pith. sign in

arxiv: 2605.18443 · v1 · pith:OSQUNHCKnew · submitted 2026-05-18 · 📡 eess.SY · cs.SY

Electric Vehicle Charging Profile Forecasting Using Hybrid Models

Riccardo Ramaschi , Mario Paolone , Sonia Leva This is my paper

Pith reviewed 2026-05-20 09:30 UTC · model grok-4.3

classification 📡 eess.SY cs.SY
keywords electric vehicle chargingcharging profile forecastinghybrid modelsfast charging stationsindividual EV forecastinformation availability
0
0 comments X

The pith

A hybrid lightweight method forecasts individual EV charging profiles before and during sessions while adapting to changing information levels.

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

The paper proposes a hybrid method for predicting how a single electric vehicle will draw power during a fast charge, both ahead of time and as the session unfolds. Individual-level forecasts like this are rarer than station-wide totals yet could let operators update overall predictions more finely and better guess when each vehicle will finish. The authors test the approach across several real EVs drawn from a public dataset and specifically measure how forecast quality shifts when the model receives different amounts of data at successive time steps.

Core claim

We propose a hybrid and lightweight method to estimate the EV charging profile before and during the charging process. Besides evaluating this method on multiple EVs from a public dataset, we also assess the impact of different level of information in the time transposition of the charging profile.

What carries the argument

Hybrid lightweight model that combines forecasting components to produce charging-profile estimates at varying stages of information availability.

If this is right

  • Supports more granular real-time updates to the station-level aggregated forecast.
  • Yields more accurate estimates of individual EV departure times.
  • Preserves forecast advantages even as the amount of known information changes during charging.

Where Pith is reading between the lines

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

  • If deployed at scale, the method could reduce reserve margins needed for grid balancing at busy charging sites.
  • The same hybrid structure might transfer to other loads whose profiles become clearer only after they start, such as heat-pump cycling.
  • Further trials with fleet data would test whether the approach still works when many vehicles share the same station.

Load-bearing premise

Varying levels of available information over time can be handled by the hybrid model without reducing forecast quality.

What would settle it

Running the hybrid model and baseline single models on a fresh set of EVs and measuring accuracy at early time steps when only partial session data is supplied would show whether the claimed benefit persists.

Figures

Figures reproduced from arXiv: 2605.18443 by Mario Paolone, Riccardo Ramaschi, Sonia Leva.

Figure 1
Figure 1. Figure 1: Overall model for charging profile estimation and its [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overall model functioning example, where [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: EMAE distribution trend at different forecast update. [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Capacity estimation model performance. 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Hour of Day 50 0 50 SoC error [%] Arrival Departure [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 7
Figure 7. Figure 7: Scenario evaluation impact on three accuracy metrics. [PITH_FULL_IMAGE:figures/full_fig_p004_7.png] view at source ↗
read the original abstract

Electric Vehicle (EV) fast charging stations require forecasting techniques both at the single charger level and aggregated level. While for the latter several models exist, forecasting individual EV charging profiles is still underexplored in literature. However, such methods may be potentially used by battery-aware scheduling, leading to a more granular update of the charging station aggregated forecast and provide a more accurate estimation of EVs departure times. Nonetheless, the variable extent of available information in time and in different settings could jeopardize these benefits. For this reason, we propose a hybrid and lightweight method to estimate the EV charging profile before and during the charging process. Besides evaluating this method on multiple EVs from a public dataset, we also assess the impact of different level of information in the time transposition of the charging profile.

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 / 2 minor

Summary. The manuscript proposes a hybrid and lightweight method to forecast individual EV charging profiles at fast charging stations both before and during the charging process. The approach is evaluated on multiple EVs drawn from a public dataset, with additional analysis of how different levels of available information affect the time-transposed charging profile forecasts. The stated motivation is to support battery-aware scheduling and improved departure-time estimation.

Significance. If the hybrid construction can be shown to preserve forecast utility under partial and time-varying inputs, the work would address an underexplored area of single-charger EV profile forecasting and could enable more granular updates to station-level predictions. The emphasis on lightweight implementation and explicit assessment of information levels is a constructive direction for practical deployment.

major comments (2)
  1. [Abstract] Abstract and motivation section: the central claim that the hybrid method maintains forecast benefits despite varying levels of available information requires a concrete adaptation mechanism (e.g., input masking, separate pre-/post-charging branches, or dynamic feature weighting). No such mechanism is described or derived, leaving open whether reduced observability degrades the profiles used for scheduling and departure-time estimation.
  2. [Evaluation] Evaluation section: the assessment of different information levels is presented as a key contribution, yet without reported error metrics, baseline comparisons, or ablation results on the public dataset it is impossible to verify that the hybrid model actually preserves accuracy under partial inputs.
minor comments (2)
  1. Clarify whether the hybrid construction combines multiple distinct models or is a single integrated architecture; the title and abstract use slightly inconsistent phrasing.
  2. Provide the exact public dataset identifier and any preprocessing steps applied to the EV charging sessions.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed review of our manuscript. We address each major comment below and indicate the revisions planned for the next version.

