The model answers one question for four cars: how much energy does the car take from its battery, and from the socket, to follow a given speed trace? Every input comes from an official or peer-reviewed source (see data-sources.md). Each car's two powertrain parameters are identified from its US certification test and then checked, without further adjustment, against measurements the identification never saw: the EU type-approval figures and, for the two cars whose reports contain them, other EPA tests of the same car.
The tables below are generated by the code (tools/update_docs.py), and tests check the numbers
in the text against the data and the model, so this document cannot drift from what the code
computes.
The force the wheels must deliver is
F = f0 + f1*v + f2*v^2 + m_i*a
where f0 + f1*v + f2*v^2 is the road load measured by coastdown (EPA target coefficients,
converted to SI units with exact factors) and m_i the inertial mass. Speed traces are
sampled once per second. Over each one-second interval the car is taken to move at the mean of
the two sampled speeds and to accelerate at their difference, so the inertial work over an
interval is exactly m_i/2 * (v_end^2 - v_start^2): summed over a trace that starts and ends
at rest, it cancels exactly, so the net work at the wheels equals the road-load work. Road load
acts only while the car moves. Distance is the trapezoidal integral of speed, which reproduces
the distances the EPA states for its two cycles.
Two sums are kept separately:
W+, the work the wheels deliver while the car is driven;W-, the work they must absorb when the car slows faster than road load alone would slow it: the braking the car has to do.
Masses. The inertial mass is the test mass plus the mass-equivalent of the parts that turn with the wheels, estimated as UN GTR No. 15 allows at 3 % of the sum of the mass in running order and 25 kg (the reading Regulation (EU) 2017/1151 states; data-sources.md, section 4). On the EPA dynamometer the test mass is the equivalent test weight; in the EU it is the WLTP test mass.
Air. The EPA coefficients refer to 20 °C and 98.21 kPa (40 CFR 1066.305), WLTP road load to 1.189 kg/m³ (GTR 15, para. 3.2.9). Density is proportional to pressure over temperature; anchored on the GTR point, the EPA air has 1.1671 kg/m³, and the quadratic term is scaled by the ratio of the two densities (1.0187) whenever the car is simulated at WLTP conditions.
battery = W+ / eta_drive - k_regen * W- + P_aux * t
mains = battery / eta_charge
| Symbol | Meaning | Source |
|---|---|---|
eta_drive |
battery-to-wheel efficiency while driving | identified per car (section 3) |
k_regen |
share of the braking work W- returned to the battery |
identified per car (section 3) |
P_aux |
the car's own consumption: control units, pumps, 12 V system | 185 W (section 3) |
t |
duration of the trace | |
eta_charge |
usable battery energy over mains energy | EPA certificate, per car (section 4) |
Over a trace that starts and ends at rest, or at a steady speed, the energy taken from the
socket splits into six parts that add up exactly: air drag and rolling resistance (the
road-load work, f2 term and f0 + f1 terms), braking not recovered
((1 - k_regen) * W-), drivetrain losses (W+ * (1/eta_drive - 1)), auxiliaries and
climate (P_aux * t), and charging losses (mains - battery).
Each EPA certificate report gives the DC energy the car drew from its battery per mile on the UDDS (city) and on the highway cycle (for the EX30, the phases from which it follows), on a dynamometer set to the same road load and test weight the model uses (docs/data-sources.md, section 7). Per kilometre, for each cycle,
W+ * (1/eta_drive) - W- * k_regen = DC_measured - P_aux * t
which is linear in 1/eta_drive and k_regen. The two cycles give two equations; they are
solved exactly, and the solution is refused if it implies an efficiency above 100 % or a
regeneration coefficient outside 0 to 1. The auxiliary load is the one input not measured on
these cars: the mean of the two values Argonne measured on stationary battery electric cars
with the climate control drawing nothing (226 W and 144 W, DOE program record, 2024). Section 5
shows that choosing either measured value instead moves the results by at most 0.2 percentage
points.
| Car | Drivetrain efficiency | Regeneration coefficient | Charging efficiency |
|---|---|---|---|
| Tesla Model 3 RWD | 0.936 | 0.774 | 0.886 |
| Volvo EX30 | 0.847 | 0.906 | 0.889 |
| Polestar 2 | 0.883 | 0.890 | 0.891 |
| VW ID.4 Pro | 0.893 | 0.815 | 0.902 |
The Tesla's 0.936 lies below the 97 % a peer-reviewed study measured as the best efficiency of a 2020 Model 3's power unit (Rosenberger et al., 2024), as a cycle average must; the lowest, 0.847, is the EX30's.
