flar said:
If the hill is stronger than the air resistance and you need to use regen to avoid speeding up then that is actually a good solution.
It is a fine solution most of the time, but not always the most efficient solution, and not always the desired solution. There are plenty of scenarios where the free kinetic energy gained from gravity can be absorbed by gravity or wind without the efficiency loss. It is also arguable that, as in the rest of you post, one should have planned for the decline and been going slower when it started, but then we are potentially in dangerous hypermiling territory again.
Note that the bulk of my post was actually a reaction to reports that people were claiming that the most efficient travel on a long flat highway involved the use of coasting to manage speed and I completely disagree with that assessment. But you are focusing on my added thoughts on hills and I agree that the cases there are more mirky. But, to expound...
For a steady state of a long downhill you are usually better reclaiming the energy into the battery before you increase wind resistance. Consider an exteme case where the car speeds up by 10MPH if you coast but then reaches a steady state where it won't get any faster and you are happy with maintaining that faster speed (safety/enjoyment-wise). In that scenario you have maximized your storage and are gaining nothing further. By introducing just a small amount of regen you will continually reclaim some energy that would have gone into increased friction (heat and turbulence) and be able to reuse it later. And you will keep reclaiming that energy, however imperfectly efficiently, for as long as the hill continues. If you simply continue coasting then you gain nothing. If you don't reach steady state, then you are likely going to reach a speed at which you no longer feel comfortable and then you will want to use the brakes. At that point you are then demonstrably and radically less efficient than using regen.
Looking at the less extreme cases, every MPH you gain reduces the efficiency of the speed gain because air resistance goes up quadratically. But, every second of regen has a consistent efficiency that lasts as long as you use the regen. I suppose there may be cases that depend on the size of the hill, its steepness, and how many MPH you gain where coasting would be a good solution too, but it would depend. And saying "I always want to use coasting because sometimes it might be the right answer" is a little short-sighted.
The presence of an immediate uphill right after the downhill also complicates things because not only do you spend the added speed immediately into a buffer against gravity, but it also reduces the amount of time you spend on the uphill.
I think the theoretical still almost never involves exactly zero energy. If you had a supercomputer analyzing the next 50 miles of roadway (and traffic?) ahead of you it could potentially model all variations of energy usage and it could come up with exactly the right amount of energy to put in or take out right at the moment for end-to-end efficiency and it would almost never be exactly 0.
But, humans can't calculate that fast and that fully, so coasting is an easy thing to comprehend that can get them more efficiency than guessing in many cases, but they aren't being optimal in the theoretical case, they are simply being better with their poor compute capacity than they could have if they just guessed at the amount of energy to use.
On a roller coaster road, humans will likely never be able to calculate any better than coasting on the downhills and spending the speed on the uphill. And they might not be far off from the efficiency of a supercomputer making carefully calculated use of regen.
On a long downhill, though, regen is probably a better answer than unchecked speed gain. If your local highways include more long slopes rarely paired with an immediate uphill then regen is probably a win. If you do have a lot of shorter back to back hills, then coasting might be an easier plan to achieve efficiency.
Also, my other point was that humans work better as a situational awareness calculating engine than as a numbers calculator. If you are looking at your numeric speed and kWh expenditure and the amount and direction of electron flow then you are using your brain for something that a computer would be better at. If instead you get your head out of the car's numbers and look ahead, even if what you see is a steep downhill leading into a steep uphill and how fast the highway is sliding beneath you, your brain will be able to do what it is naturally good at. You can train yourself as to how much speed in which situations is best so that you don't over-accumulate speed (whether the excess violates safety, or comfort, or what you can spend later on an uphill) and your foot will find 0-coasting when it needs to.
If you decide instead that "coasting" is the be-all and end-all solution as a control state, then you might get some added efficiency, but the ping-pong nature of intentionally entering that control state as a tactical end goal is limiting, and you then need a crutch to find that rigid coasting state (i.e. wanting to get exactly coasting by lifting your foot completely from the pedal).