Parametrize an RLC equivalent circuit in Simscape by measurement data via Curve Fitting or similar

Hi,

my plan is to model the behaviour of a Black Box/an unknown electrical circuit at certain points in time, e.g. switching the box on/off. Only a qualitative model needs to be created to describe the behaviour in a certain, unsteady condition. (two separate models for describing the behaviour during switching on/switching off are possible)

My initial idea was describing the box by a basic RLC circuit as shown in the appendix, with L and C to be parametrized by collected measurement data (R is already known from measurements taken at steady state conditions). So basically, the Black Box is described as a consumer with filter elements. I'm not able to collect measurement data inside of the box, so data can only be collected at the boxes' inputs. The consumer can be switched on/off externally to collect the required measurement data.

Is there an option in MATLAB/Simulink to parametrize L and C (e.g. via curve fitting) in a way that the behaviour of the real Black Box during switching can be qualitatively replicated in a Simscape simulation?

Feel free to comment, if you have alternative modelling options/circuits in mind.

Thanks in advance.

Cheers,

yinki23

Answer 1

[Electrification Enthusiasts]

Hello,

there are 2 options that come to mind immediately. 1 option would to be to use system identification toolbox to fit the transfer function behavior of the circuit, and then analytically solve for L and C components to match that transfer function. This approach requires more analytical work but probably would give you the best possible fit across operating conditions.

The other option (which is what I think you are looking for) is to numerically solve for C and L using parameter estimation (simulink design optimization toolbox) and your lab data. Matlab can vary the C and L values till the data matches as best it can. There are limitations to this approach due to local minima and other considerations when doing numerical optimization ( system id is also a numerical optimization approach so it could have similar limitations)

https://www.mathworks.com/products/sl-design-optimization.html#parameter

https://www.mathworks.com/help/sldo/ug/inverted-pendulum-parameter-estimation.html

Regards,

Joel

[yinki23]

Hi Joel,

thank you very much so far. I'm going to try out both options as soon as I got my measurement results (probably next week) and I will feed back to you then.

Greetings,

yinki23