I can't help off the top of my head with sources. Berardi wrote some articles here about insulin; you could see if there are any useful references in his articles. Once you find a few, those will cite others, and then you're good to go.
But I do have some advice about the experiment, as someone who has taught experimental psych many times at a major research university.
If you actually have to conduct the experiment, I would not do this one. First, a behavioral measure of "more energy" is gonna be tough, especially with children.
Second, children's behavior varies so much anyway that it will really be hard to tease apart energy effects of the meal from the natural variation of children's energy. Kids seem to have boundless energy for long periods, then zone out or fall deeply asleep in a matter of minutes. If there is any effect of meal, it's likely to be small, so you'd need a HUGE sample of children to discern the effect from the variability.
Third, from my own personal, anecdotal experience I would hypothesize that you won't see an effect, and that's kind of a pain. It just seems to me that young kids do very well eating lots of carbs, even simple carbs, and don't seem to need or benefit from protein as much as teenagers or adults. I suspect that most young children have great insulin sensivity and sugar handling.
Finally, and probably most importantly, it's hard to do experiments with children. They are hard to work with; you probably won't even be able to get an IRB protocol approved; if you do by some miracle get approval, you'll have so much more red tape -- extra consent forms, parental participation, extra oversight, blah blah blah. You want to stick to people 18 and over -- the magic number.
And that is what I would recommend. You could do a similar study, but with college students. College students are under lots of stress, probably starting to experience effects of poor diet that they were impervious to as children, and probably benefitting from more protein. Plus they are much easier to find in a college course setting. Be sure to administer meals and subsequent tests to each group separately; otherwise the obvious treatment difference could contaminate results.
There are still lots of confounding pitfalls to the design. If the sample is small, and the subjects are not matched, you'll really be mixing large individual differences with whatever effect the meal might have. For example, if you have 2 groups of 5, and the sugary-meal group contains 2 athletes who happen to have great insulin sensitivity, while the protein group contains a subject who has hypoglycemia but doesn't know it -- you'll find no effect of the meal, but it will be due to those individual differences confounded with meal condition.
I guess I'm getting carried away here with this post, but go ahead and post your study design if you want me to pick it apart; that's my specialty.
Hope this helps.