Size the bottom cold plate that cools a Tesla- or BYD-class EV lithium-ion pack. Change the architecture (CTP / Blade / CTC / Modular), coolant, flow rate and plate geometry to watch the total heat, coolant rise, cell maximum temperature, margin to 40°C and pump power update in real time.
Parameters
Pack architecture
Sets the heat path geometry
Cell count N
Heat per cell
W
Higher during fast charging
Coolant
Auto-loads specific heat, density, k
Coolant flow rate
L/min
Plate thickness
mm
Plate width
mm
Inlet temperature
°C
Results
—
Total heat (kW)
—
Coolant ΔT in→out (K)
—
Avg coolant temp (°C)
—
Cell max temp (°C)
—
Thermal margin (K)
—
Pump power (W)
—
Battery pack + cold plate cross-section
Coolant follows the serpentine channel under the cells, picking up heat as it goes from inlet (blue) to outlet (red).
Flow sensitivity — cell max vs flow rate
Coolant comparison (same pack & flow)
Theory & Key Formulas
$$Q_{\text{tot}} = N \cdot q_{\text{cell}}, \qquad \dot m = \dot V \cdot \rho$$
Total pack heat Q_tot (W) and mass flow m (kg/s). N: cell count, q_cell: heat per cell (W), V: volumetric flow (m³/s), ρ: coolant density (kg/m³).
Coolant rise across the plate and cell maximum temperature lifted above the mean coolant by convective + TIM resistance. c_p: specific heat, h ≈ 8000 W/m²K effective, A_plate: plate area (m²).
Channel Reynolds number (Re > 2300 → turbulent) and an empirical pump-power estimate. D_h: hydraulic diameter, μ: dynamic viscosity. Real packs vary with channel length and bend count.
About this simulator
🙋
So how do modern EVs like a Tesla Model 3 or BYD Seal actually keep their batteries cool? Is it basically the same idea as an engine radiator?
🎓
Same family of idea, but what you cool is very different. A Model 3 has about 4000 21700 cells — little finger-sized cylinders — sitting across the entire floor. Underneath them is an aluminium cold plate with a serpentine channel. About 30 L/min of 50% ethylene-glycol water runs through it, picks up the ~20 kW the pack puts out, and dumps it into the front radiator. BYD's Blade uses long, flat cells, so the plate looks different, and CTC (Cell to Chassis) packs go even further — the floor of the car itself is the cold plate.
🙋
20 kW is a lot of heat — like several home air conditioners. Why so much fuss about temperature though? Cells getting hot really is that bad?
🎓
Yeah, lithium-ion is genuinely fragile. The happy zone is 25–35°C. Above 40°C side reactions take off and capacity fade accelerates — the old rule of thumb is "every 10°C above target halves cycle life". Above 60°C the separator can fail and you get internal shorts, then thermal runaway, then fire. So thermal engineers fix a hard target — cell-max ≤ 40°C, ideally 35°C — and then look for the smallest cooling system (flow, plate size, pump power) that achieves it.
🙋
OK — then I should just crank up the flow, right? Moving the slider from 30 to 100 L/min really does drop the cell temp. But the pump power jumps too.
🎓
Exactly the trade-off. Cooling gets easier with flow, but pump power scales roughly with the square of flow, plus the line losses go up. That pump comes off your traction battery, so over-cooling literally costs you range. Production packs typically sit at 30–50 L/min and 200–400 W of pump power, targeting a 5–10 K margin. A good rule with this tool: pick a flow that keeps the coolant in-to-out ΔT around 10–15 K, and you'll be near the real-world sweet spot.
🙋
If I switch the coolant to "Water", the cell temp drops even more. So why not use water on real cars?
