Choosing an Aging Model¶

simses ships three (calendar, cyclic) aging model pairs today, each fitted to a specific commercial cell. The table below is the quick chooser; each pair is then documented in its own section with its underlying source, tested stressor envelope, and composition example.

Who this is for

Users running multi-year studies where aging matters. Aging in simses is a choice — the electrical cell model and the aging model are composed independently, so you can pair any degradation pair with any CellType that shares the same chemistry family. For the framework itself, see Degradation.

Comparison¶

Pair	Chemistry	Tested cell	Format	Capacity	Trained DoD (cycling)	Trained T (cycling)	Ships as default of
Naumann	LFP	Sony US26650FTC1	Cylindrical 26650	3 Ah	10 – 90 %	0 – 60 °C	`SonyLFP`
Gasper LFP\|Gr	LFP (large-format)	Anonymised, ~2019 vintage	Prismatic	250 Ah	80 – 100 %	10 – 45 °C	— (compose explicitly)
Gasper NMC\|Gr B1	NMC (large-format)	Anonymised, ~2020 vintage	Pouch	50 Ah	80 – 100 %	10 – 45 °C	— (compose explicitly)

All three pairs share simses's CalendarDegradation + CyclicDegradation Protocol interface, so they slot interchangeably into a DegradationModel.

Naumann (Sony LFP)¶

The default aging pair shipped with SonyLFP. Semi-empirical fits to accelerated-aging tests on the Sony/Murata US26650FTC1 3 Ah cylindrical LFP cell:

Calendar — √t stress-factor with Arrhenius temperature dependence and a cubic SOC term.
Cyclic — √FEC stress-factor with a cubic DoD term and a linear C-rate term.

Both use exact virtual-time / virtual-FEC continuation so they stay correct under varying operating conditions. Best-suited to small-format cylindrical LFP and to any study where you want simses's ready-made "aging on" behaviour via Battery(..., degradation=True). See Degradation concept for the worked walkthrough.

Sources: Naumann et al., J. Energy Storage 17 (2018); J. Power Sources 451 (2020).

from simses.battery import Battery
from simses.model.cell.sony_lfp import SonyLFP

battery = Battery(
    cell=SonyLFP(),
    circuit=(13, 1),
    initial_states={"start_soc": 0.5, "start_T": 25.0},
    degradation=True,  # picks up the Naumann pair automatically
)

Gasper LFP|Gr¶

A large-format prismatic LFP capacity-fade model from Gasper et al. 2023 (J. Energy Storage 73, 109042), ported one-to-one from NREL's BLAST-Lite. Semi-empirical: an Arrhenius calendar term with anode-potential SoC dependence, and a low-order semi-empirical cyclic term. Behaves cleanly across the full DoD range (LFP chemistry is intrinsically DoD-insensitive in this fit).

Applicability caveat

The fit corresponds to a specific ~2019-vintage 250 Ah prismatic LFP cell (anonymised in the paper). Modern (2023 +) large-format LFP cells for stationary storage — CATL LF280K, EVE LF280K, BYD blade — typically show 2 – 3× longer cycle life and comparable calendar life. Treat quantitative EOL predictions from this model as conservative lower bounds. The shape of the aging response (SoC / T / DoD / C-rate sensitivities) remains useful for aging-aware operation studies.

No resistance model

Gasper et al. 2023 publishes capacity-fade equations only. update_resistance() returns 0 on both legs — state.soh_R stays at 1.0 for the whole simulation. If your study depends on resistance rise, combine with a separate resistance-only calendar or cyclic sub-model, or use the Naumann pair.