These are effective values. For the EX30 and the Polestar 2 the regeneration coefficient is larger than the drivetrain efficiency, which no single component efficiency would allow: energy returning from the wheels passes through the same motor and inverter. The model has one efficiency for all operating points, while a real drivetrain is more efficient at some loads than at others; with two measurements and two unknowns, the parameters absorb that simplification. They are therefore not component efficiencies, and they are not presented as such. What makes them useful is that they predict measurements they were not fitted to, which section 5 tests.
The EPA certificate gives the usable battery energy (the total DC energy discharged in the full-discharge test, 40 CFR 600.116-12(a)(8)) and the AC energy taken from the mains to recharge it. Their ratio covers everything between the socket and the energy the car later draws: the charger, the battery's own losses, what the car consumes while charging, and the drain during the test's key-off soaks, which the regulation leaves out of the usable energy. It was measured with recharges at 208-240 V (docs/data-sources.md, section 7).
The same reports contain configurations of the same cars with other wheels, hence other road-load coefficients, and, for the EX30, two constant-speed phases at 65 mph. The model, with the parameters of section 3 unchanged, predicts them from the road load alone:
| Car | Measurement | Model (Wh/km) | Measured (Wh/km) | Difference |
|---|---|---|---|---|
| Volvo EX30 | UDDS, 18 inch wheels | 104.2 | 105.1 | -0.9 % |
| Volvo EX30 | highway, 18 inch wheels | 126.7 | 126.9 | -0.2 % |
| Volvo EX30 | 65 mph, mid-test, 19 and 20 inch wheels | 165.7 | 163.0 | +1.6 % |
| Volvo EX30 | 65 mph, end-of-test, 19 and 20 inch wheels | 165.7 | 162.4 | +2.0 % |
| Volvo EX30 | 65 mph, mid-test, 18 inch wheels | 167.8 | 165.0 | +1.7 % |
| Volvo EX30 | 65 mph, end-of-test, 18 inch wheels | 167.8 | 165.0 | +1.7 % |
| Polestar 2 | UDDS, 20 inch wheels | 111.0 | 112.7 | -1.5 % |
| Polestar 2 | highway, 20 inch wheels | 130.3 | 130.9 | -0.4 % |
No prediction is off by more than 2.1 %. The steady 65 mph points come out 1.6-2.0 % high; a likely reason is that at a steady cruise the drivetrain works nearer its best efficiency than the cycle average the model uses. The ID.4's report also has a second test in B mode, its stronger-regeneration drive mode, with the same road load; it is not a new check of the model, but it shows the size of such differences: 1.6 % less energy on the UDDS than in the default D mode used here, and 0.1 % on the highway cycle.
The EU figures come from a different test (WLTC, WLTP test mass, WLTP reference air) in a different region, and are never used to set anything. The model runs the car at its WLTP test mass on the WLTC, with the EPA road load (the EU road loads are not published). Consumption from the socket uses the EPA charging efficiency; range is the EPA usable battery energy divided by the modelled battery consumption. The official values are those carried by most 2024 registrations, with the 5-95 % range of the registrations where the cars differ:
| Car | Consumption, model (Wh/km) | Official | Difference | Range, model (km) | Official | Difference |
|---|---|---|---|---|---|---|
| Tesla Model 3 RWD | 131.3 | 132 | -0.6 % | 525 | 513 | +2.3 % |
| Volvo EX30 | 155.6 | 170 (170-171) | -8.5 % | 486 | 476 (473-476) | +2.0 % |
| Polestar 2 | 154.9 | 152 (150-157) | +1.9 % | 597 | 641 (616-649) | -6.9 % |
| VW ID.4 Pro | 158.4 | 163 (161-171) | -2.8 % | 542 | 542 (521-562) | 0.0 % |
For the Tesla and the ID.4 both comparisons agree within 3 %. For the other two cars the two comparisons disagree with each other: the EX30's range agrees (+2.0 %) but its consumption is 8.5 % below the official value; the Polestar 2's consumption agrees (+1.9 %) but its range falls 6.9 % short. Both comparisons start from the same modelled battery consumption and differ only in how it is converted: consumption through the EPA charging efficiency, range through the EPA usable battery energy. When they disagree, one of those conversions does not carry over to the EU car, which the next section examines.
