By Jae-Won Kim, CTO & Co-Founder, Sasung Power
3 to 5 W per meter-kelvin is the in-plane thermal conductivity of carbon fiber composite, depending on fiber type and volume fraction. Through-thickness conductivity is 0.5 to 1.0 W per meter-kelvin, creating 4 to 8 anisotropy within one material. In a conventional pouch or prismatic pack, through-thickness heat crosses a single cell, while cooling plates between cells intercept it along that short path. In a structural cell, the cell itself sits inside a load-carrying laminate, with no room for an interstitial cooling plate. The thermal architecture must therefore be rebuilt around this constraint.
Why Conventional Cooling Plates Fail
In a standard EV floor pack, cells produce heat during charge and discharge. Heat crosses the thin cell dimension to a cooling plate between cell groups, and fluid in the plate carries it to a radiator. The path from cell center to cooling plate, typically 3 to 8 mm, is short enough to keep cell-to-coolant resistance low and make 2C discharge manageable without excessive temperature rise.
A structural cell in a floor panel cannot place cooling plates between cell groups. A load-carrying laminate has no such gap, and an interstitial layer that carries no load lowers panel structural efficiency. Cell heat must cross the composite structure itself. With low through-thickness conductivity, thermal resistance from the electrochemically active zone to a panel-edge heat exchanger is much higher than in a conventional pack.
The difference can be quantified. An aluminum-plate pack provides roughly 2 to 5 K per watt of total resistance from cell center to coolant. A structural carbon fiber composite panel without internal cooling channels reaches 15 to 30 K per watt through-thickness, depending on panel thickness and fiber layup. At 0.8 W per cell during 1C discharge in a 48-cell structural floor section, the passively cooled panel develops a 12 to 24°C differential. That exceeds the 10°C uniformity target for LFP electrochemical performance.
Two Cooling Options for Structural Integration
Two practical approaches preserve the load path while addressing heat. Each introduces trade-offs.
Edge cooling uses channels integrated into the structural frame. The composite panel ends at structural ribs or sill beams. Embedding cooling-fluid channels in those members lets heat travel in-plane along the higher-conductivity fiber direction to the panel edge, then move into coolant at the rib. For 400 to 600 mm of in-plane conduction to a cooled edge rib, thermal conductivity of at least 5 W per meter-kelvin is needed to keep the center-to-edge differential below 8°C at 1C discharge. Standard epoxy-matrix CF laminates can achieve this. The farthest outer cells run hottest, so BMS state estimation must account for non-uniformity. Cell temperature cannot be assumed uniform across the panel.
Thermal interface material, or TIM, layers sit between the cell and composite. A high-conductivity TIM pad, such as graphite sheet at 200 to 400 W per meter-kelvin in-plane, spreads heat laterally within the cell and lowers effective resistance to the panel edge. The layer adds 0.3 to 0.5 mm to cell thickness and a small mass penalty, while potentially improving temperature uniformity by 30 to 40% versus a bare interface. Its structural cost depends on adhesive bond quality and whether the pad transfers shear across the interface during flexural loading. The key issue is materials qualification, not thermal modeling.
Managing Thermal Gradients in BMS Software
Either physical cooling method leaves a structural panel with temperature variation, so the BMS cannot use one thermal model for every cell. With uniform cooling, a conventional pack varies by 2 to 4°C at steady 1C discharge. With edge cooling, the same rate can produce a 6 to 10°C center-to-edge gradient in a structural panel.
LFP capacity and impedance depend on temperature. At 15°C versus 25°C, LFP internal resistance is approximately 30 to 40% higher, while usable capacity at high discharge rates falls by 8 to 12%. One SoC model applied across cells at different temperatures accumulates estimation error at a rate tied to the temperature gradient and discharge rate. When a structural panel has a known, stable center-to-edge gradient, the correction is direct: distributed thermal sensors provide local temperature for per-cell compensation in the capacity and resistance models. If ambient temperature and solar load change the gradient substantially, the model must follow thermal state dynamically instead of using a fixed offset.
We use at least one temperature sensor per 4 cells, positioned to resolve the center-to-edge gradient rather than infer it from one point. A 48-cell panel section therefore adds roughly 12 sensor nodes, compared with 4 to 6 sensors in a conventional pack with equivalent cell count.
Cold Operation: The Structural Cell Limit
Below approximately -10°C, LFP cells need preconditioning before high-rate charging to prevent lithium plating on the graphite anode. Conventional packs use resistive heaters in the cooling plate or dedicated heater foils beneath cell groups. Structural integration removes the interstitial space both methods require.
Edge heating is slow for a structural cell panel because the same resistance that impedes cooling also impedes heating. In-plane conductivity moves heat reasonably well along the fiber direction, but under typical cold-soak conditions, cells at the panel center warm minutes after cells at the heated edge, from -20°C ambient to a 0°C operating target. During this ramp, non-uniformity exceeds the steady-state condition. Center cells may remain below the safe charging threshold after edge cells reach temperature.
Our current cold-operation strategy holds charging current at 0.1C across the panel until the coldest sensor reads above 0°C. We then ramp to the target rate over a 5-minute window while monitoring per-cell impedance for plating indicators. This is more conservative than a heated conventional pack, but it defines the honest operating boundary for the current structural integration architecture. Resistive trace heating in the CF ply layup is technically feasible, yet it adds manufacturing complexity and possible heater-ply interface failures under cyclic load. We are still evaluating it.
Thermal Runaway: A Different Risk Model
In a conventional pack, cell-to-cell barriers and cooling plates shape propagation. A runaway cell transfers heat mainly through the cooling plate, which can become an unintended heat source, and through the cell-to-cell gap when that gap is too small. A propagation time of 30 to 90 seconds can allow venting and BMS disconnection of the affected module.
Structural cells are mechanically and thermally coupled through the composite laminate. During runaway, one cell produces gas and heat, but the usual vent path is unavailable because the cell is enclosed in a load-carrying composite structure. That coupling sends thermal and mechanical stress through the laminate toward neighboring cells faster than in a module with cell-to-cell gaps.
This does not mean structural cells have worse runaway behavior in every dimension. LFP has a higher onset temperature, above 270°C for the cathode exothermic reaction versus 150 to 200°C for NMC, providing more response time. The propagation path differs, so conventional module-level runaway tests alone cannot assess it. Full structural-panel testing under representative mechanical boundary conditions is required, while most safety certification programs are still developing that infrastructure for this cell form factor.