read point-by-point responses

  1. Referee: [Abstract] Abstract and motivation section: the central claim that the hybrid method maintains forecast benefits despite varying levels of available information requires a concrete adaptation mechanism (e.g., input masking, separate pre-/post-charging branches, or dynamic feature weighting). No such mechanism is described or derived, leaving open whether reduced observability degrades the profiles used for scheduling and departure-time estimation.

    Authors: The hybrid architecture consists of distinct pre-charging and during-charging components that operate on time-transposed profiles. These components are fed with different subsets of temporal features depending on the stage of the charging session, which provides the adaptation to varying information levels. We acknowledge that this design choice was not stated with sufficient explicitness in the abstract and motivation. In the revised manuscript we will add a concise description of the input selection process and its effect on forecast utility for scheduling and departure-time estimation. revision: yes

  2. Referee: [Evaluation] Evaluation section: the assessment of different information levels is presented as a key contribution, yet without reported error metrics, baseline comparisons, or ablation results on the public dataset it is impossible to verify that the hybrid model actually preserves accuracy under partial inputs.

    Authors: We agree that the current evaluation would be strengthened by additional quantitative evidence. While the manuscript already compares performance across information levels on the public dataset, we will expand the evaluation section to include explicit error metrics (MAE, RMSE), comparisons against standard baselines, and ablation results that isolate the effect of reduced temporal information. These additions will be reported for the same set of EVs used in the original experiments. revision: yes

Circularity Check

0 steps flagged

No circularity in proposed hybrid forecasting method

full rationale

The paper proposes a hybrid lightweight method for estimating EV charging profiles before and during charging, evaluated on a public dataset while assessing impacts of varying information levels. No equations, derivations, or model definitions are presented in the provided text that reduce any prediction or result to fitted inputs by construction, self-definition, or self-citation chains. The central claim rests on empirical evaluation of the proposed approach rather than any tautological renaming or forced prediction from parameters. This constitutes a standard methodological proposal without load-bearing circular steps.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Based solely on the abstract, no explicit free parameters, axioms, or invented entities are described; the hybrid method appears to combine existing forecasting techniques without introducing new postulated quantities.

pith-pipeline@v0.9.0 · 5655 in / 1019 out tokens · 28139 ms · 2026-05-20T09:30:20.223013+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

22 extracted references · 22 canonical work pages

  1. [1]

    Global EV Outlook 2025,

    IEA, “Global EV Outlook 2025,” Tech. Rep., 2025. [Online]. Available: https://www.iea.org/reports/global-ev-outlook-2025

    work page 2025

  2. [2]

    [Online]

    European Alternative Fuels Observatory, accessed on April 2026. [Online]. Available: https://alternative-fuels- observatory.ec.europa.eu/transport-mode/road

    work page 2026

  3. [3]

    Power consumption prediction for electric vehicle charging stations and forecasting income,

    K. C. Akshay, G. H. Grace, K. Gunasekaranet al., “Power consumption prediction for electric vehicle charging stations and forecasting income,” Scientific Reports, vol. 14, p. 6497, 2024

    work page 2024

  4. [4]

    Impact of electric vehicle charging demand on power distribution grid congestion,

    Y . Li and A. Jenn, “Impact of electric vehicle charging demand on power distribution grid congestion,”Proceedings of the National Academy of Sciences, vol. 121, no. 18, p. e2317599121, 2024

    work page 2024

  5. [5]

    Electric vehicle energy demand prediction techniques: An in-depth and critical systematic review,

    F. Marzbani, A. H. Osman, and M. S. Hassan, “Electric vehicle energy demand prediction techniques: An in-depth and critical systematic review,”IEEE Access, vol. 11, pp. 96 242–96 255, 2023

    work page 2023

  6. [6]

    Assessing the influence of electric vehicle charging profile modelling methods,

    T. Simolin, A. Rautiainen, P. J ¨arventausta, K. Rauma, and C. Rehtanz, “Assessing the influence of electric vehicle charging profile modelling methods,”IET Generation, Transmission & Distribution, vol. 16, no. 15, pp. 3027–3035, 2022

    work page 2022

  7. [7]

    User behavior clustering based method for ev charging forecast,

    A. Nespoli, E. Ogliari, and S. Leva, “User behavior clustering based method for ev charging forecast,”IEEE Access, vol. 11, pp. 6273–6283, 2023

    work page 2023

  8. [8]

