def model_impedance(circuit, guess, freq, impedance, name='circuit'):
model = CustomCircuit(name='Custom {}'.format(name),
circuit=circuit,
initial_guess=guess)
model.fit(freq, impedance, method='lm', bounds=(-np.inf, np.inf))
# print(repr(model.predict(np.geomspace(0.001,2e6,100),True)))
return model
def test_model_io():
# get example data
data = np.genfromtxt(os.path.join(".", "data", "exampleData.csv"),
delimiter=',')
frequencies = data[:, 0]
Z = data[:, 1] + 1j * data[:, 2]
randles = CustomCircuit(
initial_guess=[None, .005, .1, .005, .1, 0.9, .001, None],
circuit='R0-p(R1,C1)-p(R1,CPE1)-Wo1',
constants={
'R0': 0.01,
'Wo1_1': 200
})
randles.save('./test_io.json')
randles2 = CustomCircuit()
randles2.load('./test_io.json')
assert randles == randles2
randles.fit(frequencies, Z)
randles.save('./test_io.json')
randles2 = CustomCircuit()
randles2.load('./test_io.json')
assert str(randles) == str(randles2)
assert randles == randles2
fitted_template = CustomCircuit()
fitted_template.load('test_io.json', fitted_as_initial=True)
def five_transmission_calc_impedance(self):
self.data_Z, self.data_freq = file_select()
circuit = 'R0-p(CPE0,R1-p(CPE1,R2))'
initial_guess = [20, .1, .1, 50, 50, .1, 50]
circuit = CustomCircuit(circuit, initial_guess=initial_guess)
circuit.fit(self.data_freq, self.data_Z)
Z_fit = circuit.predict(self.data_freq)
fig, ax = plt.subplots()
plot_nyquist(ax, self.data_Z, fmt="o")
plot_nyquist(ax, Z_fit, fmt="-")
plt.legend(["Data", "Fit"])
self.master.withdraw()
self.eqc_calc_menu = tkinter.Toplevel()
img_canvas = tkinter.Canvas(self.eqc_calc_menu,
width=self.five_transmission_pic.width(),
height=self.five_transmission_pic.height())
img_canvas.grid(row=0, column=0)
img_canvas.create_image(0,
0,
image=self.five_transmission_pic,
anchor="nw")
img_canvas.update()
self.graph_canvas = FigureCanvasTkAgg(fig, self.eqc_calc_menu)
self.graph_canvas.draw()
self.graph_canvas.get_tk_widget().grid(row=0, rowspan=3, column=1)
return_btn = tkinter.Button(
self.eqc_calc_menu,
text="Return to Main Menu",
font=("Helvetica", 14),
bg="OliveDrab1",
command=lambda:
[self.eqc_calc_menu.destroy(),
self.master.deiconify()])
return_btn.grid(row=3, column=0)
exit_btn = tkinter.Button(
self.eqc_calc_menu,
text="Exit Program",
font=("Helvetica", 14),
bg="tomato",
command=lambda:
[self.eqc_calc_menu.destroy(),
self.master.destroy()])
exit_btn.grid(row=4, column=0)
stats = tkinter.Label(self.eqc_calc_menu,
text=circuit,
font=("Helvetica", 12),
justify="left")
stats.grid(row=2, column=0)
def test_model_io():
# get example data
data = np.genfromtxt(os.path.join(".", "data", "exampleData.csv"),
delimiter=',')
frequencies = data[:, 0]
Z = data[:, 1] + 1j * data[:, 2]
randles = CustomCircuit(
initial_guess=[None, .005, .1, .005, .1, .001, 200],
