Cells age. The voltage curve your BMS was built on does not.

Almost every state estimator ties terminal voltage to state of charge through an open-circuit voltage curve measured once, on a new cell — and then treats it as a constant for the rest of the battery's life.

It is not a constant. In the measurement above, capacity fade has already been taken into account, and the aged cells still depart from the new-cell curve by up to 116 mV. Converted through the local slope of the curve, that is a median error of 5.6 percentage points of state of charge across the 20–80% window, and more than 11 points at its worst. None of it is visible from capacity fade alone, which is the one thing a BMS does track.

Composite silicon–graphite anodes make the problem sharper. They raise energy density, and they add large voltage hysteresis on top of the drift: on the same cells at beginning of life, the charge and discharge branches are 253 mV apart at 10% SOC, so even a fully rested cell does not map one voltage onto one state of charge.

From 18 February 2027, Regulation (EU) 2023/1542 requires every electric-vehicle and industrial battery above 2 kWh placed on the EU market to report state of health and state of certified energy through the battery passport. The number stops being internal. It becomes published, and auditable.

Four situations this work fits

• Cell and pack manufacturers You will have to publish a state-of-health figure on a battery passport, and you need to know how defensible your current number is before an auditor asks.
• OEM and Tier-1 BMS teams Your estimator was parameterized on fresh cells and drifts in the field. You need to know which degradation mechanism is causing it, not only that the error grows.
• Test houses and analytics providers You have the aging campaign but not the electrochemical interpretation. You need the degradation modes separated: lithium inventory, active material, resistance.
• Second-life and warranty assessors You can measure today's capacity yourself — but only by running a full cycle on every pack, and it tells you nothing about what happens next. Degradation-mode analysis works from a partial charge, and it distinguishes a pack that has lost lithium inventory (and will fade slowly) from one losing active material (which will not).

Three ways to start

All six services, with scope and data requirements →

From first email to delivery

1. A call, or an email Thirty minutes, free, no obligation: what the cells are, how the data was collected, and the decision it has to support. If you would rather write than talk, that works too.
2. A sample of the data A representative subset is enough to begin. If you want an NDA in place first, send yours or ask for mine — but it is your call, not a step I impose.
3. Feasibility assessment Fixed price, credited back if you go ahead. You get a written verdict on what the data supports and a scoped fixed-price proposal. If the data cannot carry the analysis, you get that answer instead — together with the test campaign that would fix it.
4. The engagement Fixed sum agreed up front, invoiced half at start and half on delivery. No open-ended hourly billing.
5. Delivery A written report you own, plus a call to walk your team through the method, the numbers and where they stop holding.

Physicist, then production BMS software

Most people working on on-board state estimation are control engineers who treat the cell as a transfer function. Most people who understand the electrochemistry do not write embedded estimators. I have done both, which is why I can say not only that an estimator drifts, but which degradation mechanism is causing it.

• Named inventor on four German patent filings assigned to Dr. Ing. h.c. F. Porsche AG, two of them granted.
• Nine peer-reviewed publications, including Nature Communications, JACS and three papers in Advanced Materials.
• Two years developing battery management software for production electric vehicles at Porsche Engineering.
• An eight-cell cycle-aging campaign of my own, which is where the figure above comes from.

Patents, publications and background →