetime prediction, Science bulletin 68 (23) (2023) 3055–3079

    2023

  10. [18]

    Olmos, I

    J. Olmos, I. Gandiaga, A. Saez-de Ibarra, X. Larrea, T. Nieva, I. Aizpuru, Modelling the cycling degradation of li-ion batteries: Chemistry influenced stress factors, Journal of Energy Storage 40 (2021) 102765

    2021

  11. [19]

    Y.Yang,Y.Ye,Z.Cheng,G.Ruan,Q.Lu,X.Wang,H.Zhong,Lifecycleeconomicviabilityanalysisofbatterystorageinelectricitymarket, Journal of Energy Storage 70 (2023) 107800

    2023

  12. [20]

    J. Wen, D. Zhao, C. Zhang, An overview of electricity powered vehicles: Lithium-ion battery energy storage density and energy conversion efficiency, Renewable Energy 162 (2020) 1629–1648

    2020

  13. [21]

    H. Lv, X. Zhu, J. Mei, Y. Xia, B. Wang, Recent progress of in-situ characterization of lini1- x- y co x mn y o2 cathodes for lithium metal batteries: A mini review, Nano Research 17 (3) (2024) 1384–1401

    2024

  14. [22]

    C. Hu, M. Geng, H. Yang, M. Fan, Z. Sun, R. Yu, B. Wei, A review of capacity fade mechanism and promotion strategies for lithium iron phosphate batteries, Coatings 14 (7) (2024) 832

    2024

  15. [23]

    Harnischmacher, L

    C. Harnischmacher, L. Markefke, A. B. Brendel, L. Kolbe, Two-sided sustainability: Simulating battery degradation in vehicle to grid applications within autonomous electric port transportation, Journal of Cleaner Production 384 (2023) 135598

    2023

  16. [24]

    Zhang, J

    J. Zhang, J. Lee, A review on prognostics and health monitoring of li-ion battery, Journal of power sources 196 (15) (2011) 6007–6014

    2011

  17. [25]

    Q.Wu,M.Shahidehpour,C.Li,S.Huang,W.Wei,etal.,Transactivereal-timeelectricvehiclechargingmanagementforcommercialbuildings with pv on-site generation, IEEE Transactions on Smart Grid 10 (5) (2018) 4939–4950

    2018

  18. [27]

    C. Ma, W. Dong, Y. Shi, F. Zhang, Q. Yang, A fuzzy logic dispatching method based on model predictive control for adaptively addressing uncertain operation scenarios in multi-energy systems, Energy 322 (2025) 135724

    2025

  19. [28]

    F. A. Kassab, B. Celik, F. Locment, M. Sechilariu, S. Liaquat, T. M. Hansen, Optimal sizing and energy management of a microgrid: A joint milp approach for minimization of energy cost and carbon emission, Renewable Energy 224 (2024) 120186

    2024

  20. [29]

    R. Z. Homod, H. I. Mohammed, A. M. Sadeq, B. N. Alhasnawi, A. W. Al-Fatlawi, A. Al-Manea, O. A. Alawi, A. Alahmer, J. M. Mahdi, W.Al-Kouz,etal.,Massiveenergyreductionandstoragecapacityrelativetopcmphysicalsizebyintegratingdeeprlclusteringandmulti-stage strategies into smart bu...

    2025

  21. [30]

    R.Hamdi,E.Baccour,A.Erbad,M.Qaraqe,M.Hamdi,Lora-rl:Deepreinforcementlearningforresourcemanagementinhybridenergylora wireless networks, IEEE Internet of Things Journal 9 (9) (2021) 6458–6476

    2021

  22. [31]

    B. Xu, A. Oudalov, A. Ulbig, G. Andersson, D. S. Kirschen, Modeling of lithium-ion battery degradation for cell life assessment, IEEE Transactions on Smart Grid 9 (2) (2016) 1131–1140

    2016

  23. [32]

    Y. Bai, W. Zhang, X. Hu, A collaborative matching method for multi-energy supply systems in office buildings considering the random characteristics of electric vehicles, Energy and Buildings 303 (2024) 113809

    2024

  24. [33]

    Mohammad, R

    A. Mohammad, R. Zamora, T. T. Lie, Transactive energy management of pv-based ev integrated parking lots, IEEE Systems Journal 15 (4) (2020) 5674–5682

    2020

  25. [34]

    Buonomano, Building to vehicle to building concept: A comprehensive parametric and sensitivity analysis for decision making aims, Applied Energy 261 (2020) 114077

    A. Buonomano, Building to vehicle to building concept: A comprehensive parametric and sensitivity analysis for decision making aims, Applied Energy 261 (2020) 114077

    2020

  26. [35]

    Y. Wang, D. Xie, P. Zhao, C. Gu, X. Wang, A coordinated optimization strategy for charging station siting and ev dispatch based on response costs: A case study of chicago, Applied Energy 389 (2025) 125791

    2025

  27. [36]

    K. M. Tan, S. Padmanaban, J. Y. Yong, V. K. Ramachandaramurthy, A multi-control vehicle-to-grid charger with bi-directional active and reactive power capabilities for power grid support, Energy 171 (2019) 1150–1163

    2019

  28. [37]

    Liu, X.-F

    Z.-F. Liu, X.-F. Luo, X.-X. Hou, J.-L. Yu, J.-X. Li, Generalized energy pool-driven regional integrated energy system dispatch considering multi-time scale synergy carbon-storage game, Renewable and Sustainable Energy Reviews 217 (2025) 115752

