Prep separations are generally run under massive overload conditions, so the assumptions involved in plate height calculations via the Van Deemter or Knox equations (most importantly, the assumption that the distribution coefficients are independent of concentration) don't apply. That means that there is no "simple" set of equations that you can plug in; you have to make an assumption about the shape of the distribution isotherm. Even then, it's typically easier to "brute force
model it using sort of Craig-countercurrent model in Excel.
Unless you're really limited in how much sample you have, the easiest approach is empirical: run a series of injections with increasing load (double or triple or more) each time and see how far you can push it before the separation becomes inadequate.
As I indicated above, most of the time the load is determined by the mass (or volume) of packing.