    Electric vehicle battery remaining charging time estimation considering charging accuracy and charging profile prediction,

    J. Shi, M. Tian, S. Han, T.-Y . Wu, and Y . Tang, “Electric vehicle battery remaining charging time estimation considering charging accuracy and charging profile prediction,”Journal of Energy Storage, vol. 49, p. 104132, 2022

    work page 2022

  9. [9]

    A review of electric vehicle load open data and models,

    Y . Amara-Ouali, Y . Goude, P. Massart, J.-M. Poggi, and H. Yan, “A review of electric vehicle load open data and models,”Energies, vol. 14, no. 8, 2021

    work page 2021

  10. [10]

    Review of battery charger topologies, charging power levels, and infrastructure for plug-in electric and hybrid vehicles,

    M. Yilmaz and P. T. Krein, “Review of battery charger topologies, charging power levels, and infrastructure for plug-in electric and hybrid vehicles,”IEEE Transactions on Power Electronics, vol. 28, no. 5, pp. 2151–2169, 2013

    work page 2013

  11. [11]

    Deep learning predicts real-world electric vehicle direct current charging profiles and durations,

    S. Li, M. Zhang, R. Doel, and et al., “Deep learning predicts real-world electric vehicle direct current charging profiles and durations,”Nature Communications, vol. 16, p. 10921, 2025

    work page 2025

  12. [12]

    Analysis and modelling of dc fast-charging profiles of heterogeneous evs,

    T. Simolin, P. J ¨arventausta, and M. Paolone, “Analysis and modelling of dc fast-charging profiles of heterogeneous evs,”Transportation En- gineering, vol. 20, p. 100350, 2025

    work page 2025

  13. [13]

    Level 3 electric vehicle charging dataset,

    Distributed Electrical Systems Laboratory, “Level 3 electric vehicle charging dataset,” Accessed on Jan 2025. [Online]. Available: https://github.com/DESL-EPFL/Level-3-EV-charging-dataset

    work page 2025

  14. [14]

    Commu- nicational aspects in hierarchical real-time control of electric vehicle charging: Available information and its value,

    T. Simolin, K. Rauma, A. Rautiainen, and P. J ¨arventausta, “Commu- nicational aspects in hierarchical real-time control of electric vehicle charging: Available information and its value,”IET Generation, Trans- mission & Distribution, vol. 17, no. 14, pp. 3152–3167, 2023

    work page 2023

  15. [15]

    International Organization for Standardization,ISO 15118-20:2022 Road vehicles – Vehicle to grid communication interface – Part 20: 2nd generation network layer and application layer requirements, Std

    work page 2022

  16. [16]

    Empirical analysis of electric vehicle fast charging under cold temperatures,

    Y . Motoaki, W. Yi, and S. Salisbury, “Empirical analysis of electric vehicle fast charging under cold temperatures,”Energy Policy, vol. 122, pp. 162–168, 2018

    work page 2018

  17. [17]

    Grid-aware scheduling and control of electric vehicle charging stations for dispatching active distribution networks: Theory and experimental validation,

    R. K. Gupta, S. Fahmy, M. Chevron, R. Vasapollo, E. Figini, and M. Paolone, “Grid-aware scheduling and control of electric vehicle charging stations for dispatching active distribution networks: Theory and experimental validation,”IEEE Transactions on Smart Grid, vol. 16, no. 2, pp. 1575–1589, 2025

    work page 2025

  18. [18]

    Cross-validation,

    P. Refaeilzadeh, L. Tang, and H. Liu, “Cross-validation,”Encyclopedia of database systems, vol. 5, pp. 532–538, 2009

    work page 2009

  19. [19]

    Open-meteo.com weather api,

    P. Zippenfenig, “Open-meteo.com weather api,” 2023. [Online]. Available: https://open-meteo.com/

    work page 2023

  20. [20]

    Optuna: A next- generation hyperparameter optimization framework,

    T. Akiba, S. Sano, T. Yanase, T. Ohta, and M. Koyama, “Optuna: A next- generation hyperparameter optimization framework,” inProceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2019, p. 2623–2631

    work page 2019

  21. [21]

    Scikit-learn: Machine learning in python,

    F. Pedregosa et al., “Scikit-learn: Machine learning in python,”J. Mach. Learn. Res., vol. 12, p. 2825–2830, Nov. 2011

    work page 2011

  22. [22]

    Com- parison of training approaches for photovoltaic forecasts by means of machine learning,

    A. Dolara, F. Grimaccia, S. Leva, M. Mussetta, and E. Ogliari, “Com- parison of training approaches for photovoltaic forecasts by means of machine learning,”Applied Sciences, vol. 8, no. 2, 2018

    work page 2018