While it might be true, that computers are extremely powerful these days, you can still face applications where you have tight power constraints. There it might not be possible to use a powerful enough microcontroller or processor.
In those cases an analogue circuit might still be the way to go.
As this sounds quite vague and probably not only the asker wonders what this has to do with modelling a physical system, I'll give a short explanation what I have in mind. I can't go into too much detail but here goes:
The control system may take only 5mW of power. It is required to control a system with an update rate of 1kHz. There has to be some sort of mechanism to detect a faulty condition inside the controlled system in a very timely manner, at best in real-time (also 1kHz).
Now the control mechanism is implemented in a microcontroller which is able to handle that task quite well inside these energy boundaries. The problem was to detect the faulty condition.
For this task the analogue circuit came in. The system was modelled as an analogue circuit, and the same input was applied to the circuit as to the system. If the output differed by a to big amount, there was something wrong in the system (or the model during development stages).
Now calculating the whole model at a refresh rate of 1kHz was not possible for the microcontroller in that power budget.
Today with the advent of Cortex M4F, you might get close, they are pretty impressive, but I doubt that it is possible to squeeze it in just yet.
This might not be the modelling the asker had in mind, but it's still an application of an analogue circuit to model a physical system.