A pack charging time estimator built using the Electric Vehicle packs in the Battery Pack Database.
Dive deeper in the subject of charging: Battery Charging: CC-CV, Fast Charging & EV Charging Standards.
This post has been built based on the support of and sponsorship from: illumion, Custom Power, Quarto Technical Services, Eatron Technologies and About:Energy
How the Estimate Is Built
The note below the charts gives the short version. This section goes into more depth on where the curves come from, how the thermal simulation is put together, and just as importantly where the model is weakest, so you can judge how much to trust a given result.
A curve, not a flat average
The naive way to estimate charge time is energy needed ÷ peak power. That’s wrong by a wide margin, because peak power is rarely sustained for more than a narrow band of state of charge. To model the actual taper, we pulled every Electric Vehicle pack in the Battery Pack Database that had both a quoted peak charging power and a published 10–80% charge time, around 700 entries, and used the ratio between quoted peak power and the average power implied by the real charge time as the calibration target.
Two things fell out of that analysis:
- The ratio of average-to-peak power over a 10–80% window sits at roughly 0.65–0.70 across the dataset, packs simply don’t hold their headline number for long.
- The ratio isn’t constant: it correlates with a pack’s implied C-rate (peak kW ÷ usable kWh). Packs rated below about 1.5C typically sustain 75–85% of their peak; packs above 2.5C typically sustain 55–65%. Higher-C-rate designs are pushed harder relative to their own peak, so they taper away from it faster.
The calculator encodes this as two reference curves per chemistry. A “flat” curve calibrated against low-implied-C-rate packs and a “steep” curve calibrated against high-implied-C-rate packs, blended by the pack’s own peak-kW-to-kWh ratio. A 150kW pack on a 90kWh battery (1.7C) gets a noticeably flatter curve than the same 150kW on a 45kWh battery (3.3C), which matches what the database shows.
Chemistry shapes
Within that C-rate blending, chemistry sets the shape of the taper. NMC/NCA packs in the dataset consistently begin tapering earliest and hardest, typically dropping below 50% of peak by the time they cross ~55–65% state of charge. LFP and sodium-ion packs hold a higher fraction of peak power further into the charge. The trade-off is a slightly more gradual ramp from empty. LTO is the outlier: cell chemistry and its use in ultra-fast-charge applications (buses, some commercial platforms) mean it stays close to peak across almost the entire range, at the cost of energy density.
Temperature, cooling and voltage
Once the SoC-taper curve is set, the calculator layers a time-stepped thermal simulation on top, in 10-second increments. At every step, effective power is peak power × the lower of the SoC-taper fraction and a temperature-derate fraction, whichever constraint binds harder wins, matching how a BMS applies the tightest active limit rather than averaging its limits together.
The cold side of the temperature curve is fixed and generic: power is heavily restricted below roughly 10°C (lithium plating becomes a real risk at high charge current into a cold anode) and reaches full availability by around 20°C. This pattern is well documented across the industry and isn’t specific to any one chemistry in our model, though in reality some chemistries handle cold charging better than others.
The hot side, and the cooling behaviour, are both built relative to the maximum pack temperature you set, not a fixed threshold. Full power holds until 15°C below that ceiling, then tapers hard on approach. Cooling power only starts engaging 20°C below the ceiling and is fully active by 8°C below it. That’s a deliberate modelling choice, not an approximation of convenience: a real thermal loop doesn’t waste cooling capacity chilling a pack that’s merely warm, it holds back until the pack is actually approaching its limit. If heat generation still outpaces the cooling power you’ve specified, modelled temperature can exceed your stated maximum; the calculator flags this rather than silently capping it, because a real pack would likely throttle harder or interrupt the session at that point, which this simplified model doesn’t force.
Heat generation itself is treated as resistive (I²R): for a fixed power, current scales as 1/voltage, so heat scales as 1/voltage². An 800V pack delivering the same power as a 400V pack draws half the current and generates roughly a quarter of the heat. That’s the real engineering reason high-power fast charging has pushed toward 800V+ architectures, and it shows up directly in the database: packs at ≥600V nominal run roughly double the peak kW of packs below 500V, at a similar implied peak current. The baseline loss assumption: about 8% of power converted to heat at full peak power at a 400V reference is a reasonable order-of-magnitude figure for DC fast charging, not one derived from cell-level DCIR data (a refinement we may add in a future version, since the database does carry pack DCIR fields).
How well does it match real packs?
We checked the blended model against six real vehicles pulled from the database that weren’t part of the original curve-fitting, comparing modelled 10–80% charge time against the manufacturer-quoted figure:
| Vehicle | Pack / peak power | Modelled | Real | Difference |
| Audi Q6 e-tron (NMC) | 95 kWh / 270 kW | 23.7 min | 21 min | +13% |
| Audi Q8 e-tron 50 (NMC) | 89 kWh / 150 kW | 31.3 min | 28 min | +12% |
| BYD Seal AWD (LFP) | 82 kWh / 150 kW | 28.2 min | 26 min | +8% |
| BYD Atto 3 (LFP) | 75 kWh / 220 kW | 20.9 min | 25 min | −16% |
| Tesla Model Y LFP | 58 kWh / 170 kW | 20.9 min | 25 min | −16% |
| Kia EV6 SR (NMC) | 74 kWh / 175 kW | 25.4 min | 17 min | +49% |
Five of the six land within roughly 8–16% of the manufacturer figure, in both directions. Reasonable for a generic model built from a population average rather than a vehicle-specific one. The Kia EV6 is a genuine outlier, and it’s an instructive one: the E-GMP platform’s 800V architecture and charging curve is well known in the EV press for outperforming what its headline peak power number would suggest, largely down to aggressive preconditioning and a deliberately flat curve. It’s a reminder that any generic model, however well calibrated on average, will miss individual designs that were specifically engineered to beat the curve.
What the model doesn’t know
A few things are worth being explicit about. Manufacturer-specific BMS thermal thresholds, route-based preconditioning, and charger-vehicle negotiation logic are all proprietary and unmodelled. We use one generic temperature-derate shape across every chemistry, when in reality this varies by design. Quoted “peak power” figures themselves are sometimes short-duration marketing numbers rather than figures representative of sustained performance, which the model inherits uncritically. And the 8% baseline heat-loss assumption is an engineering approximation rather than one derived from the pack’s own internal resistance data, so absolute temperature results should be read as indicative of direction and magnitude rather than precise. Treat every number this tool produces as a planning estimate against a population average, not a substitute for manufacturer data or physical testing on a specific pack.