Using Pybamm to identify ECM parameters

[SupremeZ01]

Description

Can I treat Pybamm's DFN model as a real battery and use GITT or HPPC testing to identify the SOC-OCV curve and electrical parameters of the first-order equivalent circuit model?
I tried to do this work, but found that the error between the identified ECM model and Pybamm was a bit large. Especially when I used the ECM model and Pybamm for online closed-loop control, I found that the error between the SOC of my ECM model and the SOC calculated by Pybamm was a bit large. I don't know how to solve this problem.

Motivation

No response

Possible Implementation

No response

Additional context

No response

[valentinsulzer]

How did you calculate SOC with PyBaMM?

[SupremeZ01]

Thanks for your attention!

I used the eSOH model to calculate the SOC. The version of Pybamm is 22.9.

import pybamm

model = pybamm.lithium_ion.DFN()
parameter_values = pybamm.ParameterValues(chemistry=pybamm.parameter_sets.Chen2020)
experiment = pybamm.Experiment([
    ("Rest for 1 hour (10 s period)",
    "Charge at 0.05C until 4.2V (10 s period)",
    "Rest for 1 hour (10 s period)",
    "Discharge at 0.05C until 2.5V (10 s period)",)
])
sim = pybamm.Simulation(model, experiment=experiment, parameter_values=parameter_values)
sol = sim.solve(initial_soc = 0)

param = pybamm.LithiumIonParameters()
V_min = parameter_values.evaluate(param.voltage_low_cut_dimensional)
V_max = parameter_values.evaluate(param.voltage_high_cut_dimensional)
C_n = parameter_values.evaluate(param.n.cap_init)
C_p = parameter_values.evaluate(param.p.cap_init)
n_Li = parameter_values.evaluate(param.n_Li_particles_init)

esoh_solver = pybamm.lithium_ion.ElectrodeSOHSolver(parameter_values, param)

inputs = {"V_min": V_min, "V_max": V_max, "C_n": C_n, "C_p": C_p, "n_Li": n_Li}

esoh_sol = esoh_solver.solve(inputs)
x_0 = esoh_sol["x_0"].data[0]
x_100 = esoh_sol['x_100'].data[0]
y_0 = esoh_sol["y_0"].data[0]
y_100 = esoh_sol['y_100'].data[0]
c_n_max = parameter_values.evaluate(param.n.prim.c_max)
c_p_max = parameter_values.evaluate(param.p.prim.c_max)

x = sol['Average negative particle concentration'].entries
SOC = (x-x_0)/(x_100-x_0)

Supplement:

I used the average value of the terminal voltage of charging and discharging to obtain the OCV-SOC curve:

model = pybamm.lithium_ion.DFN()  
parameter_values = pybamm.ParameterValues(chemistry=pybamm.parameter_sets.Chen2020)

experiments = pybamm.Experiment(
    [("Rest for 1 hour (10 s period)",
    "Charge at 0.05C until 4.2V (10 s period)",
    "Rest for 1 hour (10 s period)",
    "Discharge at 0.05C until 2.5V (10 s period)",
    )])

initial_soc = 0
sim = pybamm.Simulation(model, experiment=experiments, parameter_values=parameter_values)
sol = sim.solve(initial_soc=initial_soc)

Also, I used HPPC testing to obtain the R0, R1, C1 parameter values ​​of the ECM model:

experiment = pybamm.Experiment(["Rest for 1 hour"] + [
    ("Rest for 10 s (1 s period)",
    f"Discharge at 1C for 10 s (0.1 s period)",
    "Rest for 40 s (0.2 s period)",
    f"Charge at 0.75C for 10 s (0.1 s period)",
    f"Rest for 40 s (0.2 s period)"),
],
# period = "0.2 s", 
termination = "2.5 V"
)

initial_socs = [1, 0.95, 0.9, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2, 0.1]
model = pybamm.lithium_ion.DFN()       
parameter_values = pybamm.ParameterValues(chemistry=pybamm.parameter_sets.Chen2020)
parameter_values["Ambient temperature [K]"] = 298.15

sols = []
sols_plot = []
for initial_soc in initial_socs:
    sim = pybamm.Simulation(model, experiment=experiment, parameter_values=parameter_values)
    sol = sim.solve(initial_soc=initial_soc)
    sol = sol.cycles[1]
    sols_plot.append(sol)
    sols.append({sol:str(initial_soc)})

for i,sol in enumerate(sols_plot):
        t = sol["Time [s]"].data
        t = t - t[0]
        V = sol["Terminal voltage [V]"].data
        plt.plot(t, V)

The output of HPPC testing:

So, I guess that the R0, R1, and C1 parameter values I identified with HPPC are wrong, so my ECM model cannot track Pybamm's DFN model. Maybe you can give me some other suggestion for model identification, thanks for your help.

[valentinsulzer]

Hopefully someone more familiar with ECMs can help. I'll convert this to a discussion since I don't think there are any associated bugs or feature requests

[LuppiL]

hi Supreme, I'm working on a project with similar purposes. Have you found a solution to your doubts? Is it feasible to use DFN model for ECM parametrization?

[SupremeZ01]

I gave up offline identification because the authenticity of the parameters could not be verified. I directly use RLS to identify ECM parameters online to approximate the terminal voltage.

[flag2freefish]

what is the RLS?

[SupremeZ01]

a parameter identification algorithm, recursive least squares

[Akila1993]

@SupremeZ01 Can you please let me know, for ECM parameter estimation how did you identify the initial condition of the parameters ECM model circuit?

If I am correct, we may define the initial condition if we need to identify the parameter from the RLS method. Please advise me how to approach this.