UN GTR No. 15 defines a pure electric car's consumption as the energy recharged from the mains divided by its range (Annex 8, para. 4.3.2.2), so official consumption times official range is the mains energy of a full recharge in the EU test. Dividing the usable energy the EPA measured by it gives the charging efficiency the EU figures imply, if the EU and US cars store the same usable energy. This uses no model at all:
| Car | EPA, measured | Implied by the EU values |
|---|---|---|
| Tesla Model 3 RWD | 0.886 | 0.901 |
| Volvo EX30 | 0.889 | 0.830 |
| Polestar 2 | 0.891 | 0.845 |
| VW ID.4 Pro | 0.902 | 0.876 |
For the EX30 the EU figures imply clearly larger charging losses than the EPA measured. With the implied value instead of the EPA one, the modelled consumption from the socket becomes 166.6 Wh/km against the official 170: most of the gap is charging, while the range, which does not involve charging, agrees. For the Polestar 2 the implied value is also lower, but its EU range is longer than the model predicts: the EU values imply a battery consumption on the WLTC of 128.5 Wh/km against the model's 138.1. A lower EU road load or a larger usable energy in the EU car would each explain it; the public data cannot tell them apart, so the gap is reported rather than adjusted away.
Change in the EU comparison, in percentage points of the official value (consumption / range), when one modelling choice is changed and the cars are identified again where the choice affects identification:
| Change | Tesla Model 3 RWD | Volvo EX30 | Polestar 2 | VW ID.4 Pro |
|---|---|---|---|---|
| Auxiliary load 144 W instead of 185 W | +0.2 / -0.2 | +0.1 / -0.2 | +0.2 / -0.2 | +0.2 / -0.2 |
| Auxiliary load 226 W instead of 185 W | -0.2 / +0.2 | -0.1 / +0.2 | -0.2 / +0.2 | -0.2 / +0.2 |
| No rotating-mass allowance | 0.0 / 0.0 | 0.0 / 0.0 | 0.0 / 0.0 | 0.0 / 0.0 |
| No air-density correction (EPA air for the WLTC) | -0.6 / +0.7 | -0.9 / +1.0 | -0.8 / +0.7 | -0.8 / +0.8 |
| Rolling resistance rescaled to the WLTP test mass | +0.1 / -0.1 | -0.1 / +0.1 | +1.0 / -0.9 | +0.3 / -0.3 |
| One charging efficiency for all cars (0.8920) | -0.7 / 0.0 | -0.3 / 0.0 | -0.1 / 0.0 | +1.0 / 0.0 |
No choice moves any result by more than about one point, so none of the conclusions above depends on them.
A scenario runs one car, at its WLTP test mass and in WLTP reference air, on any of the cycles (WLTC or one of its phases, UDDS, highway) or at a steady speed, and may add:
- Extra load. It adds to the inertial mass, and rolling resistance is scaled in proportion
to mass: rolling resistance is
m*g*fRR, carried by the constant and linear road-load terms (Rosenberger et al., 2024, equations 2 and 3). - Climate-control power. It adds to the auxiliary load. For orientation, Argonne measured 0.8-1.0 kW on stationary cars at 35 °C, 1.8-2.9 kW at -7 °C and 3.4-3.7 kW at -18 °C to hold a 22 °C cabin, including the battery's thermal management (DOE program record, 2024, Table 2).
Range is the usable battery energy the EPA measured, divided by the battery consumption.
- One efficiency per car. The two parameters reproduce the EPA city and highway tests exactly and unseen tests within 2.1 %; far outside those conditions (very high speeds, steep accelerations) the model has not been tested.
- Weather. Climate power is the only weather input. Cold batteries also deliver less energy and have higher resistance; that is not modelled, so cold-weather ranges are optimistic.
- Road load. The EPA coastdown is used for the EU as well; the WLTP road loads are recorded on each car's certificate of conformity but are not public.
- Charging. The charging efficiency was measured in the US with recharges at 208-240 V; charging on other supplies differs (docs/data-sources.md, sections 7 and 8).
- Flat road, no wind, and the usable battery energy of the US test cars.