If I were doing it, ... Thevenin model ... Norton model .....
Actually, in a case like this where I was expecting
grossly nonlinear behavior, I would not bother with
models at all, but would instead just map out the
entire (V,I) characteristic and present it in graphical
or tabular form. This would, of course, require using
a multiplicity of different load resistors.
I would repeat the mapping-out process for a few
different settings of the front-panel voltage-limit
and current-limit controls.