circuit='R0-p(R1,C1)-p(R1,C1)-Wo1',
constants={'R0': 0.01})
randles.fit(frequencies, Z)
randles.save('./test_io.json')
randles2 = CustomCircuit()
randles2.load('./test_io.json')
print(randles2)
assert randles == randles2
def fit(data_dir, filename, plot: str = ""):
array = open_file_numpy(data_dir, filename)
params = get_params(data_dir, filename)
print(array.shape)
data = get_data_numpy(array)
(freq_log, imped_log, phase, imag_imped, real_imped) = data
freq = np.power(10, freq_log)
# np.savetxt("zzz.csv", np.concatenate((freq, real_imped, -imag_imped), axis=1), delimiter=",")
# f, Z = preprocessing.readCSV("zzz.csv")
f = freq[:, 0]
Z = np.array(real_imped + 1j * -imag_imped, dtype=np.complex)[:, 0]
circuit_str_5a = 'R0-p(R1,CPE1)-p(R2,CPE2)-p(R3,CPE3)'
R0_guesses = [6850]
R1_guesses = [1E7, 1E6]
C1_guesses = [5E-8, 1E-8, 5E-9]
CPP_guesses = [0.8, 0.9, 1.0]
R2_guesses = [1E7, 1E5, 1E6, 1E4]
C2_guesses = [1E-6, 1E-7, 1E-8]
CPP2_guesses = [0.9, 0.8, 1.0]
R3_guesses = [1E7, 1E5, 1E6]
C3_guesses = [1E-5, 1E-6, 1E-7, 1E-8]
CPP3_guesses = [0.5]
initial_guesses_5a = []
for R0 in R0_guesses:
for R1 in R1_guesses:
for C1 in C1_guesses:
for CPP in CPP_guesses:
for R2 in R2_guesses:
for C2 in C2_guesses:
for CPP2 in CPP2_guesses:
for R3 in R3_guesses:
for CPP3 in CPP3_guesses:
for C3 in C3_guesses:
initial_guesses_5a.append([
R0, R1, C1, CPP, R2, C2, CPP2,
R3, C3, CPP3
])
num_init_guesses = len(initial_guesses_5a)
print(f"{num_init_guesses} guess combinations for {circuit_str_5a}")
assert all(initial_guesses_5a), "Cannot use 0 for any initial guesses"
chi_sentinal = 0.1
num_remaining = num_init_guesses
fitted_dict = {}
for initial_guess in initial_guesses_5a:
# print(params.voltage, initial_guess)
num_remaining -= 1
circuit = CustomCircuit(circuit=circuit_str_5a,
initial_guess=initial_guess)
circuit.fit(
f,
Z,
global_opt=False,
bounds=(
# R0 R1 C1 CPP1 R2 C2 CPP2 R3 C3 CPP3"
[6800, 0, 1E-9, 0.78, 0, 0, 0.78, 0, 0, 0.48],
[7000, np.inf, 1E-6, 1, np.inf, 1E-5, 1, np.inf, 1E-5, 0.52]))
fitted_params = circuit.parameters_
R0, R1, C1, CPP, R2, C2, CPP2, R3, C3, CPP3 = circuit.parameters_
Z_fit = circuit.predict(f)
dof = len(Z) - len(
initial_guess
) # Degrees of freedom - i.e. number of data points - number of variables
Re_Z = np.real(Z) # Extract real part of Z
Re_Z_fit = np.real(Z_fit) # Extract real part of Z_fit
variance_re = sum(
(Re_Z - Re_Z.mean())**2) / (len(Re_Z) - 1) # Variance
Chi_square_re = sum(((Re_Z - Re_Z_fit)**2) / variance_re)
Im_Z = np.imag(Z) # Extract real part of Z
Im_Z_fit = np.imag(Z_fit) # Extract real part of Z_fit
variance_im = sum(
(Im_Z - Im_Z.mean())**2) / (len(Im_Z) - 1) # Variance
Chi_square_im = sum(((Im_Z - Im_Z_fit)**2) / variance_im)
Zsq = np.sqrt(abs(Re_Z**2 + Im_Z**2))
Zsq_fit = np.sqrt(abs(Re_Z_fit**2 + Im_Z_fit**2))