Recommended pairing. Compose with the existing SonyLFP electrical model as an LFP-chemistry proxy. OCV, hysteresis, and entropy are chemistry properties and transfer cleanly across cell formats; capacity is scaled to system level via circuit=(s, p). Users who need a genuinely different capacity or a different R_int(SoC, T) map should subclass SonyLFP and override — see Extending Cell Models.

from simses.battery import Battery
from simses.degradation import DegradationModel
from simses.model.cell.sony_lfp import SonyLFP
from simses.model.degradation.gasper_lfp_gr_calendar import GasperLFPGrCalendar
from simses.model.degradation.gasper_lfp_gr_cyclic import GasperLFPGrCyclic

battery = Battery(
    cell=SonyLFP(),  # LFP-chemistry electrical proxy
    circuit=(1, 100),  # scale to system size
    initial_states={"start_soc": 0.5, "start_T": 25.0},
    degradation=DegradationModel(
        calendar=GasperLFPGrCalendar(),
        cyclic=GasperLFPGrCyclic(),
        initial_soc=0.5,
    ),
)

Gasper NMC|Gr B1¶

A large-format pouch NMC capacity-fade model from the same Gasper et al. 2023 study. Compact single-term multiplicative power law in DoD, T, and C-rate — the simplest and best-behaved of the paper's four NMC chemistries. Monotonic in every stressor, no abs() clipping, extrapolates cleanly outside the [0.8, 1] DoD training band.

Applicability caveat

The fit corresponds to a specific ~2020-vintage 50 Ah pouch NMC|Gr cell (anonymised in the paper). Manufacturer-specific behaviour of modern large-format NMC|Gr cells will differ; treat quantitative EOL predictions from this model as approximate.

No resistance model

Same as LFP|Gr — the Gasper paper publishes capacity fade only. update_resistance() returns 0 on both legs.

Recommended pairing. Compose with the existing Samsung94AhNMC electrical model as an NMC-chemistry proxy.

from simses.battery import Battery
from simses.degradation import DegradationModel
from simses.model.cell.samsung94Ah_nmc import Samsung94AhNMC
from simses.model.degradation.gasper_nmc_gr_b1_calendar import GasperNMCGrB1Calendar
from simses.model.degradation.gasper_nmc_gr_b1_cyclic import GasperNMCGrB1Cyclic

battery = Battery(
    cell=Samsung94AhNMC(),  # NMC-chemistry electrical proxy
    circuit=(96, 1),
    initial_states={"start_soc": 0.5, "start_T": 25.0},
    degradation=DegradationModel(
        calendar=GasperNMCGrB1Calendar(),
        cyclic=GasperNMCGrB1Cyclic(),
        initial_soc=0.5,
    ),
)

Why "electrical model as a chemistry proxy"?¶

The Gasper cells in the paper are anonymised — the authors publish the aging fits but not the OCV curves, R_int maps, hysteresis, thermal properties, or mass. Building dedicated GasperLFPGrCell / GasperNMCGrB1Cell classes would require inventing values for everything except capacity, nominal voltage, and a single scalar DCIR from Table 1 of the paper. simses deliberately avoids that: the aging model is exported as a pair of Protocol implementations, and users pair it with whichever electrical cell model most closely matches the chemistry of interest.

The physical rationale is that OCV(SoC), hysteresis, and entropy are chemistry properties — they're a function of the electrode materials, not of the cell format. The LFP OCV plateau near 3.3 V, the entropic-coefficient signature, the small hysteresis at 50 % SOC — all of these transfer cleanly from a 3 Ah Sony cylindrical to a 250 Ah prismatic LFP cell of the same chemistry family. What doesn't transfer is R_int in Ohms (scales with capacity and format), max C-rates, thermal mass, and cell format geometry. If any of those matter to your study, subclass the electrical cell class and override the relevant properties.

Attribution¶

The two Gasper pairs are ported from NREL's BLAST-Lite Python library (BSD-3-Clause). Original implementations by Paul Gasper (NREL). Each ported file preserves the BLAST-Lite license notice in-file as required by BSD-3.

Extending¶

To add a new aging model — either a fresh fit or a refit of an existing functional form against your own data — see Extending Degradation.