Hacker News new | past | comments | ask | show | jobs | submit login

Former flywheel energy storage startup engineer here.

The economics of flywheels for this kind of application versus just using another battery tend to rest on the purported "unlimited cycle life" of the flywheel system compared to, say, Li-ion batteries that have a very well documented finite cycle capacity that degrades even further when doing sub-optimal cycling. To a lesser extent you can also bank on lower parasitic loads during standby as the environmental requirements for a flywheel aren't as stringent as batteries that need to be either heated or cooled almost all the time in many climates.

The problem is that, by and large, "unlimited cycles" is not true. You still have huge, very high speed bearings. Motors that require routine electrical testing and can fail. And now all this stuff is sitting below ground under a massive concrete lid for containment so it's not as easy to do maintenance on compared with a similarly-sized battery system. You also need uninterruptible power supply to maintain safety and control systems when grid power is unavailable since you've still gotten a huge spinning mass that you can't slow down without somewhere to send the energy (it's possible to use braking resistors, but it's another cost).

Batteries also benefit from massive economics of scale (both on the actual cells and the power electronics) that are getting better with time and driving costs down, while flywheels have been "1 year from commercialization" for the last 25 years.

I remain skeptical of the commercial benefits vs. increasingly commoditized and readily available battery systems.

> Former flywheel energy storage startup engineer here

I love this place sometimes.

Power utility engineer here. I keep imagining a giant flywheel located at every substation- could be very useful for load balancing, voltage regulation, even frequency control. I guess what I mean is, a large spinning mass could potentially have other benefits than just storage, right? Any research into substation application at that startup?

For an energy storage system, you probably don't want a synchronous machine. You want to be able to deliver (or absorb) rated power over a wide range of rotor speeds, since the difference between your lowest and highest speed is what effectively determines your energy storage capacity. To do this with a synchronous generator you'd need a big, expensive high-speed gearbox similar to what you see in synchronous generator wind turbines. So we used an induction motor/generator + back-to-back converter to tie to the grid, which is a common design very similar to what's used in most modern wind turbines. Therefore, from the perspective of the grid, no rotational inertia is coupled to the power system and the machine doesn't exhibit the inertial response of a synchronous generator that I'm sure you're familiar with.

However, if you have a fast enough control and communications system, you can provide the desired response for both frequency and voltage control and that's a big part of what we were trying to do. You can ramp from minimum to maximum active or reactive power very, very fast (like 5 AC cycles), which is much faster than any traditional generator. Due to the fast ramping capability, we also researched the possibility of using flywheels for large unit trip contingencies in small power systems (e.g. Hawaii) where operating reserve is very expensive.

At the end of the day, the biggest problem is that anything flywheels can do, batteries can do too - and batteries are getting cheaper every year.

They do this in datacenters - mostly I think to bridge the gap between a power outage and diesel generator startup time.

You might also be interested in this:


“In the first four months of operations of the Hornsdale Power Reserve (the official name of the Tesla big battery, owned and operated by Neoen), the frequency ancillary services prices went down by 90 per cent, so that’s 9-0 per cent. And the 100MW battery has achieved over 55 per cent of the FCAS revenues in South Australia. So it’s 2 per cent of the capacity in South Australia achieving 55 per cent of the revenues in South Australia.”

I was recently reading about power factor correction and found out that "synchronous condensers", basically just large synchronous motors with no load, are sometimes used at utility scale for PFC and frequency stabilisation. Eg, here: http://www.think-grid.org/synchronous-condensers-better-grid...

Since it seems like the main source of stability in that system is the inertia in the rotor, would it be fair to describe it as a kind of flywheel? I didn't see anything about connecting an actual wheel to such a system, but it seems like it would be the same thing with more inertia, right?

Where "flywheels" in common parlance differ from synchronous condensers is that condensers run at zero torque - so they provide no active power to the system. They provide reactive power, which is needed to regulate and maintain the stability of the power system, but not active power which is used to match generation and demand or shift load. There's no actual source of energy being fed into a synchronous condenser. In a flywheel, you're drawing energy from the grid to spin up a really big mass and then storing it in rotational inertia so you can output it later very quickly. Flywheels can provide reactive power too, through their DC/AC power converters, but since you don't actually need any rotating mass to do that (recall, reactive power requires no torque), you can use a STATCOM - which is functionally like a synchronous condenser just without any moving parts.

Yes, flywheels are used for phase matching at some grid intersections. For example, http://beaconpower.com has installations that do this.

There are FERC rate schedules for this service that make it a profitable application for flywheels.

I mostly hear about flywheels used in datacenters, to bridge the time from a utility outage to diesel generator startup. That doesn't happen very often at the typical datacenter -- monthly testing and maybe a real outage or two per year.

Hospitals as well. That's called 'ride through'.

Another cool use is in railway stations to store the energy of a train as it slows into the station, then return it when it accelerates away.