variance = sum((Zsq - Zsq.mean())**2) / (len(Z) - 1) # Variance
Chi_square = sum(((Zsq - Zsq_fit)**2) / variance)
chi_square_diag = sqrt(Chi_square_re**2 + Chi_square_im**2)
red_Chi_sq = Chi_square / dof # Reduced Chi squared
std_err = np.sqrt(red_Chi_sq) * 100
fitted_dict[tuple(initial_guess)] = (fitted_params, red_Chi_sq, f, Z,
Z_fit, circuit_str_5a,
params.voltage)
divisor = 60 if num_init_guesses > 60 else num_init_guesses
if num_remaining in range(1, num_init_guesses,
int(num_init_guesses / divisor)):
print(f"{initial_guess} - {num_remaining} guesses left")
# if chi_square_diag < chi_sentinal or chi_square_diag < 0.005 or Chi_square_re < chi_re_sentinal or Chi_square_re < 0.0038:
if red_Chi_sq < chi_sentinal:
chi_sentinal = red_Chi_sq * 1.02
print(
f"Fit {params.voltage}, red_chi2={red_Chi_sq:.8f}, chi2_re={Chi_square_re:.3f}, chi2_im={Chi_square_im:.3f}, std_err={std_err:.4f}, circuit={circuit_str_5a}\t {num_remaining} guesses left"
)
fitted_tuple = sorted(fitted_dict.items(), key=lambda item: item[1][1])
sorted_fitted_dict = OrderedDict(fitted_tuple)
# pprint.pprint(sorted_fitted_dict)
return sorted_fitted_dict
def fit(data_dir, filename, plot: str = ""):
array = open_file_numpy(data_dir, filename)
params = get_params(data_dir, filename)
print(array.shape)
data = get_data_numpy(array)
(freq_log, imped_log, phase, imag_imped, real_imped) = data
freq = np.power(10, freq_log)
# np.savetxt("zzz.csv", np.concatenate((freq, real_imped, -imag_imped), axis=1), delimiter=",")
# f, Z = preprocessing.readCSV("zzz.csv")
f = freq[:, 0]
Z = np.array(real_imped + 1j * -imag_imped, dtype=np.complex)[:, 0]
circuit_str_5a = 'R0-p(R1-p(R2,CPE2),CPE1)'
R0_guesses = [6850]
R1_guesses = [1E7, 1E6, 1E8, 5E8]
C1_guesses = [5E-8, 1E-8, 5E-9, 5E-7, 1E-6] # , 5E-9, 5E-8, 1E-8]
CPP_guesses = [0.8, 0.9, 1.0]
R2_guesses = [1E5, 1E6, 1E7, 1E8]
C2_guesses = [5E-6, 1E-6, 1E-7, 1E-8]
CPP2_guesses = [0.8, 0.9, 1.0]
initial_guesses_5a = []
for R0 in R0_guesses:
for R1 in R1_guesses:
for C1 in C1_guesses:
for CPP in CPP_guesses:
for R2 in R2_guesses:
for C2 in C2_guesses:
for CPP2 in CPP2_guesses:
initial_guesses_5a.append([R0, R1, R2, C2, CPP2, C1, CPP])
num_init_guesses = len(initial_guesses_5a)
print(f"{num_init_guesses} guess combinations")
assert all(initial_guesses_5a), "Cannot use 0 for any initial guesses"
chi_sentinal = 1.0
chi_re_sentinal = 1.0
num_remaining = num_init_guesses
fitted_dict = {}
for initial_guess in initial_guesses_5a:
# print(params.voltage, initial_guess)
num_remaining -= 1
circuit = CustomCircuit(circuit=circuit_str_5a, initial_guess=initial_guess)
circuit.fit(f, Z, global_opt=False, bounds=(
# R0 R1 R2 C2 CPP2 C1 CPP1
[6800, 0, 0, 1E-9, 0.78, 1E-9, 0.78],
# [7000, 1E8, 1E-5, 1, 1E8, 1E-5, 1]
[7000, np.inf, np.inf, 1E-5, 1, 1E-5, 1]
