The InChI library (e.g. via the CDK) can be used to list a subset of tautomers for a given chemical graph, but in order to decide which one to use in QSAR modeling, there must be some ranking. One based on energy sounds like a good first approach.
Roger Sayle took me to school on this very topic when he visited our group recently. Basically the take home message is that any tautomer “force field” should take conjugation into consideration. He gives a number of such examples in his recent perspective. I’d certainly start with something simple as Table 2 of this paper, though modify the first rule to take into account conjugation.