v22.4 - 2022-04-30
- Added a casadi version of the IDKLU solver, which is used for
model.convert_to_format = "casadi"
(#2002)
- Corrected legend order in "plot_voltage_components.py", so each entry refers to the correct overpotential. (#2061)
- Remove old deprecation errors, including those in
parameter_values.py
that caused the simulation if, for example, the reaction rate is re-introduced manually (#2022)
- Changed domain-specific parameter names to a nested attribute, e.g.
param.c_n_max
is nowparam.n.c_max
(#2063)
v22.3 - 2022-03-31
- Added "Discharge energy [W.h]", which is the integral of the power in Watts, as an optional output. Set the option "calculate discharge energy" to "true" to get this output ("false" by default, since it can slow down some of the simple models) (#1969))
- Added an option "calculate heat source for isothermal models" to choose whether or not the heat generation terms are computed when running models with the option
thermal="isothermal"
(#1958)
- Fix bug where sensitivity calculation failed if len of
calculate_sensitivities
was less thaninputs
(#1897) - Fixed a bug in the eSOH variable calculation when OCV is given as data (#1975)
- Fixed a bug where isothermal models did not compute any heat source terms (#1958)
- Dropped support for Windows 32-bit architecture (#1964)
v22.2 - 2022-02-28
- Isothermal models now calculate heat source terms (but the temperature remains constant). The models now also account for current collector heating when
dimensionality=0
(#1929) - Added new models for power control and resistance control (#1917)
- Initial concentrations can now be provided as a function of
r
as well asx
(#1866)
- Fixed a bug where thermal submodels could not be used with half-cells (#1929)
- Parameters can now be imported from a directory having "pybamm" in its name (#1919)
scikit.odes
andSUNDIALS
can now be installed usingpybamm_install_odes
(#1916)
- The
domain
setter andauxiliary_domains
getter have been deprecated,domains
setter/getter should be used instead. Thedomain
getter is still active. We now recommend creating symbols withdomains={...}
instead ofdomain=..., auxiliary_domains={...}
, but the latter is not yet deprecated (#1866)
v22.1 - 2022-01-31
- Half-cell models can now be run with "surface form" (#1913)
- Added option for different kinetics on anode and cathode (#1913)
- Allow
pybamm.Solution.save_data()
to return a string if filename is None, and added json to_format option (#1909) - Added an option to force install compatible versions of jax and jaxlib if already installed using CLI (#1881)
- The
Symbol
nodes no longer subclassesanytree.NodeMixIn
. This removes some checks that were not really needed (#1912)
- Parameters can now be imported from any given path in
Windows
(#1900) - Fixed initial conditions for the EC SEI model (#1895)
- Fixed issue in extraction of sensitivites (#1894)
v21.12 - 2021-12-29
- Added new kinetics models for asymmetric Butler-Volmer, linear kinetics, and Marcus-Hush-Chidsey (#1858)
- Experiments can be set to terminate when a voltage is reached (across all steps) (#1832)
- Added cylindrical geometry and finite volume method (#1824)
PyBaMM
is now importable inLinux
systems wherejax
is already installed (#1874)- Simulations with drive cycles now support
initial_soc
(#1842) - Fixed bug in expression tree simplification (#1831)
- Solid tortuosity is now correctly calculated with Bruggeman coefficient of the respective electrode (#1773)
v21.11 - 2021-11-30
- The name of a parameter set can be passed to
ParameterValues
as a string, e.g.ParameterValues("Chen2020")
(#1822) - Added submodels for interface utilisation (#1821)
- Reformatted SEI growth models into a single submodel with conditionals (#1808)
- Stress-induced diffusion is now a separate model option instead of being automatically included when using the particle mechanics submodels (#1797)
Experiment
s with drive cycles can be solved (#1793)- Added surface area to volume ratio as a factor to the SEI equations (#1790)
- Half-cell SPM and SPMe have been implemented (#1731)
- Fixed
sympy
operators forArctan
andExponential
(#1786) - Fixed finite volume discretization in spherical polar coordinates (#1782)
- Fixed bug when using
Experiment
with a pouch cell model (#1707) - Fixed bug when using
Experiment
with a plating model (#1707) - Fixed hack for potentials in the SPMe model (#1707)
- The
chemistry
keyword argument inParameterValues
has been deprecated. UseParameterValues(chem)
instead ofParameterValues(chemistry=chem)
(#1822) - Raise error when trying to convert an
Interpolant
with the "pchip" interpolator to CasADI (#1791) - Raise error if
