Battery thermal model calculator
Simulate how hot the cells of a lithium-ion pack get. Each cell is a 1-RC equivalent circuit whose resistance changes with charge and temperature, makes I²R heat and loses it to the air. Add a faulty cell, pick a charge, discharge or cycling current, and see every cell's temperature, SOC and the pack voltage. It opens on the 10-cell Simulink project of Skill-Lync's battery courses, and a second tab checks your answers. Free, runs in your browser.
1-RC model, MathWorks / Huria 2012 cell tables · RK4, checked against scipy Radau · Updated 20 Nov 2026
Pack and load
Fault cell
Thermal
Positive current discharges; the sine starts on a discharge half-cycle. Frequency takes 1/600 or 0.001667. Start SOC 37.26% is the course script's charge deficit Qe_init = 15.6845 Ah of 25 Ah. Area and m·cp are the course pouch cell (0.0084 × 0.215 × 0.220 m, 2.04 MJ/m³·K). Up to 20 cells and 6 hours.
Set the inputs your model used above, then type the results it gave: peak temperatures, pack voltage range, and the "heat transfer coefficient" if your report quotes one. Leave any box empty to skip it.
- Average heat: 12.51 W per healthy cell, 57.71 W in the fault cell. h·A = 0.510 W/K, time constant m·cp/(h·A) = 1591 s (26.5 min).
- The healthy cells and fault cell 5 pass 40 °C, the top of the cell tables; above it the model holds the 40 °C values. For comparison, A123's 26650 datasheet rates its cell for −30 to 55 °C.
- After 12 whole cycles the SOC is up 2.41 points with zero net charge: the capacity table changes with temperature.
- Ageing (Wang 2011 LFP fit, C/2): healthy cells fade 1.82× and the fault cell 19.1× as fast per Ah as at 25 °C over this run.
What is the battery pack thermal modelling project?
"For a 10 cell series lithium ion battery model, simulate the thermal effects and compare life cycle performance at various temperatures, charge & discharge rates using MATLAB." It is the second project of Skill-Lync's battery courses, after the 18 kWh pack design, and Skill-Lync read all 519 public reports of it on 20 Nov 2026. 335 build in a faulty cell and 381 model convection. Almost all of them start from MathWorks' Simscape lithium pack example: ten cells, each a 1-RC equivalent circuit, one with a fault, all losing heat to the air.
The electrical part of each cell is a voltage source Em, a series resistance R0 and one resistor-capacitor pair R1, C1. All four, and the capacity, come from lookup tables over state of charge and temperature:
The thermal part is one lumped mass per cell. It gains the heat dissipated in R0 and R1, loses heat by convection, and (if you join cells with conduction blocks) trades heat with its neighbours:
The calculator solves exactly these equations for every cell with fixed half-second Runge-Kutta steps, about 14,400 of them for the 2-hour course run, in a fraction of a second.
Which cell data does the course model use?
The tables come from the MathWorks initialisation script for its 1-RC lithium cell, built on Huria, Ceraolo, Gazzarri and Jackey (2012), who fitted them to pulse tests of a 31 Ah NMC pouch cell at 5, 20 and 40 °C. Two versions show up in public reports, with identical Em, R0, R1 and C1 tables:
| Quantity | Course value | Where it comes from |
|---|---|---|
| Table breakpoints | SOC 0, 0.1, 0.25, 0.5, 0.75, 0.9, 1 · 5, 20, 40 °C | Script |
| Capacity | 24 / 25 / 26 Ah (2012 script) · 28.01 / 27.63 / 27.64 Ah (2022 .mat) | Script; the tested cell was 31 Ah |
| Em at 20 °C | 3.506 V (SOC 0) to 4.193 V (SOC 1) | Script |
| R0 at 20 °C | 8.2 to 8.7 mΩ (11 to 12 mΩ at 5 °C) | Script |
| R1 · C1 at 20 °C | 1.6 to 2.9 mΩ · 12,447 to 40,764 F | Script |
| Cell size | 0.0084 × 0.215 × 0.220 m → A = 0.101908 m² | Script and paper |
| Heat capacity | 2.04 MJ/(m³·K) × 3.973 × 10⁻⁴ m³ = 810.5 J/K | Script and paper |
| Convection h | 5 W/(m²·K), "5 to 25 for natural convection" | Script and paper |
| Start | Qe = 15.6845 Ah (SOC 37.26%), 20 °C (2012) or 25 °C (2022) | Script |
One line of the script trips people up: cell_Cp_heat = cell_rho_Cp * cell_volume; %J/kg/K. A volumetric heat capacity times a volume is a heat capacity in J/K, not a specific heat in J/(kg·K). It only works because the script sets the cell mass to 1 kg, so mass × "specific heat" comes out at the right 810.5 J/K. Change the mass to the real cell's and the thermal mass changes with it, which is wrong.
