Feature comparison

This page places ev-flow alongside the open-source EV charging generators and simulators it relates to most directly. It is a companion to When to use ev-flow: use that page to choose a tool, and this one to check capabilities cell by cell.

ev-flow on PyPI, pev_synth in Python

You pip install ev-flow and import pev_synth — one project, two names.

How to read the table

Yes = first-class, documented, callable without re-implementing it. Partial = supported with user wiring, or only along part of the dimension. No = not supported. — = not a stated capability of that tool in its published descriptor; treat as "not claimed here, see that tool's own docs" rather than as a negative we assert. Cells are grounded in the ev-flow paper's related-work survey and the differentiation notes; we do not invent capabilities of peer tools.

Tools compared

Tool	Primary role	Reference
ev-flow (pev_synth)	Generator — NHTS-grounded, per-vehicle	This package; Travacca 2026
emobpy	Generator — mobility → charging demand	Gaete-Morales et al. 2021
VencoPy	Flexibility-envelope tool from travel survey	Wulff et al. 2021
RAMP-mobility	Bottom-up stochastic EV load profiles	Mangipinto et al. 2022
datafev	Scenario generator + scheduler benchmark	Gümrükcü et al. 2023
ACN-Sim	Per-EV simulator on real workplace data	Lee et al. 2021

SimBEV, SpiceEV, EV-SDG, and SPEECh are discussed in the notes below; they are named in the paper's related work but omitted from the table to keep it legible (SpiceEV and ACN-Sim are simulators rather than generators; EV-SDG is session-only; SPEECh is an aggregate-demand model whose K=16 parameters ev-flow itself consumes).

Capability matrix

Capability	ev-flow	emobpy	VencoPy	RAMP-mobility	datafev	ACN-Sim
Regional grounding	8 U.S. regions (NHTS)	European (German)	European (MiD)	European countries	Parametric	Caltech/JPL facilities
Survey-microdata basis	Yes (NHTS 2017)	Partial (empirical stats)	Yes (MiD survey)	Partial (stochastic, survey-informed)	No (parametric specs)	No (real ACN telemetry)
Output: per-vehicle profiles	Yes	Yes	No (flexibility envelopes)	No (aggregate load)	Partial (sessions/scenarios)	Yes
Output: aggregate / flexibility	Yes (load curve)	Partial	Yes (envelopes)	Yes (load)	Partial	Partial
Residential charging	Yes	Yes	Yes	Yes	Partial	No
Workplace charging	Yes (cohort caveat)	Yes	Partial	—	Partial	Yes
Public / DC-fast charging	No	Yes	—	—	Partial	No
Continuous-time SoC trajectory	Yes	Yes	Partial	—	No	Yes
Charging-session records	Yes	Partial	Partial	—	Yes	Yes
15-min / hourly raster	Yes	Yes (configurable)	Partial	Yes	Partial	Yes
Timezone-aware (UTC-canonical)	Yes	Partial	—	—	—	Partial
Charging-control / strategy modeling	No (by design)	No	No	No	Yes (scheduler lib)	Yes (algorithms)
Vehicle-to-grid / bidirectional	No	No	No	No	—	Yes
Validation runner vs published bounds	Yes (classified)	No	Partial (paper-only)	—	No	—
Reproducible from a single seed	Yes	Partial	—	Partial	—	—
Distributed on PyPI	Yes (ev-flow)	Yes	Yes	Yes	Yes	Yes (acnportal)
License	MIT	Open (see project)	Open (see project)	Open (see project)	Open (see project)	Open (see project)

Honest caveat

This matrix reflects the ev-flow author's reading of each peer tool's published descriptor and documentation, not exhaustive hands-on testing of every cell. Peer projects evolve; a — means the capability is not a claimed feature in the sources we consulted, not that it is impossible. Where you depend on a specific capability of another tool, confirm it against that tool's current docs. ev-flow is not universally superior: on several axes a peer tool is the more mature choice (see below).

Where peers lead ev-flow

Being candid about non-overlap, the open ecosystem is ahead of ev-flow on several axes:

• Non-U.S. coverage. emobpy, VencoPy, RAMP-mobility, and SimBEV are purpose-built for European mobility surveys; ev-flow is U.S.-only and NHTS-2017-bound.
• Charging-strategy libraries. datafev and SpiceEV ship non-trivial scheduling algorithms on top of the data; ev-flow ships none and delegates control entirely.
• Charging physics & V2G. ACN-Sim models the charger/battery interaction at higher fidelity (taper, three-phase) and supports bidirectional flow; ev-flow uses a deliberately simple, one-way, constant-current SoC ledger.
• Real-telemetry replay. ACN-Sim is built around a real workplace dataset; ev-flow is synthetic by design.

What ev-flow uniquely combines

No single peer tool offers all of the following in one open, MIT-licensed, pip install-able package:

1. NHTS-microdata donor-stitched travel weeks (not parametric sampling over aggregate statistics);
2. per-vehicle Profile objects unifying trips, plug status, sessions, and SoC, indexable by ev_id, with a Fleet aggregation API;
3. eight pre-baked U.S. regions spanning CAISO / NYISO / ISO-NE / PJM / ERCOT / WECC, each with its own CBSAs, time zone, sales-mix source, and winter uplift;
4. residential and workplace profile types;
5. UTC-canonical storage with tz-aware query, DST-verified; and
6. an honestly-classified validation runner — the reference bay_area residential profile rolls up to 11 PASS / 0 FAIL / 6 EXPLAINED_FAIL / 4 EXPLAINED_SKIP / 0 ERROR, with every non-PASS row tied to a sourced reason.

The contribution is the union, not dominance on any one column. See When to use ev-flow for the use-case-by-use-case guidance.