How to master battery aging

A battery aging model predicts how a cell loses capacity and power over time and use. Correlative models fit measured data and work where data exists. A physics-based aging model identifies the mechanisms behind the decline, so it can simulate aged cells under conditions you never tested. However, the goal is not simply to model aging. The goal is to release the finished battery system for series production as soon as possible, without any aging issue. That only works with a combination of tests and models, because you cannot test for as long as your statements have to hold. That is why Batemo's approach combines testing and modeling.

Cycle aging comes from charging and discharging. Calendar aging happens while the cell sits, and depends on temperature, state of charge and time. In a product that stands still most of the day, calendar aging can dominate. A test plan has to cover both, because the operating strategy that helps one does not always help the other.

Aging tests take months, often years, because the cell has to live through the wear you want to predict. Accelerated tests shorten this but change the conditions. Typical test matrices try to finish the tests within six months, which brings its own problems. To speed things up, a first estimate from Batemo's aging database is available right away, and Batemo can put your results in context. On that basis, we propose tests and build the best connection between test and model, so you reach series release earlier.

Only with care. Methods based on correlation are accurate where data was fitted. The range you need is usually the one beyond the data. A model built on the physical processes stays meaningful outside the tested points. Key scenarios must be checked by a test, because anything else is unreliable.

Six drivers matter: the temperature range, time at a high state of charge, depth of discharge, charge rate, discharge rate and cycle count. The calculator on this page rates them as an aging risk from very low to very high for your operating strategy. It is a screening, not a lifetime prediction: it tells you which driver to look at first.

Across 198 validated cells the median model deviation is 0.8 %, measured against validation experiments. Batemo backs its models with a guarantee: if you can show us another battery cell model that is more accurate, we will refund your money. We publish the accuracy of every model on its own cell page on this website, and for every model we provide Excel reports that show its accuracy under the validation conditions in detail.