Most lithium-ion cells are rated to operate from –40 to 80 °C, yet few ever experience the extremes for more than a few minutes. In real-world operation, temperature tends to converge within a much narrower band. This is typically between 30 and 50 °C. That reality should guide how battery models are trained if they are to provide insights that truly matter.
Why cells heat up in the first place
Ohmic and internal resistance losses
Every ampere driven through a cell encounters resistance in the electrodes and current collectors. The familiar I²R term results in heat generation.
Entropic heating from chemical reactions
Lithium intercalation is not perfectly reversible. Entropic effects contribute additional heating or cooling depending on whether the cell is charging or discharging, and on its local State of Charge (SoC).
These sources of heat are unavoidable. That’s why battery packs are often coupled with thermal management systems (whether forced-air or liquid cooling). With active control, excess heat is removed and the cell often stabilises between 30 and 50 °C. This leads to more consistent power delivery, slower degradation, and predictable voltage polarisation.
Train where the data live
A model built with uniform data coverage from –40 to 80 °C spreads accuracy too thin. It dilutes predictive power where it’s needed. By concentrating data points around common operating conditions, models become far more precise in forecasting capacity fade, resistance growth, and instantaneous State of Power (SoP). It also reduces testing time and cost. At the start of the design process it’s critical to understand the duty cycle. In motorsport and aviation, for example, packs are often pre-conditioned on the grid or in the hangar. These use cases focus on high performance, narrow temperature windows, and aggressive swings in depth of discharge. Automotive or drone applications, by contrast, may involve cold starts or passive cooling, where accuracy at low temperatures is critical, particularly in the first few minutes.
Validation at the limits is still required. Engineers need to know: can the battery start at –30 °C, and how fast will it warm into its nominal zone? Test data is therefore needed at the extremes to anchor the model, but focus most of the coverage where the battery will operate most frequently.
WLTP drive cycle of a battery tested at -5°C – which demonstrates self-heating and thermal management
While SoC reflects how much energy remains, SoP tells you whether the system can perform when demanded. Temperature governs both. A cell at 10 % SoC may supply full power at 35 °C, but the same cell at –20 °C might fall below its voltage limit. Models can answer the real question: will the battery deliver when it’s needed?
Focusing modelling efforts on realistic usage conditions, and validating only where it counts, leads to better-performing models, reduced development time, and more reliable system design. By aligning models with real-world behaviour, engineers can make faster, more confident decisions and unlock the full value of battery performance data.
About:Energy
A data company that offers software to enhance battery performance throughout its lifetime. Its platform, “The Voltt”, reduces time and costs in developing new electric products for various industries. Utilizing advanced battery testing methods, this platform facilitates cell selection, benchmarking, and system design. Catering to anyone aiming to improve battery performance metrics, it efficiently trims down cell selection times from months to minutes through its comprehensive cell library. Users can leverage this resource to benchmark their emerging cell chemistries, maintaining a competitive edge by evaluating rival performance. Providing cloud-based tools and downloadable models, “The Voltt” aids in optimizing system designs by offering insights into battery pack behaviour