Concatenation
is used directly withVariable
objects (concatenation
should be used instead) (#1789) - Made jax, jaxlib and the PyBaMM JaxSolver optional (#1767, #1803)
v21.10 - 2021-10-31
- Summary variables can now be user-determined (#1760)
- Added
all_first_states
to theSolution
object for a simulation with experiment (#1759) - Added a new method (
create_gif
) inQuickPlot
,Simulation
andBatchStudy
to create a GIF of a simulation (#1754) - Added more examples for the
BatchStudy
class (#1747) - SEI models can now be included in the half-cell model (#1705)
- Half-cell model and lead-acid models can now be simulated with
Experiment
s (#1759) - Removed in-place modification of the solution objects by
QuickPlot
(#1747) - Fixed vector-vector multiplication bug that was causing errors in the SPM with constant voltage or power (#1735)
v21.9 - 2021-09-30
- Added thermal parameters (thermal conductivity, specific heat, etc.) to the
Ecker2015
parameter set from Zhao et al. (2018) and Hales et al. (2019) (#1683) - Added
plot_summary_variables
to plot and compare summary variables (#1678) - The DFN model can now be used directly (instead of
BasicDFNHalfCell
) to simulate a half-cell (#1600)
- Dropped support for Python 3.6 (#1696)
- The substring 'negative electrode' has been removed from variables related to SEI and lithium plating (e.g. 'Total negative electrode SEI thickness [m]' replaced by 'Total SEI thickness [m]') (#1654)
v21.08 - 2021-08-26
This release introduces:
- the switch to calendar versioning: from now on we will use year.month version number
- sensitivity analysis of solutions with respect to input parameters
- several new models, including many-particle and state-of-health models
- improvement on how CasADI solver's handle events, including a new "fast with events" mode
- several other new features, optimizations, and bug fixes, summarized below
- Added submodels and functionality for particle-size distributions in the DFN model, including an example notebook (#1602)
- Added UDDS and WLTC drive cycles (#1601)
- Added LG M50 (NMC811 and graphite + SiOx) parameter set from O'Regan 2021 (#1594)
pybamm.base_solver.solve
function can take a list of input parameters to calculate the sensitivities of the solution with respect to. Alternatively, it can be set toTrue
to calculate the sensitivities for all input parameters (#1552)- Added capability for
quaternary
domains (in addition toprimary
,secondary
andtertiary
), increasing the maximum number of domains that aSymbol
can have to 4. (#1580) - Tabs can now be placed at the bottom of the cell in 1+1D thermal models (#1581)
- Added temperature dependence on electrode electronic conductivity (#1570)
pybamm.base_solver.solve
function can take a list of input parameters to calculate the sensitivities of the solution with respect to. Alternatively, it can be set toTrue
to calculate the sensitivities for all input parameters (#1552)- Added a new lithium-ion model
MPM
or Many-Particle Model, with a distribution of particle sizes in each electrode. (#1529) - Added 2 new submodels for lithium transport in a size distribution of electrode particles: Fickian diffusion (
FickianSingleSizeDistribution
) and uniform concentration profile (FastSingleSizeDistribution
). (#1529) - Added a "particle size" domain to the default lithium-ion geometry, including plotting capabilities (
QuickPlot
) and processing of variables (ProcessedVariable
). (#1529) - Added fitted expressions for OCPs for the Chen2020 parameter set (#1526)
- Added
initial_soc
argument toSimualtion.solve
for specifying the initial SOC when solving a model (#1512) - Added
print_name
to some symbols (#1495, #1497) - Added Base Parameters class and SymPy in dependencies (#1495)
- Added a new "reaction-driven" model for LAM from Reniers et al (2019) (#1490)
- Some features ("loss of active material" and "particle mechanics") can now be specified separately for the negative electrode and positive electrode by passing a 2-tuple (#1490)
plot
andplot2D
now take and return a matplotlib Axis to allow for easier customization (#1472)ParameterValues.evaluate
can now return arrays to allow function parameters to be easily evaluated (#1472)- Added option to save only specific cycle numbers when simulating an
Experiment
(#1459) - Added capacity-based termination conditions when simulating an
Experiment
(#1459) - Added "summary variables" to track degradation over several cycles (#1459)
- Added
ElectrodeSOH
model for calculating capacities and stoichiometric limits (#1459) - Added Batch Study class (#1455)
- Added
ConcatenationVariable