))
fitted_params = circuit.parameters_
R0, R1, R2, C2, CPP2, C1, CPP = circuit.parameters_
Z_fit = circuit.predict(f)
# dof = len(Z)-len(initial_guess) # Degrees of freedom - i.e. number of data points - number of variables
Re_Z = np.real(Z) # Extract real part of Z
Re_Z_fit = np.real(Z_fit) # Extract real part of Z_fit
variance_re = sum((Re_Z-Re_Z.mean())**2) / (len(Re_Z)-1) # Variance
Chi_square_re = sum(((Re_Z-Re_Z_fit)**2)/variance_re)
Im_Z = np.imag(Z) # Extract real part of Z
Im_Z_fit = np.imag(Z_fit) # Extract real part of Z_fit
variance_im = sum((Im_Z-Im_Z.mean())**2) / (len(Im_Z)-1) # Variance
Chi_square_im = sum(((Im_Z-Im_Z_fit)**2)/variance_im)
variance = sum((Z-Z.mean())**2) / (len(Z)-1) # Variance
Chi_square = sum(((Z-Z_fit)**2)/variance)
chi_square_diag = sqrt(Chi_square_re**2 + Chi_square_im**2)
red_Chi_sq = Chi_square_re # Reduced Chi squared
fitted_dict[tuple(initial_guess)] = (fitted_params, red_Chi_sq, f, Z, Z_fit, circuit_str_5a, params.voltage)
if num_remaining in range(1, num_init_guesses, int(num_init_guesses / 60)):
print(f"{initial_guess} - {num_remaining} guesses left")
if chi_square_diag < chi_sentinal or chi_square_diag < 0.01 or Chi_square_re < chi_re_sentinal or Chi_square_re < 0.005:
if chi_square_diag < chi_sentinal:
chi_sentinal = chi_square_diag * 1.1
if Chi_square_re < chi_re_sentinal:
chi_re_sentinal = Chi_square_re * 1.1
print(f"Fit {params.voltage}, chi2_re={Chi_square_re:.3f}, chi2_im={Chi_square_im:.3f}, chi2_diag={chi_square_diag:.3f}, circuit={circuit_str_5a}\t {num_remaining} guesses left")
if plot in ("all", "both"):
import matplotlib.pyplot as plt
fig, ax = plt.subplots()
plot_nyquist(ax, Z, fmt='o')
plot_nyquist(ax, Z_fit, fmt='-')
plt.legend(['Data', 'Fit'])
plt.title(f"{params.voltage}_{red_Chi_sq:.4f}")
plt.gcf().text(
0.8, 0.7,
f" R0={R0:.0f}\n\n R1={R1:.0f}\n C1= {C1}\n CPE1={CPP}\n\n R2={R2:.0f}\n C2={C2}\n CPE2={CPP2}\n\n {circuit_str_5a}",
fontsize=12)
fig.set_size_inches(8.5, 5.5, forward=True)
chi_str = f"{red_Chi_sq: .6f}".replace('.', '_')
plt.savefig(Rf'{data_dir}\all\{params.voltage}_chi_{chi_str}_{circuit_str_5a}.png',
bbox_inches='tight', dpi=500, transparent=True)
plt.close()
# plt.show()
fitted_tuple = sorted(fitted_dict.items(), key=lambda item: item[1][1])
sorted_fitted_dict = OrderedDict(fitted_tuple)
# pprint.pprint(sorted_fitted_dict)
return sorted_fitted_dict
import matplotlib.pyplot as plt
from impedance import preprocessing
from impedance.models.circuits import CustomCircuit
from impedance.visualization import plot_nyquist
circuit = 'R0-p(R1,C1)-p(R2-Wo1,C2)'
initial_guess = [.1, .01, .1, .05, .001, 20, 1]
circuit = CustomCircuit(circuit, initial_guess=initial_guess)
if __name__ == '__main__':
