Determining the state space equation of a MISO system based on input and output data series

Question

I am building a model predictive controller for a reactor with two input streams and an output product. We control the input flow rates and measure the output concentration.

I have input and output values data and aim to build a model predictive controller by trying out different flow rates and measuring concentration on a physical setup.

The first step would be to identify the system to build a process model. I explored methods such as N4SID, but I was unable to find a library in Python that is equivalent to Systems Identification Toolbox in MATLAB.

I came across this library called SIPPY (SIPPY Github) and used their system identification functionality.

input = np.hstack((F_Na, F_Ag)).T # Input flow rates with shape (2, N)
output = sgf.T # Output shape (1, N)

sys_id = sippy.system_identification(y=output, u=input, id_method="N4SID", SS_fixed_order=4, centering="MeanVal")
xid, yid = sippy.functionsetSIM.SS_lsim_process_form(sys_id.A, sys_id.B, sys_id.C, sys_id.D, input, sys_id.x0)

The below two time series are the input data series

The below image depicts the output data series

After running SIPPY's system_identification function, the output generated from simulating the state equation is

I would appreciate help in the following areas.

1. Why is the generated output simulation beginning from zero instead of 400's as in the target
2. Suggest better approach or libraries for the same