, which is automatically created when variables are concatenated (#1453) - Added "fast with events" mode for the CasADi solver, which solves a model and finds events more efficiently than "safe" mode. As of PR #1450 this feature is still being tested and "safe" mode remains the default (#1450)
- Models that mostly use x-averaged quantities (SPM and SPMe) now use x-averaged degradation models (#1490)
- Improved how the CasADi solver's "safe" mode finds events (#1450)
- Perform more automatic simplifications of the expression tree (#1449)
- Reduce time taken to hash a sparse
Matrix
object (#1449)
- Fixed bug with
load_function
(#1675) - Updated documentation to include some previously missing functions, such as
erf
andtanh
(#1628) - Fixed reading citation file without closing (#1620)
- Porosity variation for SEI and plating models is calculated from the film thickness rather than from a separate ODE (#1617)
- Fixed a bug where the order of the indexing for the entries of variables discretised using FEM was incorrect (#1556)
- Fix broken module import for spyder when running a script twice (#1555)
- Fixed ElectrodeSOH model for multi-dimensional simulations (#1548)
- Removed the overly-restrictive check "each variable in the algebraic eqn keys must appear in the eqn" (#1510)
- Made parameters importable through pybamm (#1475)
- Refactored the
particle
submodel module, with the models having no size distribution now found inparticle.no_distribution
, and those with a size distribution inparticle.size_distribution
. Renamed submodels to indicate the transport model (Fickian diffusion, polynomial profile) and if they are "x-averaged". E.g.,FickianManyParticles
andFickianSingleParticle
are nowno_distribution.FickianDiffusion
andno_distribution.XAveragedFickianDiffusion
(#1602) - Changed sensitivity API. Removed
ProcessedSymbolicVariable
, all sensitivity now handled within the solvers andProcessedVariable
() (#1552) - The
Yang2017
parameter set has been removed as the complete parameter set is not publicly available in the literature (#1577) - Changed how options are specified for the "loss of active material" and "particle cracking" submodels. "loss of active material" can now be one of "none", "stress-driven", or "reaction-driven", or a 2-tuple for different options in negative and positive electrode. Similarly "particle cracking" (now called "particle mechanics") can now be "none", "swelling only", "swelling and cracking", or a 2-tuple (#1490)
- Changed the variable in the full diffusion model from "Electrolyte concentration" to "Porosity times concentration" (#1476)
- Renamed
lithium-ion
folder tolithium_ion
andlead-acid
folder tolead_acid
in parameters (#1464)
v0.4.0 - 2021-03-28
This release introduces:
- several new models, including reversible and irreversible plating submodels, submodels for loss of active material, Yang et al.'s (2017) coupled SEI/plating/pore clogging model, and the Newman-Tobias model
- internal optimizations for solving models, particularly for simulating experiments, with more accurate event detection and more efficient numerical methods and post-processing
- parallel solutions of a model with different inputs
- a cleaner installation process for Mac when installing from PyPI, no longer requiring a Homebrew installation of Sundials
- improved plotting functionality, including adding a new 'voltage component' plot
- several other new features, optimizations, and bug fixes, summarized below
- Added
NewmanTobias
li-ion battery model (#1423) - Added
plot_voltage_components
to easily plot the component overpotentials that make up the voltage (#1419) - Made
QuickPlot
more customizable and added an example (#1419) Solution
objects can now be created by stepping different models (#1408)- Added Yang et al 2017 model that couples irreversible lithium plating, SEI growth and change in porosity which produces a transition from linear to nonlinear degradation pattern of lithium-ion battery over extended cycles(#1398)
- Added support for Python 3.9 and dropped support for Python 3.6. Python 3.6 may still work but is now untested (#1370)
- Added the electrolyte overpotential and Ohmic losses for full conductivity, including surface form (#1350)
- Added functionality to
Citations
to print formatted citations (#1340) - Updated the way events are handled in
CasadiSolver
for more accurate event location (#1328) - Added error message if initial conditions are outside the bounds of a variable (#1326)
- Added temperature dependence to density, heat capacity and thermal conductivity (#1323)
- Added temperature dependence to the transference number (
t_plus
) (#1317) - Added new functionality for
