Simulations with O’Regan 2022 parameter set not working at 2C 0°C, 2C 10°C and WLTP 0°C

PyBaMM Q&A Models
1

Hello,
I am working with the O’Regan parameter set to get temperature prediction for LGM50 to compare the predictions with the model that I have developed. I am using this code as the reference:

https://docs.pybamm.org/en/latest/source/examples/notebooks/models/simulating-ORegan-2022-parameter-set.html

data_file = loadmat('WLTP_2_CYCLES_FULL_SET.mat')
Crate, Temp = "2C",0
t,T,T_inf,I,SOC_0 =load_data(Crate,Temp)

options = {"thermal":"lumped","dimensionality":0,"cell geometry":"arbitrary"}
model = pybamm.lithium_ion.DFN(options=options)
#O'Regan parameter set
param = pybamm.ParameterValues("ORegan2022")
solver = pybamm.IDAKLUSolver()

param["Total heat transfer coefficient [W.m-2.K-1]"] = 14
param["Ambient temperature [K]"] = T_inf
param["Initial temperature [K]"] = T_inf
param["Current function [A]"]    = I

var_pts = {"x_n": 30, "x_s": 30, "x_p": 30, "r_n": 40, "r_p": 40}

submesh_types = model.default_submesh_types
submesh_types["negative particle"] = pybamm.MeshGenerator(
    pybamm.Exponential1DSubMesh, submesh_params={"side": "right"}
)
submesh_types["positive particle"] = pybamm.MeshGenerator(
    pybamm.Exponential1DSubMesh, submesh_params={"side": "right"}
)
sim =  pybamm.Simulation(model,parameter_values=param,solver=solver
sol = sim.solve(t_eval=t, initial_soc=SOC_0)

When I run the simulation for the 2C 0°C, 2C 10°C and WLTP 0°C conditions I am getting the following error

---------------------------------------------------------------------------
SolverError                               Traceback (most recent call last)
Cell In[106], line 1
----> 1 sol = sim.solve(t_eval=t, initial_soc=SOC_0)

File ~\AppData\Local\anaconda3\envs\battery\Lib\site-packages\pybamm\simulation.py:488, in Simulation.solve(self, t_eval, solver, save_at_cycles, calc_esoh, starting_solution, initial_soc, callbacks, showprogress, inputs, **kwargs)
    474             if dt_eval_max > dt_data_min + sys.float_info.epsilon:
    475                 warnings.warn(
    476                     f"""
    477                     The largest timestep in t_eval ({dt_eval_max}) is larger than
   (...)
    485                     stacklevel=2,
    486                 )
--> 488     self._solution = solver.solve(
    489         self.built_model, t_eval, inputs=inputs, **kwargs
    490     )
    492 elif self.operating_mode == "with experiment":
    493     callbacks.on_experiment_start(logs)

File ~\AppData\Local\anaconda3\envs\battery\Lib\site-packages\pybamm\solvers\base_solver.py:903, in BaseSolver.solve(self, model, t_eval, inputs, nproc, calculate_sensitivities)
    901 ninputs = len(model_inputs_list)
    902 if ninputs == 1:
--> 903     new_solution = self._integrate(
    904         model,
    905         t_eval[start_index:end_index],
    906         model_inputs_list[0],
    907     )
    908     new_solutions = [new_solution]
    909 else:

File ~\AppData\Local\anaconda3\envs\battery\Lib\site-packages\pybamm\solvers\idaklu_solver.py:679, in IDAKLUSolver._integrate(self, model, t_eval, inputs_dict)
    677     return newsol
    678 else:
--> 679     raise pybamm.SolverError("idaklu solver failed")

SolverError: idaklu solver failed

The simulation works fine under conditions at:

0.5C 0°C
0.5C 10°C
0.5C 25°C
1C  0°C
1C  10°C
1C  25°C
2C 25°C
WLTP 10°C
WLTP 25°C

Any insight on why the solver fails at 2C 0°C, 2C 10°Cand WLTP 0°C?

Any help would be much appreciated

2

Hi @ashkal, please try updating to the latest version of pybamm, which includes solver robustness improvements

3

Hi @Marc,

interesting feedback. I was looking through the discourse when I came across this post because I noticed something interesting regarding similar issues as stated in the post above: When running relative simple simulations (CCCV ~1-2C, maybe lithium plating, DFN) the solver has no issues at all when using e.g. Pybamm 25.10.1 with pybammsolvers 0.3.3 and casadi 3.6.7 but doesn’t even manage to complete the first step in the first cycle (convergence failure with a very small |h|) when using pybamm 26.3 (e.g. standard version when running “pip install pybamm”) which uses casadi 3.7.2 and pybammsolvers 0.6.0 - so in this case the latest or later versions of pybamm are not more robust, on the contrary. Do you have any idea why that is or what’s causing it regarding the change in the casadi and pybammsolvers versions?

(And what can be done about this? As updating is not an option in this case and generally switching between pybamm versions or solver versions until one finds a combination that works for a specific case can’t really be the way for the community)

best,