# Read and fit EIS data
frequencies, Z = preprocessing.readZPlot(
'C:\\Users\\luowe\\Desktop\\IRP_code\\Battery_Data\\50th\\10.z')
frequencies, Z = preprocessing.ignoreBelowX(frequencies, Z)
circuit.fit(frequencies, Z)
para = circuit.parameters_
# Generate training data
f = 'training data.txt'
with open(f, "a") as file:
file.write('50' + ',' + str(para[0]) + ',' + str(para[1]) + ',' +
str(para[3]) + '\n')
file.close()
# Plot data curve and fit curve
Z_fit = circuit.predict(frequencies)
fig, ax = plt.subplots()
plot_nyquist(ax, Z, fmt='o')
plot_nyquist(ax, Z_fit, fmt='-')
plt.legend(['Data', 'Fit'])
plt.show()
def test_CustomCircuit():
initial_guess = [.01, .005, .1, .005, .1, .001, 200]
custom_string = 'R0-p(R1,C1)-p(R2,C2)-Wo1'
custom_circuit = CustomCircuit(initial_guess=initial_guess,
circuit=custom_string)
# check get_param_names()
full_names, all_units = custom_circuit.get_param_names()
assert full_names == ['R0', 'R1', 'C1', 'R2', 'C2', 'Wo1_0', 'Wo1_1']
assert all_units == ['Ohm', 'Ohm', 'F', 'Ohm', 'F', 'Ohm', 'sec']
# check _is_fit()
assert not custom_circuit._is_fit()
# check predictions from initial_guesses
high_f = np.array([1e9])
assert np.isclose(np.real(custom_circuit.predict(high_f)[0]),
initial_guess[0])
# __str()__
initial_guess = [.01, .005, .1]
custom_string = 'R0-p(R1,C1)'
custom_circuit = CustomCircuit(initial_guess=initial_guess,
circuit=custom_string)
assert str(custom_circuit) == \
'\nCircuit string: R0-p(R1,C1)\n' + \
'Fit: False\n' + \
'\nInitial guesses:\n' + \
' R0 = 1.00e-02 [Ohm]\n' + \
' R1 = 5.00e-03 [Ohm]\n' + \
' C1 = 1.00e-01 [F]\n'
custom_circuit.fit(f, Z)
assert custom_circuit._is_fit()
custom_circuit.plot(f_data=f, Z_data=Z)
# constants and _ in circuit and no name
circuit = 'R_0-p(R_1,C_1)-Wo_1'
constants = {'R_0': 0.02, 'Wo_1_1': 200}
initial_guess = [.005, .1, .001]
custom_circuit = CustomCircuit(initial_guess=initial_guess,
constants=constants,
circuit=circuit,
name='Test')
assert str(custom_circuit) == \
'\nName: Test\n' + \
'Circuit string: R_0-p(R_1,C_1)-Wo_1\n' + \
'Fit: False\n' + \
'\nConstants:\n' + \
' R_0 = 2.00e-02 [Ohm]\n' + \
' Wo_1_1 = 2.00e+02 [sec]\n' + \
'\nInitial guesses:\n' + \
' R_1 = 5.00e-03 [Ohm]\n' + \
' C_1 = 1.00e-01 [F]\n' + \
' Wo_1_0 = 1.00e-03 [Ohm]\n'
# incorrect number of initial guesses
with pytest.raises(ValueError):
initial_guess = [.01, .005, .1, .005, .1, .001, 200]
custom_string = 'R0-p(R1,CPE1)-p(R1,C1)-Wo1'
custom_circuit = CustomCircuit(initial_guess=initial_guess,
circuit=custom_string)
# no initial guesses supplied before fitting
with pytest.raises(ValueError):
custom_circuit = CustomCircuit()
custom_circuit.fit(f, Z)
# incorrect circuit element in circuit
with pytest.raises(ValueError):
custom_circuit = CustomCircuit('R0-NotAnElement', initial_guess=[1, 2])