Interpolant
(#1312) - Added option to express experiments (and extract solutions) in terms of cycles of operating condition (#1309)
- The event time and state are now returned as part of
Solution.t
andSolution.y
so that the event is accurately captured in the returned solution (#1300) - Added reversible and irreversible lithium plating models (#1287)
- Reformatted the
BasicDFNHalfCell
to be consistent with the other models (#1282) - Added option to make the total interfacial current density a state (#1280)
- Added functionality to initialize a model using the solution from another model (#1278)
- Added submodels for active material (#1262)
- Updated solvers' method
solve()
so it can take a list of inputs dictionaries as theinputs
keyword argument. In this case the model is solved for each input set in the list, and a list of solutions mapping the set of inputs to the solutions is returned. Note thatsolve()
can still take a single dictionary as theinputs
keyword argument. In this case the behaviour is unchanged compared to previous versions.(#1261) - Added composite surface form electrolyte models:
CompositeDifferential
andCompositeAlgebraic
(#1207)
- Improved the way an
Experiment
is simulated to reduce solve time (at the cost of slightly higher set-up time) (#1408) - Add script and workflow to automatically update parameter_sets.py docstrings (#1371)
- Add URLs checker in workflows (#1347)
- The
Solution
class now only creates the concatenatedy
when the user asks for it. This is an optimization step as the concatenation can be slow, especially with larger experiments (#1331) - If solver method
solve()
is passed a list of inputs as theinputs
keyword argument, the resolution of the model for each input set is spread across several Python processes, usually running in parallel on different processors. The default number of processes is the number of processors available.solve()
takes a new keyword argumentnproc
which can be used to set this number a manually. - Variables are now post-processed using CasADi (#1316)
- Operations such as
1*x
and0+x
now directly returnx
(#1252)
- Fixed a bug on the boundary conditions of
FickianSingleParticle
andFickianManyParticles
to ensure mass is conserved (#1421) - Fixed a bug where the
PolynomialSingleParticle
submodel gave incorrect results with "dimensionality" equal to 2 (#1411) - Fixed a bug where volume averaging in 0D gave the wrong result (#1411)
- Fixed a sign error in the positive electrode ohmic losses (#1407)
- Fixed the formulation of the EC reaction SEI model (#1397)
- Simulations now stop when an experiment becomes infeasible (#1395)
- Added a check for domains in
Concatenation
(#1368) - Differentiation now works even when the differentiation variable is a constant (#1294)
- Fixed a bug where the event time and state were no longer returned as part of the solution (#1344)
- Fixed a bug in
CasadiSolver
safe mode which crashed when there were extrapolation events but no termination events (#1321) - When an
Interpolant
is extrapolated an error is raised forCasadiSolver
(and a warning is raised for the other solvers) (#1315) - Fixed
Simulation
andmodel.new_copy
to fix a bug where changes to the model were overwritten (#1278)
- Removed
Simplification
class and.simplify()
function (#1369) - All example notebooks in PyBaMM's GitHub repository must now include the command
pybamm.print_citations()
, otherwise the tests will fail. This is to encourage people to use this command to cite the relevant papers (#1340) - Notation has been homogenised to use positive and negative electrode (instead of cathode and anode). This applies to the parameter folders (now called
'positive_electrodes'
and'negative_electrodes'
) and the options ofactive_material
andparticle_cracking
submodels (now called'positive'
and'negative'
) (#1337) Interpolant
now takesx
andy
instead of a singledata
entry (#1312)- Boolean model options ('sei porosity change', 'convection') must now be given in string format ('true' or 'false' instead of True or False) (#1280)
- Operations such as
1*x
and0+x
now directly returnx
. This can be bypassed by explicitly creating the binary operators, e.g.pybamm.Multiplication(1, x)
(#1252) 'Cell capacity [A.h]'
has been renamed to'Nominal cell capacity [A.h]'
.'Cell capacity [A.h]'
will be deprecated in the next release. (#1352)
v0.3.0 - 2020-12-01
This release introduces a new aging model for particle mechanics, a new reduced-order model (TSPMe), and a parameter set for A123 LFP cells. Additionally, there have been several backend optimizations to speed up model creation and solving, and other minor features and bug fixes.