🎓
Good catch. Pure water has c_p = 4186 J/kg·K, about 25% more than EG-50 (3300), so it carries more heat per kg. But it freezes at 0°C and corrodes aluminium and copper. In a Canadian winter at −20°C an all-water loop would burst, so it's just not allowed. EG-50 freezes at −37°C and comes pre-mixed with inhibitors. R134a is a refrigerant — paired with a chiller it can drop battery temperature below ambient, which is why it's used for sustained fast charging. Dielectric oil is the new kid: cells are dunked directly in oil ("immersion cooling"), giving great temperature uniformity and short-circuit immunity, at the cost of lower c_p — you compensate with more flow or phase change.
🙋
Pushing the plate width to 1500 mm gives me a huge margin. Are bigger plates just always better?
🎓
More area helps because the cell-to-coolant ΔT shrinks, yes. But you're paying in chassis space, mass, cost, and channel pressure-drop. Also, transfer is a balance of area and velocity: if you stretch the plate without re-doing the channels, velocity drops, the flow goes laminar, and the real h falls — this simple model fixes h at 8000 so you won't see that, but it bites in reality. A Model 3 plate is ~500 mm wide × ~1500 mm long because that's the cell footprint; BYD Blade is taller because the cells are longer. "Smallest plate that fits the cells" is the design rule.
Frequently asked questions
Lithium-ion cells live longest and perform best between 25 and 35°C. Above 40°C the SEI layer grows faster and capacity fade accelerates roughly exponentially — every 10°C rise is often quoted as halving cycle life. Above 60°C the separator can fail, leading to internal shorts and thermal runaway. Tesla, BYD and other OEMs typically target a 35–45°C cell-maximum and design the cold plate with at least 5K of margin against the 40°C ceiling.
CTP and BYD's Blade remove the module casings to gain volumetric efficiency, which leaves less air gap between cells and concentrates the heat path through a single bottom (or side) cold plate. The plate therefore tends to be large and thin. Modular packs run cooling to each module independently — thermally easier but with worse volumetric efficiency and more parts. CTC (Cell to Chassis) goes further and uses the vehicle floor itself as cold plate and structural member, the most integrated form available today.
Pure water has the highest specific heat (4186 J/kg·K) and is thermally ideal, but it freezes at 0°C and corrodes the cooling loop. EVs need antifreeze and corrosion inhibition, so the de-facto standard is 50% ethylene glycol in water: freezing point ≈ −37°C and specific heat ≈ 3300 J/kg·K (about 79% of water). Roughly 25% more flow restores the cooling capacity. R134a enables sub-ambient cooling via a heat pump but adds system complexity, while dielectric oil supports immersion cooling — lower heat capacity but excellent uniformity and short-circuit immunity.
Production EV cold plates use serpentine channels to force turbulent flow and a thermal interface material (TIM) between cell base and plate. The coolant-side convective h is roughly 2000–5000 W/m²K in turbulent conditions, and the effective end-to-end value including the TIM is about 6000–10000 W/m²K. This tool uses h = 8000 W/m²K, which reproduces Tesla Model 3 numbers (≈4000 cells, ≈20 kW heat, EG-50 at 30 L/min) with cell-max ≈ 30–35°C and a 5–10 K margin. Use it for first-pass sizing; final designs must check local hot-spots with 3-D CFD.
Real-world applications
Tesla Model 3 / Model Y CTP bottom cooling: roughly 4000 21700 cells share a single bottom cold plate. With 30 L/min of EG-50 the plate handles about 20 kW and keeps the cell maximum below 40°C even on a 250 kW Supercharger session. The defaults in this tool match that configuration and reproduce the typical 10–12 K coolant rise and ~290 W of pump power seen in disassembly reports.
BYD Blade (LFP) long-cell cooling: Blade cells are roughly 960 mm long, so the cold plate is also long and narrow. LFP chemistry tolerates higher cell-temperatures than NMC but generates a touch more heat per kWh, so select "Blade (BYD)" in the tool and try 6–8 W per cell to see how much margin you have. Designers normally aim for ≥10 K margin at 40 L/min of EG-50.
CTC (Cell to Chassis) integrated cooling: in packs such as Zeekr 001 or Tesla's 4680 structural pack, the pack bottom plate becomes the vehicle floor and the cold plate at once. That saves mass and parts but reduces channel design freedom because it must also pass crash load. Pick "CTC" to inspect the high-density regime.
