EV battery cooling keeps a lithium-ion pack inside a working temperature band while limiting how far individual cells drift from one another, and the method that achieves both depends on measured cell heat generation, the uniformity target the cells carry, and the packaging available. Air, indirect liquid cold plates, direct refrigerant, immersion, heat pipes and phase-change buffering each reach a different physical limit first. Heat generation has to come from cell-level data across the expected current, state of charge and temperature range. Pack capacity multiplied by C-rate describes electrical throughput, so it is not a substitute.
How EV Battery Cooling Works and What It Is Actually Controlling
An EV battery cooling system has to hold two targets at once. One is an absolute temperature band for the cells; the other is a ceiling on how far cells may drift from one another. Which target binds first depends on the flow path, the packaging and the cell-to-coolant interface.
Heat generation inside the pack has several contributors. Irreversible heat comes from ohmic and polarization losses in the cells, and reversible entropic heat comes from the electrochemistry itself. Busbars, tabs and contact resistances add more, so joints often become hot spots before cell bodies do. All of these move with current, internal resistance, state of charge, temperature, aging and charge direction. A cooling load worth designing to therefore comes from calorimetry, a supplier heat-generation map, or a validated electrothermal model run across your actual duty cycle.
Published temperature bands differ, and the useful question is which one each figure answers. An E-Mobility Engineering feature on EV battery cooling puts best performance and longest life at roughly 20–25 °C, a longevity optimum. Wider operating tolerances are also quoted. A design that treats a tolerance as its target can pass its own acceptance test and still age faster than the cell datasheet assumes. Both kinds of figure are industry generalizations, and the approved cell specification takes precedence over either.
Normal thermal management is not a propagation-safety strategy, and treating it as one is a common design error. Normal cooling controls working temperature, spread, aging and auxiliary energy use. Thermal runaway and propagation safety is a separate discipline: detection, venting, isolation, occupant warning and cell-to-cell propagation resistance. For road vehicles it is addressed by ISO 6469-1:2019/Amd 1:2022, which covers safety management of thermal propagation for rechargeable energy storage systems. A cooling loop sized for normal operation is not a propagation strategy, and no version of that standard prescribes a cooling method, a plate size or an optimum cell temperature.
Why Liquid Cooling Does Not Automatically Beat Air Cooling
Surface heat flux, the flow actually available, and the uniformity target decide this comparison, not pack capacity or vehicle class. Air-cooling capability moves with air mass flow, allowable air temperature rise, inlet temperature, cell-to-air surface area, duct bypass and flow distribution, fan static pressure, ambient conditions, and whether the air is conditioned first. Liquid capability moves with plate area, contact resistance, flow rate, inlet temperature, channel geometry, pressure drop and chiller capacity. Neither list collapses into a single wattage threshold, and a published threshold usually turns out to be one supplier’s product envelope.
What can be stated is the scale of the gap. For the same heat removed at the same fluid temperature rise, air needs on the order of 3,000 times the volumetric flow of water. The arithmetic is straightforward: at about 25 °C, air carries roughly 1.005 kJ/kg·K at a density near 1.225 kg/m³, or about 1.2 kJ per cubic metre per kelvin, while water carries about 4.18 kJ/kg·K at roughly 1,000 kg/m³, or about 4,180 kJ per cubic metre per kelvin. The figures assume sensible heat only, atmospheric pressure and pure water; a glycol mixture carries less. In practice an air-cooled design tends to run up against the duct volume and fan power the packaging allows, and where that happens it can lose uniformity before it loses average temperature.
The two paths also combine. Phase-change material can buffer peaks inside an air-cooled enclosure, and a mostly air-cooled system can carry a small liquid loop serving only the highest-flux modules. Stationary storage sits under different packaging and service constraints, and it is a separate comparison.
Six Cooling Methods and the Limit Each One Reaches First
No single method wins on heat load alone. Which of the six reaches its limit first depends on surface heat flux, packaging volume, and whether the fluid can safely touch live cells.
| Method | Heat path | More suitable when | Limit reached first |
|---|---|---|---|
| Forced or conditioned air | Cell surface to air to duct or enclosure | Heat flux is low, airflow is available, wider spread is tolerable | Low volumetric heat capacity and airflow maldistribution |
| Indirect liquid, cold plate | Cell to interface material to plate to coolant | Higher heat flux, or spread has to be tight | Interface resistance, pressure drop, leak management |
| Direct refrigerant | Cell or plate to evaporating refrigerant | Packaging favors tying the pack into the refrigerant circuit | Refrigerant distribution and superheat control across parallel branches |
| Immersion, direct dielectric | Cell outer surface to dielectric fluid | Surface heat flux is high or packaging is very compact | Fluid viscosity, pumping power, compatibility, serviceability |
| Heat pipe | Cell to phase change inside a sealed pipe, then to a sink | Heat has to move out of a confined spot to a cooler region | Capillary limit, orientation, and the sink still has to exist |
| PCM buffering | Cell to latent heat storage | Transient peaks need smoothing | Stored heat still needs a regeneration or rejection path |
Maturity is uneven across that list. Air cooling and indirect liquid cooling are established in production, and refrigerant-based approaches appear in production vehicles. Immersion and large-scale PCM systems remain more common in pilot programs and advanced development, with indirect liquid cooling the most established approach for high-energy passenger packs today. Immersion also carries a practical caveat already visible in the table: some dielectric fluids evaluated for it are viscous enough to raise pump power noticeably, so the fluid choice, not the heat transfer, tends to set the limit.