- Added a submodel for particle mechanics (#1232)
- Added a notebook on how to speed up the solver and handle instabilities (#1223)
- Improve string printing of
BinaryOperator
,Function
, andConcatenation
objects (#1223) - Added
Solution.integration_time
, which is the time taken just by the integration subroutine, without extra setups (#1223) - Added parameter set for an A123 LFP cell (#1209)
- Added variables related to equivalent circuit models (#1204)
- Added the
Integrated
electrolyte conductivity submodel (#1188) - Added an example script to check conservation of lithium (#1186)
- Added
erf
anderfc
functions (#1184)
- Add (optional) smooth approximations for the
Minimum
,Maximum
,Heaviside
, andAbsoluteValue
operators (#1223) - Avoid unnecessary repeated computations in the solvers (#1222)
- Rewrite
Symbol.is_constant
to be more efficient (#1222) - Cache shape and size calculations (#1222)
- Only instantiate the geometric, electrical and thermal parameter classes once (#1222)
- Quickplot now works when timescale or lengthscale is a function of an input parameter (#1234)
- Fix bug that was slowing down creation of the EC reaction SEI submodel (#1227)
- Add missing separator thermal parameters for the Ecker parameter set (#1226)
- Make sure simulation solves when evaluated timescale is a function of an input parameter (#1218)
- Raise error if saving to MATLAB with variable names that MATLAB can't read, and give option of providing alternative variable names (#1206)
- Raise error if the boundary condition at the origin in a spherical domain is other than no-flux (#1175)
- Fix boundary conditions at r = 0 for Creating Models notebooks (#1173)
- The parameters "Positive/Negative particle distribution in x" and "Positive/Negative surface area to volume ratio distribution in x" have been deprecated. Instead, users can provide "Positive/Negative particle radius [m]" and "Positive/Negative surface area to volume ratio [m-1]" directly as functions of through-cell position (x [m]) (#1237)
v0.2.4 - 2020-09-07
This release adds new operators for more complex models, some basic sensitivity analysis, and a spectral volumes spatial method, as well as some small bug fixes.
- Added variables which track the total amount of lithium in the system (#1136)
- Added
Upwind
andDownwind
operators for convection (#1134) - Added Getting Started notebook on solver options and changing the mesh. Also added a notebook detailing the different thermal options, and a notebook explaining the steps that occur behind the scenes in the
Simulation
class (#1131) - Added particle submodel that use a polynomial approximation to the concentration within the electrode particles (#1130)
- Added
Modulo
,Floor
andCeiling
operators (#1121) - Added DFN model for a half cell (#1121)
- Automatically compute surface area to volume ratio based on particle shape for li-ion models (#1120)
- Added "R-averaged particle concentration" variables (#1118)
- Added support for sensitivity calculations to the casadi solver (#1109)
- Added support for index 1 semi-explicit dae equations and sensitivity calculations to JAX BDF solver (#1107)
- Allowed keyword arguments to be passed to
Simulation.plot()
(#1099) - Added the Spectral Volumes spatial method and the submesh that it works with (#900)
- Fixed bug where some parameters were not being set by the
EcReactionLimited
SEI model (#1136) - Fixed bug on electrolyte potential for
BasicDFNHalfCell
(#1133) - Fixed
r_average
to work withSecondaryBroadcast
(#1118) - Fixed finite volume discretisation of spherical integrals (#1118)
t_eval
now gets changed to alinspace
if a list of length 2 is passed (#1113)- Fixed bug when setting a function with an
InputParameter
(#1111)
- The "fast diffusion" particle option has been renamed "uniform profile" (#1130)
- The modules containing standard parameters are now classes so they can take options
(e.g.
standard_parameters_lithium_ion
is nowLithiumIonParameters
) (#1120) - Renamed
quick_plot_vars
tooutput_variables
inSimulation
to be consistent withQuickPlot
. Passingquick_plot_vars
toSimulation.plot()
has been deprecated andoutput_variables
should be passed instead (#1099)
v0.2.3 - 2020-07-01
This release enables the use of Google Colab for running example notebooks, and adds some small new features and bug fixes.
- Added JAX evaluator, and ODE solver (#1038)
- Reformatted Getting Started notebooks (#1083)
- Reformatted Landesfeind electrolytes (#1064)
- Adapted examples to be run in Google Colab (#1061)
- Added some new solvers for algebraic models (#1059)
- Added
length_scales
attribute to models (#1058) - Added averaging in secondary dimensions (#1057)
- Added SEI reaction based on Yang et. al. 2017 and reduction in porosity (#1009)
- Reformatted CasADi "safe" mode to deal with events better (#1089)
- Fixed a bug in
InterstitialDiffusionLimited
(#1097) - Fixed
Simulation
to keep different copies of the model so that parameters can be changed between simulations (#1090) - Fixed
model.new_copy()
to keep custom submodels (#1090) - 2D processed variables can now be evaluated at the domain boundaries (#1088)
- Update the default variable points to better capture behaviour in the solid particles in li-ion models (#1081)
- Fix
QuickPlot
to display variables discretised by FEM (in y-z) properly (#1078) - Add length scales to
EffectiveResistance
models (#1071) - Allowed for pybamm functions exp, sin, cos, sqrt to be used in expression trees that are converted to casadi format (#1067)
- Fix a bug where variables that depend on y and z were transposed in
QuickPlot
(#1055)
Simulation.specs
andSimulation.set_defaults
have been deprecated. Users should create a newSimulation
object for each different case instead (#1090)- The solution times
t_eval
must now be provided toSimulation.solve()
when not using an experiment or prescribing the current using drive cycle data (#1086)
v0.2.2 - 2020-06-01
New SEI models, simplification of submodel structure, as well as optimisations and general bug fixes.