Up-front sizing for fast charging and high-speed driving: a 250 kW Supercharger V3 or sustained autobahn cruise can push cell heat to 10–20 W. Set the per-cell heat to 10 W and check whether margin survives; if not, increase flow before redesigning the plate. Detailed transients still need 3-D CFD, but this lumped model is enough for steady-state sizing.
Common misconceptions & cautions
The biggest trap is designing to the average and forgetting the hot-spot. This tool returns a pack-average cell maximum; the real outlet end, channel corners and cell-centre interior all sit 3–5 K higher. CFD or instrumented testing must confirm the local peak — add a 5 K safety pad on top of this number before committing. Early Model S packs suffered outlet-side hot-spots and that drove several channel redesigns in later generations.
Second trap: throwing flow at the problem and ignoring pump power. Double the flow and the loop's pressure drop nearly quadruples, so pump power often goes up by 3–4×. The pump electricity comes straight out of your traction range. The empirical pump formula here ignores valves, fittings and filters; bias your estimate up by ~50%. The real design objective is "the smallest flow that keeps cell-max ≤ 40°C".
Third: treating one coolant as universal. EG-50 is the world default, but cold-climate variants use 60% glycol and warm-climate variants 40%. Specific heat drops about 5% per +10% glycol, so the required flow shifts by region. Looking forward, dielectric-oil immersion cooling is being qualified as a thermal-runaway mitigation, and new chemistries (solid-state, sodium-ion) change the heat-generation profile entirely. The presets here are representative of 2026 production; treat them as a starting point, not gospel.
How to Use
Enter the number of cells in your pack (e.g., 96 for a BYD Blade 44.9 kWh module) and heat generation per cell in watts (typical 2–8 W per cell at 3C discharge).
Set coolant flow rate in LPM (1.5–4.0 LPM typical for automotive glycol-water mix) and cold plate thickness in mm (3–8 mm aluminum or copper sandwich).
Run the simulator to obtain total heat load, coolant temperature rise, maximum cell temperature, and pump power consumption; adjust parameters iteratively to keep cell max temperature below 55°C and thermal margin above 10 K.
Worked Example
Tesla Model 3 LR pack equivalent: 96 cells × 5.2 W heat = 499.2 W total load. Coolant (50% ethylene glycol) flows at 2.8 LPM through a 5 mm aluminum cold plate. Inlet temperature 25°C, specific heat 3.51 kJ/kg·K, density 1.055 kg/L. Temperature rise ΔT = 499.2 W / (2.8 LPM × 1.055 kg/L × 60 s/min × 3510 J/kg·K) ≈ 1.8 K. Average coolant temp 26.9°C. With convection coefficient ~2800 W/m²·K and 0.18 m² plate area, cell max temperature reaches 48.2°C. Thermal margin 6.8 K. Pump power ~35 W at 2.8 LPM with 0.25 bar pressure drop.
Practical Notes
CTP (cell-to-pack) architectures eliminate module frames; increase coolant coverage by 18–22% versus traditional 4680-tube designs, reducing required flow rate by ~0.6 LPM.
For fast charging (>150 kW DC), heat generation jumps to 12–15 W/cell; increase flow to 3.5+ LPM or reduce plate thickness to 4 mm for faster transient response.
Glycol-water mixture thermal conductivity (0.42 W/m·K) is 15% lower than pure water; account for 2–3°C higher cell temperatures in design margin calculations.
Pump cavitation risk exists below 0.5 LPM; verify NPSH with your coolant loop pressure relief setting (typically 1.0–1.5 bar).
🎬 Watch it in motion
Chemistry Becomes Power | How a Battery Works
Why 3.7 V reads as 3.4 V | How a Battery works #Shorts
Why voltage sags under heavy load | How a Battery works #Shorts
Why two metals make electricity | How a Battery works #Shorts
What happens inside a battery? | How a Battery works #Shorts
