I've always thought that fluids should have some sort of "energy" on the tile representing the flow (energy of 0 being stagnant water, brooks might be 4, major rivers 10, whatever).
Realistically they should, but that makes the fluid physics massively more complicated.
I thought I should give more of a reason why it makes the physics more complicated: water (I am pretty sure) teleports when it travels through a system in DF, so the energy would not be coherent and you could (possibly) make an even worse exploit with a cunning arrangement of tunnels, assuming that water is some sort of "object" that moves around and not a property of an empty tile.
While I agree there are more pressing bugs/quirks, it still warrants some (more) thought.
Hmmm, if a water wheel needed the water to fall to generate power, let's instead assume it needed it to fall 3 z levels, since the building is three tiles long anyway and water wheels are circular. That means you'd need 4 screw pumps to get the water high enough to fall over the wheel if you drew water from the same basin as the wheel sits on. That leaves a net power of 50 ( 90 - 10 x 4 ). If you halve a waterwheel's power output, then it at least breaks even which would be the case in a frictionless world. Bump up a waterwheel's power consumption to 15 and it is a loosing system and will never work.
A more complicated but perhaps better solution may be to use the fact that a waterwheel is already aware if it sits on top of moving water. If you change that to being aware of how many tiles of moving water are within the imaginary 3x3x1 volume which the wheel takes up, with a max count of three, then you might have something. You could say the output of the wheel is 15 x n, where n is the number of moving tiles of water. The wheel would consume 10 itself, leaving a max output of 35. Change pumps to use 15 power. Then for max power, you need 60 power to pump the water and you will always need 25 more power than what the wheel can possibly provide.
Of course, that nerfs waterwheels significantly, but it is sorta odd that you can currently power a pump from 80 tiles away. Assuming DF tiles are around 2 meters, that means a 9 foot water wheel can power a pump that can move eight cubic meters of water two meters high from one and two thirds football fields away. In terms of the wooden bearings alone that seems rather impossible.