- Reformatted
Geometry
andMesh
classes (#1032) - Added arbitrary geometry to the lumped thermal model (#718)
- Allowed
ProcessedVariable
to handle cases wherelen(solution.t)=1
(#1020) - Added
BackwardIndefiniteIntegral
symbol (#1014) - Added
plot
andplot2D
to enable easy plotting ofpybamm.Array
objects (#1008) - Updated effective current collector models and added example notebook (#1007)
- Added SEI film resistance as an option (#994)
- Added
parameters
attribute topybamm.BaseModel
andpybamm.Geometry
that lists all of the required parameters (#993) - Added tab, edge, and surface cooling (#965)
- Added functionality to solver to automatically discretise a 0D model (#947)
- Added sensitivity to
CasadiAlgebraicSolver
(#940) - Added
ProcessedSymbolicVariable
class, which can handle symbolic variables (i.e. variables for which the inputs are symbolic) (#940) - Made
QuickPlot
compatible with Google Colab (#935) - Added
BasicFull
model for lead-acid (#932) - Added 'arctan' function (#973)
- Implementing the use of GitHub Actions for CI (#855)
- Changed default solver for DAE models to
CasadiSolver
(#978) - Added some extra simplifications to the expression tree (#971)
- Changed the behaviour of "safe" mode in
CasadiSolver
(#956) - Sped up model building (#927)
- Changed default solver for lead-acid to
CasadiSolver
(#927)
- Fix a bug where slider plots do not update properly in notebooks (#1041)
- Fix storing and plotting external variables in the solution (#1026)
- Fix running a simulation with a model that is already discretized (#1025)
- Fix CI not triggering for PR. (#1013)
- Fix schedule testing running too often. (#1010)
- Fix doctests failing due to mismatch in unsorted output.(#990)
- Added extra checks when creating a model, for clearer errors (#971)
- Fixed
Interpolant
ids to allow processing (#962) - Fixed a bug in the initial conditions of the potential pair model (#954)
- Changed simulation attributes to assign copies rather than the objects themselves (#952)
- Added default values to base model so that it works with the
Simulation
class (#952) - Fixed solver to recompute initial conditions when inputs are changed (#951)
- Reformatted thermal submodels (#938)
- Reformatted electrolyte submodels (#927)
- Reformatted convection submodels (#635)
- Geometry should no longer be given keys 'primary' or 'secondary' (#1032)
- Calls to
ProcessedVariable
objects are now made using dimensional time and space (#1028) - For variables discretised using finite elements the result returned by calling
ProcessedVariable
is now transposed (#1020) - Renamed "surface area density" to "surface area to volume ratio" (#975)
- Replaced "reaction rate" with "exchange-current density" (#975)
- Changed the implementation of reactions in submodels (#948)
- Removed some inputs like
T_inf
,R_g
and activation energies to some of the standard function parameters. This is because each of those inputs is specific to a particular function (e.g. the reference temperature at which the function was measured). To change a property such as the activation energy, users should create a new function, specifying the relevant property as aParameter
orInputParameter
(#942) - The thermal option 'xyz-lumped' has been removed. The option 'thermal current collector' has also been removed (#938)
- The 'C-rate' parameter has been deprecated. Use 'Current function [A]' instead. The cell capacity can be accessed as 'Cell capacity [A.h]', and used to calculate current from C-rate (#952)
v0.2.1 - 2020-03-31
New expression tree node types, models, parameter sets and solvers, as well as general bug fixes and new examples.
