Well I see that most offered no real help. My suggestion, stop thinking of this as a "gravety" engine, it isn't.
Do some research on hydo-electric plants.
Your two biggest issues will be friction of the moving parts, use really god bearings. The other issue is "head".
Thats actually a semi-technical ter that describes your waterfall. You are going to need
feet of "head". Do your research to find out how to
compute it, and make it according to your design parameters.
You might also try not using a bicycle generator. They typically have large turning resistance, and you won't be able to fix that. Instead, use a
small DC electric motor, say in the 6 - 12 volt range. for a load, use an LED and resistor. Doing this will enable you to evaluate the performance of
a small model, and help in the decision to scale it up.
Stay away from gears and pullys as much as ossible. Balance will not have any effect on your system, so that is a design consideration you don't need.
Again, your biggest issues are "turning resistance" and "head".
I originally thought your idea was ... well not worth much. However,since it can't achieve perpetual motion; to much energy lost in bearings, and
"fuel loss" (water evaporation). But, that is no reason it wont work. Water is cheap (usually) and a bit of WD-40 will keep the bearings happy. If
large enough, I see no reason that it won't work and provide very cheap energy.
Good luck.
ETA. I just had a thought. for your turbine / paddle wheel, try a Tesla Turbine. They are typically more efficient and easier to build. You'll have to
do your homework on that, but, I think it will be well worth it.
edit on 21-5-2012 by AnthraAndromda because: (no reason given)