Multiple input parameters

[Tangzl922]

Hi,

in the tutorial there is only example with one input parameter, could you please show me how to simulate with multiple parameters as input? Or maybe further, how to carry out the parametrisation for meanwhile multiple parameters?

Best regards,

[valentinsulzer]

What have you tried?

[Tangzl922]

Hi, I tried these codes

options = {"SEI": "reaction limited"}
model = pybamm.lithium_ion.DFN(options = options)

geometry = model.default_geometry


param = pybamm.ParameterValues("Chen2020")
param["Current function [A]"] = param["Current function [A]"]*0.05


param.update({"Initial concentration in negative electrode [mol.m-3]": "[input]",
              "Initial concentration in positive electrode [mol.m-3]": "[input]",
             "Maximum concentration in negative electrode [mol.m-3]": "[input]",
             "Maximum concentration in positive electrode [mol.m-3]": "[input]"})

param.process_geometry(geometry)
param.process_model(model)


var = pybamm.standard_spatial_vars
var_pts = {var.x_n: 30, var.x_s: 30, var.x_p: 30, var.r_n: 10, var.r_p: 10}
mesh = pybamm.Mesh(geometry, model.default_submesh_types, var_pts)


disc = pybamm.Discretisation(mesh, model.default_spatial_methods)
disc.process_model(model)

t_eval = np.linspace(0, 3600, 1000)
solver = pybamm.CasadiSolver(mode="safe", atol=1e-9, rtol=1e-6)


solution = solver.solve(model, t_eval/0.05)
plt.plot(solution["Discharge capacity [A.h]"](t_eval/0.05)*1000, solution["Terminal voltage [V]"](t_eval/0.05), label="Discharge rate = 0.05C")
plt.xlabel('Discharge capacity [mA.h]')
plt.ylabel('Terminal voltage [V]')



plt.legend()
plt.show()

and I got this error

My purpose is to find the values for these 4 parameters to fit my experimental data. Later on I tried

def sum_of_squares(parameters):
    print('solving for "initial concentration in negative electrode" =', parameters[0])
    siimulation = solver.solve(
    model, t_eval,
    inputs = {"Initial concentration in negative electrode [mol.m-3]": parameters[0],
              "Initial concentration in positive electrode [mol.m-3]": parameters[1],
             "Maximum concentration in negative electrode [mol.m-3]": parameters[2],
             "Maximum concentration in positive electrode [mol.m-3]": parameters[3]})["Terminal voltage [V]"](t_eval)
    return np.sum((simulation - V_exp) ** 2)

I also tried to give a timescale:

# load model
options = {"SEI": "reaction limited", "timescale": 3600/0.05}
model = pybamm.lithium_ion.DFN(options = options)
# create geometry
geometry = model.default_geometry

# load parameter values and process model and geometry
param = pybamm.ParameterValues("Chen2020")
param["Current function [A]"] = param["Current function [A]"]*0.05

# set "Current function" as an input parameter
param.update({"Initial concentration in negative electrode [mol.m-3]": "[input]",
              "Initial concentration in positive electrode [mol.m-3]": "[input]",
             "Maximum concentration in negative electrode [mol.m-3]": "[input]",
             "Maximum concentration in positive electrode [mol.m-3]": "[input]"})

param.process_geometry(geometry)
param.process_model(model)

# set mesh
var = pybamm.standard_spatial_vars
var_pts = {var.x_n: 30, var.x_s: 30, var.x_p: 30, var.r_n: 10, var.r_p: 10}
mesh = pybamm.Mesh(geometry, model.default_submesh_types, var_pts)

# discretise model
disc = pybamm.Discretisation(mesh, model.default_spatial_methods)
disc.process_model(model)

t_eval = np.linspace(0, 3600, len(C_exp))
solver = pybamm.CasadiSolver(mode="safe", atol=1e-9, rtol=1e-6)


solution = solver.solve(model, t_eval/0.05)
plt.plot(solution["Discharge capacity [A.h]"](t_eval/0.05)*1000, solution["Terminal voltage [V]"](t_eval/0.05), label="Discharge rate = 0.05C")
plt.xlabel('Discharge capacity [mA.h]')
plt.ylabel('Terminal voltage [V]')



plt.legend()
plt.show()

Then I got this

[Tangzl922]

Hi,

it seems that I have solved the problem. I gave the inputs into the solver and it works fine now with the simulation results. However, when I tried the data fitting there's problem like this

Does it mean that maybe my boundaries are too wide so that some attempts of parameter values can lead to error of the solver?

[rtimms]

Yeah you’ll need to handle cases where the optimiser attempts values that cause the solver to fail. E.g. try/except block.

[Tangzl922]

Hi,
thank you for your reply. I met another problem that the scipy.optimize.minimize solver only finds the local optimum which extremely depends on the starting value. Sometimes when I give the starting values for the four input parameters it will not even change after fitting. Do you have any idea about this?