That has two levels of indirection in it 
If the digital computer was simply tasked with doing the same work, but in its domain, than for any analog computer we have ever made, digital will beat it. This has probably been true since the mid-70’s (earlier if you allow for machines like the CDC6600).
But simulating the actual physical system. Much much harder question. You now want to ask questions about fidelity of the simulation. Just how good is your simulation? You could probably knock up an analog computer in Spice quite quickly. But a simple version won’t be true to the real device. You should be adding all the second order effects in, indeed maybe modelling it down to the individual transistor, resistor and capacitor. With all their non-linearities. Now the task is much harder.
If we open the definitions wider we can get to the limits of digital quickly. Human brain? Perfectly good analog computer. Indeed often modelled with very similar components. And it fits the definition, since it represents many physical properties in its own domain (charge and chemical concentrations). No way to we get even close with digital.
Then there was the recent thread on a 300 atom quantum computer. That is an analog computer. Its whole purpose is to represent quantum physics problems from one area, in an analogous form in its own quantum domain, one where it is tractable to set up and then to interrogate it for the answers. It was cheerfully pointed out in the article that no possible digital computer could manage what it did if it was to try to simulate its action.