Without any cooling design, the time constant and the steady rise follow from two numbers, the heat made and h·A:
For the course cell τ = 1591 s, about 27 minutes, and h·A = 0.510 W/K, so every watt of heat held for long enough lifts the cell by 1.96 K. That is why a 2-hour run nearly settles, and why the fault cell's extra heat matters so much.
What are the answers for the course pack?
Ten cells, cell 5 faulty with the factors most reports type (capacity × 0.95, Em × 0.9, R0 × 5, R1 × 5, C1 × 0.95), a 50 A peak sine current at 1/600 Hz, 20 °C ambient, h = 5 W/(m²·K), 7,200 s. The healthy cells reach 45.0 °C and the fault cell 135.3 °C. The pack voltage swings between 29.28 and 43.15 V; 27 reports quote "30 to 43 volts", which is this swing. Solved by Skill-Lync with scipy's Radau solver at a relative tolerance of 10⁻¹⁰ and checked with LSODA:
| t (s) | Healthy cell (°C) | Fault cell 5 (°C) | SOC, healthy | SOC, fault cell |
|---|---|---|---|---|
| 0 | 20.00 | 20.00 | 37.26% | 37.26% |
| 600 | 27.74 | 55.49 | 38.22% | 39.67% |
| 1,200 | 33.04 | 79.93 | 38.86% | 39.67% |
| 1,800 | 36.67 | 96.68 | 39.29% | 39.67% |
| 2,400 | 39.16 | 108.16 | 39.58% | 39.67% |
| 3,000 | 40.86 | 116.04 | 39.67% | 39.67% |
| 3,600 | 42.02 | 121.44 | 39.68% | 39.68% |
| 4,200 | 42.82 | 125.15 | 39.68% | 39.68% |
| 4,800 | 43.37 | 127.69 | 39.68% | 39.68% |
| 5,400 | 43.74 | 129.43 | 39.68% | 39.68% |
| 6,000 | 44.00 | 130.63 | 39.68% | 39.68% |
| 6,600 | 44.18 | 131.45 | 39.68% | 39.68% |
| 7,200 | 44.30 | 132.01 | 39.68% | 39.68% |
The healthy cells average 12.51 W of heat over the run, the fault cell 57.71 W, about 4.6 times as much: five times the resistance, a little offset by the RC branch. That is the whole story of the fault cell. Its temperature is not a sign of runaway in this model; it is simply heat over h·A. Left running, it settles near 20 + 57.7 / 0.510 ≈ 133.3 °C, with its resistance held at the 40 °C values because the tables stop there.
Two things look like errors and are not. The SOC never comes back to where it started: the capacity table grows from 25 Ah at 20 °C to 26 Ah at 40 °C, so SOC = 1 − Qe/C(T) climbs from 37.26% to 39.68% with zero net charge. And the healthy cells settle at 44.3 °C rather than rising forever: by 7,200 s the run is 4.5 time constants long.
How does the C-rate and the ambient temperature change the heat?
