BasicDFN and DFN model

[xyyrly]

Hi Team,

Does BasicDFN model still work now? I used BasicDFN and DFN to calculate the same event but got different results. For example, the "Negative electrode interfacial current density [A.m-2]" even has quite different trends:

[valentinsulzer]

That is strange, they are supposed to give identical answers, it is even tested here: https://github.com/pybamm-team/PyBaMM/blob/develop/tests/integration/test_models/test_full_battery_models/test_lithium_ion/test_compare_basic_models.py#L13

[xyyrly]

I don't know why it doesn't work... Here is a simple example, and their variable "Negative electrode potential [V]" cannot match with each either.
model0= pybamm.lithium_ion.BasicDFN()
model_1=pybamm.lithium_ion.DFN()
parameter_value = pybamm.ParameterValues("Ecker2015")
parameter_value["Current function [A]"]=1.5652
simulation = pybamm.Simulation(model0,parameter_values=parameter_value)
simulation1 = pybamm.Simulation(model1,parameter_values=parameter_value)
t_eval = np.linspace(0, 0.005*5208, 5209)
solution_0 = simulation.solve(t_eval)
solution_0 = simulation.solution
solution_1 = simulation1.solve(t_eval)
solution_1 = simulation1.solution
plt.plot(solution_0["Negative electrode potential [V]"].entries[:,-1],label="BasicDFN")
plt.plot(solution_1["Negative electrode potential [V]"].entries[:,-1],label="DFN")
plt.xlabel("x/m")
plt.ylabel("Negative electrode potential [V]")
plt.legend()

[rtimms]

There is an error in the boundary conditions in BasicDFN which only shows up when the diffusivity is nonlinear. We should take surf(D(c)) and not D(surf(c)).

The particle boundary conditions in BasicDFN should be replaced with

        # Boundary conditions must be provided for equations with spatial derivatives
        self.boundary_conditions[c_s_n] = {
            "left": (pybamm.Scalar(0), "Neumann"),
            "right": (
                -j_n / (param.F * pybamm.surf(param.n.prim.D(c_s_n, T))),
                "Neumann",
            ),
        }
        self.boundary_conditions[c_s_p] = {
            "left": (pybamm.Scalar(0), "Neumann"),
            "right": (
                -j_p / (param.F * pybamm.surf(param.p.prim.D(c_s_p, T))),
                "Neumann",
            ),
        }

[rtimms]

Fixed here: #3207

[xyyrly]

Thank you very much. Now it works well.

[bobonice]

@rtimms Excellent job! After fixing this issue, another strange problem is that differences in the results of two models still persist. While this disparity may not be readily discernible in the figures, it becomes more evident when examining the numerical values. Additionally, the differences between the two gradually almost increase over time and distance, which could be a consequence of errors accumulating. And the other variables also exhibit the same issue, such as "Negative particle surface concentration [mol.m-3]". I would appreciate it if you could shed some light on the factors contributing to the differences in results between the BasicDFN and DFN models.

[brosaplanella]

If you adjust the solver tolerances and/or refine the mesh, does the error decrease? It could be a matter of numerical accuracy. Here is an example on how to change solver settings and change the mesh.

[valentinsulzer]

Alternatively can you check how the variance is when solving the same model twice