Battery cycle-calendar life model updated to use sum not min#1239
Conversation
|
@brtietz The tests are failing because the error between Old Mac/Linux/Windows vs New Mac is greater than the abs tolerance of 0.01 for some of the battery tests. Mind if I increase the tol for this particular test? |
Increasing the tol seems good to me. Maybe 0.1? |
Yep. Tests are passing now |
brtietz
left a comment
There was a problem hiding this comment.
Results look good, tests all change in the direction that we expected. Do we need to feed any of this back to Kandler and Paul?
I don't think it's necessary |
I modified the cycle calendar model to take the sum instead of the min. Here are the results of 25-yr battery capacity percent (replacements are off) across the different chemistries and models. Overall, there are 6675 cycles. Average DOD is ~50%
The table models have more degradation now.
However, whereas in the Old column, the NMC/Gr table and algebraic model both had more degradation than LFP, in the New version, the LFP table has less degradation than NMC/Gr Alg. The LMO/LTO table doesn’t change because the calendar loss is 0.
This is fine because the calendar-cycle models aren’t comparable to the algebraic ones—apples and oranges.
References (added 3/30/2026):
Gaspar, P et al. (2022). Machine-Learning Assisted Identification of Accurate Battery Lifetime Models with Uncertainty. https://doi.org/10.1149/1945-7111/ac86a8.
Gaspar, P et al. (2023). Degradation and modeling of large-format commercial lithium-ion cells as a function of chemistry, design, and aging conditions. https://doi.org/10.1016/j.est.2023.109042