The challenge asks you to compare temperatures and charge-discharge rates. One healthy course cell discharged from full to empty at constant current, h = 5 W/(m²·K), with 1C taken as 25 A:
| Ambient | C-rate (A) | Time to empty | Temperature rise | Peak | Voltage at empty |
|---|---|---|---|---|---|
| 5 °C | 0.5C (12.5 A) | 116.7 min | +4.7 K | 9.7 °C | 3.263 V |
| 5 °C | 1C (25 A) | 59.7 min | +13.2 K | 18.2 °C | 3.188 V |
| 5 °C | 2C (50 A) | 31.2 min | +36.0 K | 41.0 °C | 3.001 V |
| 5 °C | 3C (75 A) | 20.8 min | +62.9 K | 67.9 °C | 2.745 V |
| 20 °C | 0.5C (12.5 A) | 120.8 min | +3.3 K | 23.3 °C | 3.367 V |
| 20 °C | 1C (25 A) | 61.4 min | +11.6 K | 31.6 °C | 3.243 V |
| 20 °C | 2C (50 A) | 31.2 min | +34.7 K | 54.7 °C | 3.001 V |
| 20 °C | 3C (75 A) | 20.8 min | +61.2 K | 81.2 °C | 2.745 V |
| 40 °C | 0.5C (12.5 A) | 124.8 min | +3.1 K | 43.1 °C | 3.386 V |
| 40 °C | 1C (25 A) | 62.4 min | +11.3 K | 51.3 °C | 3.258 V |
| 40 °C | 2C (50 A) | 31.2 min | +34.6 K | 74.6 °C | 3.001 V |
| 40 °C | 3C (75 A) | 20.8 min | +61.0 K | 101.0 °C | 2.745 V |
The heat goes with the square of the current, so doubling it from 1C to 2C triples the rise here (from +11.6 to +34.7 K at 20 °C) even though the discharge is half as long. The ambient barely changes the rise; it shifts the whole curve. Cold cells have a higher R0 (11 to 12 mΩ at 5 °C against 8 to 9 at 20 °C) and less capacity, so at 5 °C and C/2 the cell gives 24.31 Ah instead of 25.16. At 2C and above every cell warms past 40 °C and ends on the 26 Ah of the hottest column.
As a check on the model, Huria et al. ran a 31 A discharge from 20 °C for 3,000 s and report "approximately 16 °C" of rise. The course tables give +16.6 K for the same current, with the cell emptying at 2999 s (25 Ah, where the tested cell had 31). A 1C charge from empty at 25 °C (2022 tables, 27.6 A) takes 60 minutes and adds +13.7 K.
How much cooling does the pack need?
Same course run, only h changed. h = 5 is still air; forced air is roughly 10 to 100 W/(m²·K):
| h (W/m²·K) | Healthy cell peak | Fault cell peak |
|---|---|---|
| 5 (course) | 45.0 °C | 135.3 °C |
| 10 | 33.0 °C | 79.9 °C |
| 25 | 25.6 °C | 45.9 °C |
| 50 | 23.1 °C | 34.1 °C |
| 100 | 21.8 °C | 28.2 °C |
Going from 5 to 25 W/(m²·K) brings the healthy cells from 45.0 to 25.6 °C. Conduction between neighbours spreads the fault cell's heat but does not remove it: with G = 0.401 W/K between each pair of cells (401 W/(m·K) copper, 1 cm², 0.1 m), the fault cell peaks at 91.0 °C instead of 135.3, its neighbour cell 4 rises to 59.7 °C, and the far end of the pack stays at 45.7 °C. Set "Cell-to-cell G" in the calculator to try your own (G = k·A/L of the conductive block).
How do you compare cycle life at different temperatures?
The challenge also asks for life cycle performance, and the Simscape model has no ageing in it. 283 reports mention cycle life, but only 24 put a number of cycles anywhere in the text and 2 mention Arrhenius. A widely used semi-empirical fit is Wang et al. (2011), from cycling tests of A123 LFP 26650 cells (2.2 Ah) at 15 to 60 °C and C/2 to 10C:
Qloss is the capacity lost in percent, R = 8.314 J/(mol·K), T in kelvin, Ah the charge throughput and A a factor fitted for each C-rate: 31,630 at C/2, 21,681 at 2C, 12,934 at 6C, 15,512 at 10C. Full cycles at 100% depth of discharge until the cell has lost 20% of its capacity:
| C-rate | 15 °C | 25 °C | 35 °C | 45 °C | 60 °C |
|---|---|---|---|---|---|
| C/2 | 16,925 | 7,588 | 3,584 | 1,774 | 669 |
| 2C | 22,063 | 10,032 | 4,801 | 2,407 | 923 |
| 6C | 18,338 | 8,659 | 4,293 | 2,224 | 893 |
| 10C | 4,282 | 2,100 | 1,078 | 577 | 243 |
Every 10 K roughly halves the life: about 7,588 cycles at 25 °C, 1,774 at 45 °C (C/2). The C-rate effect in the fit is not monotonic: 2C and 6C come out better than C/2 because of the fitted A, and 10C cuts the life to about 28% of the C/2 figure. Two cautions. The fit is for LFP cells, not the NMC cell of the course tables, so use it for relative comparisons, as the calculator does with its ageing factor (the Arrhenius term averaged over the simulated temperature). And above about 60 °C it is outside its data.