Cell format shifts where these limits fall. It changes the available contact area, the internal conduction path and the direction from which cooling works best, so the comparison needs the actual cell dimensions and the surfaces available for thermal contact. Flow layout matters for the same reason. In a single-pass serpentine channel, cells near the outlet can run hotter because the coolant has already absorbed heat upstream; parallel or multi-inlet layouts can reduce or redistribute that gradient.
Which Pack Variables Decide the Method, and Which to Establish First
Seven variables drive the decision. Each traces back to one of three origins: a cell datasheet, a validation requirement, or the duty cycle the pack actually runs.
| Variable | Where the value comes from | What it constrains |
|---|---|---|
| Cell heat generation | Calorimetry, supplier heat-generation map, or validated electrothermal model | Every downstream sizing step |
| Cell-to-cell spread limit | Cell and pack validation requirements, stated with the measurement condition | Whether air remains an option at all |
| Coolant temperature rise | Flow rate, plate area, channel layout | One term of the spread budget |
| Interface resistance | Bond-line thickness, filler conductivity, surface flatness | Achievable cell temperature at a given plate temperature |
| Enclosure and airflow path | Ingress rating, service access, packaging volume | Feasibility of conditioned air |
| Working fluid | Electrical conductivity, low-temperature viscosity, material compatibility | Leak consequence and service regime |
| Charge derating and shutdown thresholds | Approved cell specification and BMS safety concept | Sustained fast-charge behavior |
Two of the seven are worth locking first. Cell heat generation comes first, because plate area, channel count, flow rate and pump selection all take it as an input. The cell-to-cell spread limit comes next; it is set by cell and pack validation requirements rather than by the designer’s preference, which is what makes it hard to relax late. Enclosure type, flow path and working fluid can be revisited, but not for free: change any of them and the thermal-resistance network, the flow and pressure drop, the pump selection, the material compatibility and the low-temperature behavior all shift, so the sizing has to be re-checked rather than assumed to hold.
Stated against our own interest: where a pack’s measured dissipation is modest and the packaging allows conditioned air, a liquid loop adds pumps, fittings, coolant service and leak detection for margin the duty cycle may never call on. A spread budget on paper will show that faster than a supplier comparison will.
Coolant conductivity limits that changed in 2025
Coolant selection has tightened, and older coolant specifications may not reflect it. Two constraints matter when picking a fluid.
- For platforms sold into China, a national EV-coolant standard now limits the electrical conductivity of the coolant as supplied, with a permitted rise in service. Confirm the current requirement and its in-service limit with your coolant supplier before specifying a fluid for that market.
- ASTM D8566-24, Standard Specification for Glycol-Based Electric Vehicle Coolant with Low Electrical Conductivity, was approved by ASTM in 2024 under subcommittee D15.26 and specifies a maximum electrical conductivity of 100 µS/cm for fresh coolant. Check whether your loop is specified to this document or to a legacy engine-coolant specification, since conventional engine coolants sit orders of magnitude higher.
Neither constraint sets an optimum cell temperature, a uniformity target or a cooling flow rate. They bound the fluid, and the thermal design stays yours.
Interface Resistance Between the Cell and the Cold Plate
The bond line, not the plate, often sets the achievable cell temperature. Interface resistance is a series element: it rises with thickness and falls with conductivity and area, so a modest gap filled with a mediocre material can undo a well-designed plate. This is the interface a Trumonytechs cold plate is specified around — the plate defines the coolant-side path, while the material and stack-up between cell and plate govern how much of the plate’s capability actually reaches the cell. The E-Mobility Engineering feature makes the physics explicit: a 10 mm gap filled with a 10 W/m·K material carries about the same thermal resistance as a 1 mm gap filled with a 1 W/m·K material, all else being equal.
That relation is often read as a cost lever, but it is a statement about thermal resistance, not total cost. Halving the bond line and doubling filler conductivity move the interface resistance by a similar amount; which is cheaper depends on the project, because thickness is a tolerance and flatness decision while conductivity is a price decision. Neither is free to change in isolation. A workable interface still has to satisfy gap and flatness, tolerance absorption, assembly stress, curing and bonding needs, contact resistance at both faces, and manufacturing yield. Selecting the thermal interface material is therefore a gap, flatness, assembly-stress, curing and bonding decision as much as a conductivity one, and Trumonytechs supplies the interface materials that sit in this bond line rather than a single premium grade to be dropped in everywhere. The 10 mm-against-1 mm comparison holds only for one-dimensional conduction with contact resistance at the two surfaces set aside, and that contact resistance grows when either surface is out of flat.
