buzzp
I have to take issue with your comment. CT accuracy generally gets worse as the current falls below the rated current. The CT accuracy get better above the rated current, up to a point where core saturation starts to take effect.
However, for most protection application, the concern is not the accuracy at or around the rated current, but the performance of the CT during transient conditions.
The SEL paper tgott suggests is a good starting point. However, there are a few assumptions they make that I take some issue with, in particular when they discuss the internal burden of the CT and give rules of thumb for ohms/turn. While this generally holds true for bushing/slip-over CTs (i.e. 1 primary turn), the user has no way of knowing for sure the number of secondary turns on CTs with the possibility of multiple-turn primaries. This is esspecially true for insulated CTs where the ratio is 800:5A or below. Bottom line is that if you want to pick CT ratios based on this method, you either need to specify the maximum ohms/turn, or ask manufacturer for a an excitation curve and secondary resistance figure during the proposal stage.