- Store variable slices in model for inspection (#925)
- Added LiNiCoO2 parameter set from Ecker et. al. (#922)
- Made t_plus (optionally) a function of electrolyte concentration, and added (1 + dlnf/dlnc) to models (#921)
- Added
DummySolver
for empty models (#915) - Added functionality to broadcast to edges (#891)
- Reformatted and cleaned up
QuickPlot
(#886) - Added thermal effects to lead-acid models (#885)
- Added a helper function for info on function parameters (#881)
- Added additional notebooks showing how to create and compare models (#877)
- Added
Minimum
,Maximum
andSign
operators (#876) - Added a search feature to
FuzzyDict
(#875) - Add ambient temperature as a function of time (#872)
- Added
CasadiAlgebraicSolver
for solving algebraic systems with CasADi (#868) - Added electrolyte functions from Landesfeind (#860)
- Add new symbols
VariableDot
, representing the derivative of a variable wrt time, andStateVectorDot
, representing the derivative of a state vector wrt time (#858)
- Filter out discontinuities that occur after solve times (#941)
- Fixed tight layout for QuickPlot in jupyter notebooks (#930)
- Fixed bug raised if function returns a scalar (#919)
- Fixed event handling in
ScipySolver
(#905) - Made input handling clearer in solvers (#905)
- Updated Getting started notebook 2 (#903)
- Reformatted external circuit submodels (#879)
- Some bug fixes to generalize specifying models that aren't battery models, see #846
- Reformatted interface submodels to be more readable (#866)
- Removed double-counted "number of electrodes connected in parallel" from simulation (#864)
- Changed keyword argument
u
for inputs (when evaluating an object) toinputs
(#905) - Removed "set external temperature" and "set external potential" options. Use "external submodels" option instead (#862)
v0.2.0 - 2020-02-26
This release introduces many new features and optimizations. All models can now be solved using the pip installation - in particular, the DFN can be solved in around 0.1s. Other highlights include an improved user interface, simulations of experimental protocols (GITT, CCCV, etc), new parameter sets for NCA and LGM50, drive cycles, "input parameters" and "external variables" for quickly solving models with different parameter values and coupling with external software, and general bug fixes and optimizations.
- Added LG M50 parameter set from Chen 2020 (#854)
- Changed rootfinding algorithm to CasADi, scipy.optimize.root still accessible as an option (#844)
- Added capacitance effects to lithium-ion models (#842)
- Added NCA parameter set (#824)
- Added functionality to
Solution
that automatically getst_eval
from the data when simulating drive cycles and performs checks to ensure the output has the required resolution to accurately capture the input current (#819) - Added
Citations
object to print references when specific functionality is used (#818) - Updated
Solution
to allow exporting to matlab and csv formats (#811) - Allow porosity to vary in space (#809)
- Added functionality to solve DAE models with non-smooth current inputs (#808)
- Added functionality to simulate experiments and testing protocols (#807)
- Added fuzzy string matching for parameters and variables (#796)
- Changed ParameterValues to raise an error when a parameter that wasn't previously defined is updated (#796)
- Added some basic models (BasicSPM and BasicDFN) in order to clearly demonstrate the PyBaMM model structure for battery models (#795)
- Allow initial conditions in the particle to depend on x (#786)
- Added the harmonic mean to the Finite Volume method, which is now used when computing fluxes (#783)
- Refactored
Solution
to make it a dictionary that contains all of the solution variables. This automatically createsProcessedVariable
objects when required, so that the solution can be obtained much more easily. (#781) - Added notebook to explain broadcasts (#776)
- Added a step to discretisation that automatically compute the inverse of the mass matrix of the differential part of the problem so that the underlying DAEs can be provided in semi-explicit form, as required by the CasADi solver (#769)
- Added the gradient operation for the Finite Element Method (#767)
- Added
InputParameter
node for quickly changing parameter values (#752) - Added submodels for operating modes other than current-controlled (#751)
- Changed finite volume discretisation to use exact values provided by Neumann boundary conditions when computing the gradient instead of adding ghost nodes(#748)
- Added optional R(x) distribution in particle models (#745)
- Generalized importing of external variables (#728)
- Separated active and inactive material volume fractions (#726)
- Added submodels for tortuosity (#726)
- Simplified the interface for setting current functions (#723)
- Added Heaviside operator (#723)
- New extrapolation methods (#707)
- Added some "Getting Started" documentation (#703)
- Allow abs tolerance to be set by variable for IDA KLU solver (#700)
- Added Simulation class (#693) with load/save functionality (#732)
- Added interface to CasADi solver (#687, #691, #714). This makes the SUNDIALS DAE solvers (Scikits and KLU) truly optional (though IDA KLU is recommended for solving the DFN).