Which mistakes show up in published reports?
From all 519 public reports, read 20 Nov 2026; the Check tab catches the first three:
- A sensor reading quoted as a coefficient. 30 reports give a negative "heat transfer coefficient" such as −21.32 W/(m²·K) and a "thermal conductivity" of −10.66 W/(m·K), read off the probe of a heat transfer block. Those probes output heat flow in W and the sign is the direction of flow. h and k are inputs and always positive.
- Kelvin and Celsius mixed. Simscape reports temperatures in kelvin. A "peak of 44 K" is a Celsius number in the wrong unit.
- One cell's voltage given as the pack's. The pack of ten swings 29.28 to 43.15 V; a single cell about a tenth of that.
- Life cycle discussed without a number. The life comparison is usually a paragraph of general statements. The table above, or the ageing factor in the calculator, turns your simulated temperatures into a ratio you can defend.
- A fault cell at 135 °C reported without comment. With five times the resistance and still air, the model's fault cell runs 90 K hotter than its neighbours, well beyond the 40 °C top of the tables. That is the result worth discussing: what cooling, current limit or cell monitoring would catch it.
The thermal loop in a few lines of MATLAB, the same equations as this page for one healthy cell with fixed R0, R1 and C1 (the calculator also looks them up as SOC and temperature change):
% the thermal part of each cell, the same equations as this page (one cell, Euler steps)
R0 = 0.0085; R1 = 0.0029; C1 = 12447; % 20 C, SOC 0 row of the course tables
mcp = 810.5; hA = 5 * 0.101908; Tamb = 293.15;
dt = 0.5; t = 0:dt:7200; I = 50 * sin(2*pi*t/600);
V1 = 0; T = Tamb; Tlog = zeros(size(t));
for k = 1:numel(t)
Q = I(k)^2 * R0 + V1^2 / R1; % W: heat made in the cell
V1 = V1 + dt * (I(k)/C1 - V1/(R1*C1));
T = T + dt * (Q - hA*(T - Tamb)) / mcp; % W / (J/K) = K/s
Tlog(k) = T;
end
plot(t/60, Tlog - 273.15); xlabel('t (min)'); ylabel('T (C)')Skill-Lync ran a line-by-line port of it: the cell ends at 46.9 °C after 7,200 s, against 44.3 °C from the full model, which looks R1 and C1 up at the real SOC of about 39% (where R1 is lower) instead of holding the SOC-0 values.
What this calculator does not cover
- Reversible (entropic) heat. Heat is I²R0 + V1²/R1. The entropic term I·T·dU/dT can add or subtract a noticeable share at low currents; it needs dU/dT data the course tables do not have.
- Temperatures above 40 °C or below 5 °C use the nearest table column (no extrapolation). Simscape's own lookup settings may extrapolate instead, so very hot or cold results can differ.
- Temperature differences inside a cell, coolant flow, cold plates and thermal runaway. Each cell is one lumped temperature; for spatial fields and cooling channel design you need CFD or a finite-element thermal model.
- Ageing during the run. Capacity and resistance stay as tabulated; the ageing factor only compares rates.
Equivalent circuit cell models, pack thermal models and SOC estimation are the core of Skill-Lync's battery technology courses, built in MATLAB and Simulink the way battery management and EV teams work.
Frequently asked questions
What temperature should the 10-cell pack reach in the thermal modelling project?
