Sure, all the ones we know about today. It's a bit of an anthropic argument, though. We only know about the coal seams that have survived long enough for us to find them; and, almost universally, we've chosen to exploit the seams we've found. That leaves two questions unanswered:

• What would the statistical yearly probability of a coal-seam fire be for seams we find, check up on, but never mine? (We know of a few seams in protected areas, but I don't think we know of enough to have any statistical power. Also, the natural catching-fire rate might be, say, one per thousand years, which would be a bit like the pitch-drop experiment in its requirements, yet would still mean that coal seams catch fire pretty much "as soon as" they reach the surface, in geological time.)

• More interestingly: how many coal-seam fires have happened historically in the Earth's 4.2 billion years being around, that finished and burned out all the coal, and now there's just a cave there? Or ex-cave sedimentary layer, given that these caves tend to collapse? Can we tell when such a cave/layer was previously a coal seam? (I assume so, because soot, but can pressure-treated soot be differentiated from other kinds of naturally-occurring mineral striae?)