- Added option to use CasADi's Algorithmic Differentiation framework to calculate Jacobians (#687)
- Added method to evaluate parameters more easily (#669)
- Added
Jacobian
class to reuse known Jacobians of expressions (#665) - Added
Interpolant
class to interpolate experimental data (e.g. OCP curves) (#661) - Added interface (via pybind11) to sundials with the IDA KLU sparse linear solver (#657)
- Allowed parameters to be set by material or by specifying a particular paper (#647)
- Set relative and absolute tolerances independently in solvers (#645)
- Added basic method to allow (a part of) the State Vector to be updated with results obtained from another solution or package (#624)
- Added some non-uniform meshes in 1D and 2D (#617)
- Now simplifying objects that are constant as soon as they are created (#801)
- Simplified solver interface (#800)
- Added caching for shape evaluation, used during discretisation (#780)
- Added an option to skip model checks during discretisation, which could be slow for large models (#739)
- Use CasADi's automatic differentation algorithms by default when solving a model (#714)
- Avoid re-checking size when making a copy of an
Index
object (#656) - Avoid recalculating
_evaluation_array
when making a copy of aStateVector
object (#653)
- Fixed a bug where current loaded from data was incorrectly scaled with the cell capacity (#852)
- Moved evaluation of initial conditions to solver (#839)
- Fixed a bug where the first line of the data wasn't loaded when parameters are loaded from data (#819)
- Made
graphviz
an optional dependency (#810) - Fixed examples to run with basic pip installation (#800)
- Added events for CasADi solver when stepping (#800)
- Improved implementation of broadcasts (#776)
- Fixed a bug which meant that the Ohmic heating in the current collectors was incorrect if using the Finite Element Method (#767)
- Improved automatic broadcasting (#747)
- Fixed bug with wrong temperature in initial conditions (#737)
- Improved flexibility of parameter values so that parameters (such as diffusivity or current) can be set as functions or scalars (#723)
- Fixed a bug where boundary conditions were sometimes handled incorrectly in 1+1D models (#713)
- Corrected a sign error in Dirichlet boundary conditions in the Finite Element Method (#706)
- Passed the correct dimensional temperature to open circuit potential (#702)
- Added missing temperature dependence in electrolyte and interface submodels (#698)
- Fixed differentiation of functions that have more than one argument (#687)
- Added warning if
ProcessedVariable
is called outside its interpolation range (#681) - Updated installation instructions for Mac OS (#680)
- Improved the way
ProcessedVariable
objects are created in higher dimensions (#581)
- Time for solver should now be given in seconds (#832)
- Model events are now represented as a list of
pybamm.Event
(#759 - Removed
ParameterValues.update_model
, whose functionality is now replaced byInputParameter
(#801) - Removed
Outer
andKron
nodes as no longer used (#777) - Moved
results
to separate repositories (#761) - The parameters "Bruggeman coefficient" must now be specified separately as "Bruggeman coefficient (electrolyte)" and "Bruggeman coefficient (electrode)"
- The current classes (
GetConstantCurrent
,GetUserCurrent
andGetUserData
) have now been removed. Please refer to thechange-input-current
notebook for information on how to specify an input current - Parameter functions must now use pybamm functions instead of numpy functions (e.g.
pybamm.exp
instead ofnumpy.exp
), as these are then used to construct the expression tree directly. Generally, pybamm syntax follows numpy syntax; please get in touch if a function you need is missing. - The current must now be updated by changing "Current function [A]" or "C-rate" instead of "Typical current [A]"
v0.1.0 - 2019-10-08
This is the first official version of PyBaMM. Please note that PyBaMM in still under active development, and so the API may change in the future.
- Single Particle Model (SPM)
- Single Particle Model with electrolyte (SPMe)
- Doyle-Fuller-Newman (DFN) model
with the following optional physics:
- Thermal effects
- Fast diffusion in particles
- 2+1D (pouch cell)
- Leading-Order Quasi-Static model
- First-Order Quasi-Static model
- Composite model
- Full model
with the following optional physics:
- Hydrolysis side reaction
- Capacitance effects
- 2+1D
- Finite Volume (1D only)
- Finite Element (scikit, 2D only)
- Scipy
- Scikits ODE
- Scikits DAE
- IDA KLU sparse linear solver (Sundials)
- Algebraic (root-finding)