With the MathWorks course tables, 20 °C ambient, h = 5 W/(m²·K) and a 50 A peak sine current at 1/600 Hz for 7,200 s, the healthy cells reach 45.0 °C and a fault cell with 5 times the resistance reaches 135.3 °C. The pack voltage swings between 29.28 and 43.15 V. Solved by Skill-Lync with scipy (Radau and LSODA) and re-solved in this page's browser code.
How is the heat generated in a lithium-ion cell calculated?
In a 1-RC equivalent circuit model, the heat made is I²R0 + V1²/R1: the Joule loss in the series resistance plus the loss in the RC branch resistance. At 25 A (1C), near full charge and 20 °C, this course cell makes 5.3 W in R0 at the first instant and about 6.4 W once the RC branch has charged, a minute or two later. The reversible (entropic) heat, I·T·dU/dT, is left out of this model; it matters at low currents.
What is the thermal time constant of a battery cell?
τ = m·cp / (h·A). For the course pouch cell (810.5 J/K, 0.1019 m², h = 5 W/(m²·K)) it is 1591 s, about 26.5 minutes. After 3τ (80 minutes) the cell is 95% of the way to its steady temperature, which is the heat divided by h·A.
Why is a negative heat transfer coefficient wrong?
The heat transfer coefficient h is a property you type into the convection block, so it is always positive. The sensor on a conductive or convective heat transfer block outputs the heat flow in watts, and its sign only tells you the direction. 30 public reports of this project quote a negative 'heat transfer coefficient' such as −21.32 W/(m²·K), which is really a heat flow reading.
How does temperature affect lithium-ion battery cycle life?
Strongly. In the Wang et al. (2011) fit for LFP 26650 cells, capacity fade per Ah cycled grows by 2.22 times from 25 °C to 45 °C at C/2. Cycled at C/2 and 100% depth of discharge, the fit gives about 7600 cycles to 20% capacity loss at 25 °C and about 1770 at 45 °C.
Why does the SOC go up when the battery heats with no charge going in?
Because in the table-based model SOC = 1 − Qe/C(T), and the capacity table rises from 25 Ah at 20 °C to 26 Ah at 40 °C. In the course run the SOC read at whole cycles climbs from 37.26% to 39.68% while the net charge is zero. It is a feature of the model, not an error in your coulomb counting.
Related on Skill-Lync
- Battery pack calculator (cells in series and parallel for any kWh, the project before this one)
- 2D heat conduction solver (temperature fields when one lump per cell is not enough)
- Damped pendulum calculator (ODE solvers and their tolerances)
- Battery technology courses on Skill-Lync
- Electric vehicle courses on Skill-Lync
- Student project: thermal modeling of a 10S1P battery pack
- All free engineering tools
Sources
- T. Huria, M. Ceraolo, J. Gazzarri, R. Jackey, "High Fidelity Electrical Model with Thermal Dependence for Characterization and Simulation of High Power Lithium Battery Cells", IEEE International Electric Vehicle Conference, March 2012: 31 Ah NMC cell at 5, 20, 40 °C, one RC branch, h = 5 W/(m²·K), 2.04 MJ/(m³·K), 0.0084 × 0.215 × 0.22 m, about 16 °C rise for 31 A over 3,000 s.
- MathWorks, initialisation script of the Simscape ssc_lithium_cell_1RC example (2012) and a 2022 battery_init.mat export, as pasted in full in two public Skill-Lync reports: all lookup tables, capacity, Qe_init, h and cell size on this page.
- J. Wang, P. Liu, J. Hicks-Garner et al., "Cycle-life model for graphite-LiFePO4 cells", Journal of Power Sources 196 (2011) 3942-3948; equation and A values as reproduced in T. Yuksel et al., Journal of Power Sources 338 (2017) 49-64, eq. 29 and table 4.
- Skill-Lync, all 519 public student reports titled "Thermal modeling of battery pack", read 20 Nov 2026: model structure, fault factors, currents, quoted results and the mistakes above.
- Skill-Lync, scripts/verify/battery-thermal.py: solves every case on this page with scipy's Radau and LSODA solvers at a relative tolerance of 10⁻¹⁰, using its own table interpolation, and computes the cycle-life table. This page re-solves everything in the browser code and checks its numbers against that script at build time.