Coolant flow rate follows from the heat load and the rise you allow. From q = ṁ × Cp × ΔT, a load of 5 kW held to a 3 K coolant rise needs about 5,000 ÷ (3,300 × 3) ≈ 0.5 kg/s. The 3.3 kJ/kg·K figure comes from pure-component values for a 50/50 water–glycol mixture with ideal mixing assumed, so substitute your own measured load and your own fluid datasheet. What that calculation sizes is the coolant rise, one term in the cell-to-cell spread budget. The rest comes from manifold flow distribution, in-plane conduction across the plate, interface resistance, differences in per-cell heat generation, edge effects at the module boundary, and sensor placement and tolerance. Check what coolant rise a stated spread budget can absorb before comparing plate options.
When a plate is sized from total heat load alone and nobody writes the spread budget down, the pack can meet its average temperature target and still fail the uniformity check near the outlet. The correction then tends to arrive as more flow instead of a different manifold layout. In packs cycling daily in high ambient heat, the cells nearest the coolant outlet and the plate-to-cell bond line are reasonable first places to re-check when measured spread starts to widen.
From Heat Generation Data to a Cooling Method Decision
Two inputs decide an EV battery cooling method above all others: a measured or modelled heat-generation profile, and a cell-to-cell spread limit stated with its measurement condition. Once both exist, method selection, plate sizing and flow rate follow with much less guesswork, and most of the disagreement about air against liquid turns out to be disagreement about those two numbers. Both are project-level and depend on your own duty cycle, so they are worth establishing on a sustained-rate test before you choose hardware.
If you already have a heat-generation profile and a stated uniformity target, the next step is a channel layout and interface stack checked against both, which is where Trumonytechs’ battery pack thermal management work starts. If you have neither, start with calorimetry or a supplier heat-generation map, because every sizing decision after that point inherits its accuracy.
FAQ
Does a parked EV still need battery cooling?
Sometimes, and whether it does is a control-strategy question, not a driver one. Some platforms keep thermal management active while parked in high ambient heat to limit aging, at a cost in auxiliary energy and standby range. Whether a given platform does depends on its control strategy, ambient boundaries, and the SOC and auxiliary-energy budget it is validated to, so confirm the behavior against the platform’s thermal control specification rather than assuming it.
Can a pack be converted from air cooling to liquid cooling later?
Rarely without a pack redesign. Plate mounting, cell spacing, sealing and coolant routing are physical decisions fixed at pack layout, unlike control set points such as target coolant temperature, which stay adjustable afterwards. Treat the cooling method as settled once the enclosure geometry is frozen.
How often does coolant need replacing in a battery loop?
On the interval in the vehicle or battery-system service specification. Fluid-supplier data on inhibitor life and material compatibility should support that interval, and the loop’s metals, seals and conductivity drift all bear on it.
What has to change before immersion cooling suits a production pack?
Three things: a pumping-power budget that survives the fluid’s viscosity, seals and service access designed for a wetted pack, and volume supply of a qualified fluid. Heat transfer is usually not the limiting factor.
What does battery preconditioning do for charging speed?
Preconditioning brings the pack toward the window where the BMS will allow higher charge current, which is why one charger can deliver very different curves on a cold pack and a preconditioned one. Derating and shutdown thresholds come from the approved cell specification, and preconditioning also draws pack energy, so extra thermal capability costs range.
Further Reading
- ASTM D8566-24, Glycol-Based Electric Vehicle Coolant with Low Electrical Conductivity — Standard (T1). Supports the coolant electrical-conductivity limit used when specifying a fluid; applies to fresh glycol-based EV coolant, not legacy engine coolant.
- Thermal Management of Lithium-Ion Batteries: Phase Change Materials and Air-Cooling with Fins (arXiv 2503.10244) — Peer-reviewed-style research (T2). Supports how PCM and finned designs improve temperature uniformity relative to air cooling; results are from cylindrical-cell simulation, not a production pack.
Related Articles
- Exploring Types of Battery Cooling Systems: Comprehensive Guide — A wider catalog of active and passive cooling types behind the six methods compared here.
- Battery Thermal Management Ensures Safety And Performance — How temperature and cell-to-cell uniformity drive lithium battery life and safety across BTMS choices.
- What Causes Thermal Runaway in Batteries? — The propagation-safety discipline this article deliberately leaves out: what triggers runaway and how BTMS limits it.
- The Importance Of Thermal Management In Electric Vehicles — Why EV thermal management spans the battery, power electronics and motor, not just the pack.
- Why Battery Protection Matters Electric Vehicle — How pack protection and the BMS extend cell and module life alongside cooling.
