Transfer function estimation with frequency response. Python equivalent of MATLAB tfest (but the algorithm is not exactly the same)

Only works with SISO systems for now.

pip install tfest

To get a scipy transfer function:

import tfest
# u: input
# y: output
te = tfest.tfest(u, y)
# n_zeros, n_poles
te.estimate(3, 4, time=1)
te.get_transfer_function()

"time" is simulation length in seconds. To plot its bode diagram:

te.plot_bode()

Default method to calculate the frequency response Y/U is H1 estimator, if you want to use H2 or frequency/frequency just set the method to "h2" or "fft" and specify the time length of the simulation (in seconds, default is 1):

te.estimate(3, 4, method="h2", time=1)
te.estimate(3, 4, method="fft", time=1)

To use L2 regularization set the value of lambda l1 (default l1=0):

te.estimate(3, 4, time=1, l1=0.1)