At first, I agreed with the general consensus of separate tanks. But I thought about it, and if you do separate tanks, then a lot of the variables will be different. I.e. ferts amount, position of lighting, nutrients. However, at the same time, if you just put dividers, as fishscale mentions, some substrates like aquasoil has nutrients in it that could move onto other substrates or water column, skewing your results. So I guess to truly single out substrate as the only dependent variable, you'll have to be very exact with separate tanks.
For ferts, it seems kind of dangerous, because being once a very mischevious 5th grader myself, I could've thought up of at least 5 stupid things to do with it.
1. Dare my friends to drink the liquid dose or lick the powdered form, or do it myself.
2. Put it into the teacher's morning coffee.
3. Dump it on the school lawn during recess.
4. Dump all of it at once into the tank.
5. Give it to some kid who doesn't know what it is, telling them it is magic water.
So unless there's a safe place to store it, I wouldn't recommend it. Even then, what if some kid who was as crazy as me tried to drink the tank water? Also, ferts can have different effects on substrate. I'm thinking of CEC values. For example, sand has a low CEC value compared to peat moss. So that could mess up your results as well.
But it's only a 5th grade experiment, so it probably doesn't matter whether it's separate tanks or dividers. Every other variable is really simple to keep constant anyway.