    2025

  29. [38]

    Chen, I.-Y

    W.-H. Chen, I.-Y. L. Hsieh, Techno-economic analysis of lithium-ion battery price reduction considering carbon footprint based on life cycle assessment, Journal of Cleaner Production 425 (2023) 139045

    2023

  30. [39]

    C. Wu, H. Han, S. Gao, Y. Liu, Coordinated scheduling for multimicrogrid systems considering mobile energy storage characteristics of electric vehicles, IEEE Transactions on Transportation Electrification 9 (1) (2022) 1775–1783

    2022

  31. [40]

    First Author et al.:Preprint submitted to ElsevierPage 35 of 36 Short Title of the Article

    H.Farzin,M.Fotuhi-Firuzabad,M.Moeini-Aghtaie,Apracticalschemetoinvolvedegradationcostoflithium-ionbatteriesinvehicle-to-grid applications, ieee transactions on sustainable energy 7 (4) (2016) 1730–1738. First Author et al.:Preprint submitted to ElsevierPage 35 of 36 Short Titl...

    2016

  32. [41]

    Y. Wu, X. Huang, L. Huang, J. Chen, Strategies for rational design of high-power lithium-ion batteries, Energy & Environmental Materials 4 (1) (2021) 19–45

    2021

  33. [42]

    Redondo-Iglesias, P

    E. Redondo-Iglesias, P. Venet, S. Pelissier, Efficiency degradation model of lithium-ion batteries for electric vehicles, IEEE Transactions on Industry Applications 55 (2) (2018) 1932–1940

    2018

  34. [43]

    Xiong, P

    R. Xiong, P. Wang, Y. Jia, W. Shen, F. Sun, Multi-factor aging in lithium iron phosphate batteries: Mechanisms and insights, Applied Energy 382 (2025) 125250

    2025

  35. [44]

    Kassem, C

    M. Kassem, C. Delacourt, Postmortem analysis of calendar-aged graphite/lifepo4 cells, Journal of Power Sources 235 (2013) 159–171

    2013

  36. [45]

    E. M. Krieger, J. Cannarella, C. B. Arnold, A comparison of lead-acid and lithium-based battery behavior and capacity fade in off-grid renewable charging applications, Energy 60 (2013) 492–500

    2013

  37. [46]

    Nájera, J

    J. Nájera, J. Arribas, R. De Castro, C. Núñez, Semi-empirical ageing model for lfp and nmc li-ion battery chemistries, Journal of Energy Storage 72 (2023) 108016

    2023

  38. [47]

    J.Wang,P.Liu,J.Hicks-Garner,E.Sherman,S.Soukiazian,M.Verbrugge,H.Tataria,J.Musser,P.Finamore,Cycle-lifemodelforgraphite- lifepo4 cells, Journal of power sources 196 (8) (2011) 3942–3948

    2011

  39. [48]

    B. Xu, J. Zhao, T. Zheng, E. Litvinov, D. S. Kirschen, Factoring the cycle aging cost of batteries participating in electricity markets, IEEE Transactions on Power Systems 33 (2) (2017) 2248–2259

    2017

  40. [49]

    B.Xu,Y.Shi,D.S.Kirschen,B.Zhang,Optimalbatteryparticipationinfrequencyregulationmarkets,IEEETransactionsonPowerSystems 33 (6) (2018) 6715–6725

    2018

  41. [50]

    W. Zhao, J. Zheng, L. Zou, H. Jia, B. Liu, H. Wang, M. H. Engelhard, C. Wang, W. Xu, Y. Yang, et al., High voltage operation of ni-rich nmc cathodes enabled by stable electrode/electrolyte interphases, Advanced energy materials 8 (19) (2018) 1800297

    2018

  42. [51]

    Z. Yang, H. Charalambous, S. E. Trask, A. Montoya, A. Jansen, K. M. Wiaderek, I. Bloom, Extreme fast charge aging: Effect of electrode loading and nmc composition on inhomogeneous degradation in graphite bulk and electrode/electrolyte interface, Journal of Power Sources 549 (2...

    2022

  43. [52]

    A.Chiş,J.Lundén,V.Koivunen,Reinforcementlearning-basedplug-inelectricvehiclechargingwithforecastedprice,IEEETransactionson Vehicular Technology 66 (5) (2016) 3674–3684

    2016

  44. [53]

    R. S. Sutton, A. G. Barto, Reinforcement Learning: An Introduction, 2nd Edition, The MIT Press, 2018. URLhttp://incompleteideas.net/book/the-book-2nd.html

    2018

  45. [54]

    V. Mnih, K. Kavukcuoglu, D. Silver, A. A. Rusu, J. Veness, M. G. Bellemare, A. Graves, M. Riedmiller, A. K. Fidjeland, G. Ostrovski, et al., Human-level control through deep reinforcement learning, nature 518 (7540) (2015) 529–533

    2015

  46. [55]

    X.Xu,L.Guan,Z.Wang,R.Yao,X.Guan,Adouble-layerforecastingmodelforpvpowerforecastingbasedongru-informer-svrandblending ensemble learning framework, Applied Soft Computing (2025)

    2025

  47. [56]

    R. Gao, L. Du, P. N. Suganthan, Q. Zhou, K. F. Yuen, Random vector functional link neural network based ensemble deep learning for short-term load forecasting, Expert Systems with Applications 206 (2022) 117784. First Author et al.:Preprint submitted to ElsevierPage 36 of 36 S...

    2022

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.