Technology
Simulation works for almost everything. For batteries, it did not.
Model the components, calculate their behavior, answer the design question before anything is built: that method is proven across engineering. It fails on batteries for one reason: the available cell models do not hold. This page explains why, and what we do instead.
Physics-based models · parameterized in our own laboratory in Karlsruhe
Discuss your battery challengeThe same design question has a different answer every time.
A battery for a smartphone has almost nothing in common with one for an electric vehicle, and neither resembles a power tool or a stationary storage unit. Cell technology also moves quickly. So the requirements differ by application and by year, which is exactly the situation simulation is good for. These are the questions our customers actually arrive with:
Two established approaches, and why neither is enough.
This is a methodological distinction, not a product claim. Both approaches below are legitimate and widely used; each fails on a different axis, and the axes are what matter when you choose one.
| Approach | How it works | What it costs you | Verdict |
|---|---|---|---|
| Equivalent circuit modelsECM | Resistors, capacitors and impedance elements imitate the cell’s terminal behavior. The parameters are multi-dimensional lookup tables built from measurements. | The model structure does not match the physics of the cell, and nonlinearities are not properly represented. It reproduces the operating points it was fitted to. | Too inaccurate |
| Microstructural modelsFEM | The electrode microstructure is reconstructed and the system of partial differential equations is solved spatially, using finite element tools. | Error-prone in setup, parameterization takes several months, a single simulation usually takes hours, and ensuring the model is valid is left to you. | Too complex, too slow |
| BatemoPhysics-based, measured | All relevant processes inside the cell are reproduced: electrical, chemical and thermodynamic, coupled, across length scales from micrometer particles to meter-scale packs and time scales from milliseconds to months. | It needs a battery laboratory. Chemistry labs for disassembly, custom cell building, in-house analysis, and measurements across every electrical, thermal and mechanical boundary condition. | Seconds, on an office computer |
The third row is not free either. It is simply a cost we carry instead of you. A physics-based model is only as good as the measurements it is parameterized from, which is why the laboratory is the condition for this approach.
Three things, and the combination is the point.
Each of the three exists elsewhere. Held together (a physics-based model, parameters from our own measurements, and the experience of having done it across the market), they answer questions none of them answers alone.
01 Battery modeling
The model describes the physics, not the curve.
Every relevant process inside the cell is reproduced: physical, chemical and thermodynamic, coupled to each other, dependent on operating condition and age. They run on different scales at once: from electrode particles a few micrometers across to packs measured in meters, from milliseconds to months.
02 Battery parameterization
The parameters come from our own laboratory.
Chemistry labs for cell disassembly, the ability to build custom cells, and in-house analysis. On top of that, measurements across the full range of electrical, thermal and mechanical boundary conditions, and optimized methods to identify the model parameters from them.
03 Battery data
The experience covers the market, not a segment.
Hundreds of cells have been through the laboratory: measured over the entire operating range, completely disassembled, and modeled. All chemistries, all cell designs, all sizes, all industries.
Inside a 21700 cell
Opened, layer by layer.
A cylindrical cell is four thin layers, wound up and sealed in a can. Scroll to open it. Drag to turn it.
- The cell. A 21700: 21 mm across, 70 mm long. We open every cell in our library.
- The can. Under the can and the cap with the positive terminal sits the winding.
- The layers. One long strip, rolled up tight: cathode, separator, anode, separator.
- The model. Our physics-based model describes what happens in these layers: electrochemistry, transport and heat, across the whole cell.
The same models are a training data source.
A data-driven algorithm is only as good as the data it learned from, and battery training data has a specific problem: it has to cover operating conditions that a test campaign cannot reach in the time available, across fresh and aged cells, across every cell type you use. Running those experiments when the data is actually needed is expensive, slow, and usually impossible.
A validated physics-based model removes that constraint. It runs in seconds inside an automation backend, so thousands of training profiles can be generated overnight. Because the model is physical, it also exposes the internal quantities a real cell will not give you, which is what lets an algorithm be checked against physics rather than against its own training set. Using the model library, the training basis stays consistent across every cell type from day one.
The questions this usually comes up with (state of charge under extreme conditions, state of health, whether an estimator generalises at all) are the ones on the BMS page →
“If you can show us another battery cell model that is more accurate, we will refund your money.”
Everything we build has to be as fast, as physical and as accurate as we can make it. This is where we put money behind that.
How accuracy is measured
Against validation experiments covering the cell’s full operating area as specified by the manufacturer, plus measurements in the interpolation and extrapolation regime that the model’s creator has never seen. The measurement time splits three ways:
- 50 %
- known to us and to a competitor
- 25 %
- known only to us
- 25 %
- known only to the competitor, not to us
That last quarter is the part that matters. A model that only performs on data its author chose has proven nothing.
Where this actually gets used.
The same models answer very different questions depending on what you are deciding. Each of these has its own page, with the specifics.
- Aging and lifetime
- Cell benchmarking
- Fast charging
- Battery cell design
- Thermal modeling
- BMS development
- System optimization
Or start from the products: see which model answers which question →
Two ways forward.
If you need the data
See the measurements the models are built on.
Batemo Insights holds the measured behavior behind every cell in the public library, including what the public Explorer deliberately does not show.
If you need the decision
Bring us the battery question you are stuck on.
An engineer who has measured these cells reads your message, not a mailbox and not a sales desk that forwards it.
Or start with the measurements: explore the public Batemo Cell Explorer →