One of the reasons this was needed for a long time is that DC traction power systems used rectifiers that only worked one-way, so you couldn't just spit the regenerated braking energy back out on the main AC grid. So without somewhere to store it, regenerative braking would cause temporary overvoltages on your traction power system.

With modern power electronic converters, it's probably cheaper and easier to just use the main AC grid as the sink for the regen braking power.

I'd read about European high speed trains using regenerative braking and the grid, thanks for explaining why that wasn't what was being talked about! I had no idea about this earlier system.

How does this work? If the flywheel is in the station, how does it “catch” the train’s energy?

They are on a common electrical system. The train uses regenerative braking to put power into the 'grid'. The flywheel senses this and draws power to spin up. When the train tries to draw a huge current to speed up, the flywheel senses this and dumps power to meet the demand.

Makes sense. I was imagining some kind of steampunkish mechanical contraption.

There was something a bit like that for trucks a few years ago, that captured pressurised exhaust while the truck was engine braking and then used it (not sure exactly how) to help the truck accelerate again. Apparently it saved significant amounts of fuel in stop-start situations (rubbish trucks etc.), not sure what happened to it though.

Giant gears that mesh with the sides of the train would be pretty awesome.

Well there's rack railways[1] which are pretty close. I've been on one in India[2] and it was amazing, although a bit scary when (halfway up) one of our fellow passengers told us that one of the trains on that route had derailed and fallen into a ravine a few months prior, killing everyone on board...

[1] https://en.wikipedia.org/wiki/Rack_railway

[2] https://en.wikipedia.org/wiki/Nilgiri_Mountain_Railway

I've heard of such systems using pressurized hydraulic fluid in delivery vehicles. Feeding the energy back into the car's kinetic energy worked like an automatic transmission's torque converter, I believe.

Yeah, thinking about it I'd guess that the exhaust-gas system didn't take off because of noise. Exhaust brakes are noisy and all the applications that made sense were suburban or inner city. Hydraulic would be much quieter.

Or battery systems, like cars. "mild hybrids" only need a relatively small battery. It's a common technology on buses, for example, which do a lot of stop-and-go.

> You also need uninterruptible power supply to maintain safety and control systems when grid power is unavailable since you've still gotten a huge spinning mass that you can't slow down without somewhere to send the energy

Why? If it's safe in it's vault when powered, why is it unsafe when in it's vault unpowered? Hell, you could use one flywheel to power the rest, and as they lose too much velocity the next one becomes the generator, until either they're all spun down, or power comes back, then you have at least some of the flywheels ready to go. Now, let's say you just don't do anything, they're spinning away slowly slowing, then power comes back and they don't have to spin back up from a dead stop. Why is that not true?

The rotor spins in a vacuum to reduce standby friction losses, and it requires power to maintain that vacuum. The bearings also need active cooling. Without these support systems you won't have a catastrophic failure, but you'll be damaging your equipment and reducing its lifespan.

As for using the flywheel(s) themselves as the source of backup power, that was our original design and definitely feasible at a conceptual level, but there's a lot of engineering in getting that to work properly while maintaining grid code compliance. You need your grid-tie inverter (which also provides the 60 Hz AC used by the support systems) to disconnect from the grid and transition to island mode /without interruption/ very, very fast (since utilities have standards on how fast generators need to disconnect during a system fault) and basically it required us to write our own firmware for the VFDs we were using which in turn invalidated their safety certifications. So definitely a solvable problem but we just didn't get there.

I would think that a momentary loss of power wouldn't be a big deal, even with active cooling and a vacuum pump. As long as your disconnect is fast, if it takes 500-1000ms for the pump and cooling to come back online from flywheel power, that seems like a much easier solution than worrying about five nines. The wear in that second can't be significant.

Gut feel suggests flywheels would last longer, be less expensive to recondition.

I can only make a (barely) educated guess at the difference between Li-ion battery cycle life (single digit thousands of cycles to 80% capacity seems to be what I see everywhere?) compared to bearing replacement schedules and motor/controller maintenance (and I don't have even best guess anecdotal data for this? A little Googling suggests some Rolls Royce airliner jet engines have 15,000 hours between overhauls, but that at least one has made 42,000 hors without an overhaul).

I'd _guess_ you probably don't dump energy back into the flywheel as fast as you pull it out? (I base this on calculating a Tesla 100KWhr battery requires 600(+)KW to recharge in 10 mins, and if you could pull that off the grid easily, you'd just do that. They seem to get enough grid power to charge a Tesla in ~1hr, so they've got ~100KW available I guess?) For back of the envelope calculations I'm gonna use 1 hour as "one cycle" (discharge in 10 mins, recharge in 50 mins seems a reasonable/conservative estimate) - that'd implie a flywheel with similar bearing longevity to a 747 engine bearings would last about 10 times as long (15k - 40k hours) as a Li-ion battery takes to drop to 80% capacity (say 1.5k to 4k cycles?).

The big difference would be a flywheel with new bearings is "as good as new", whereas there's nothing besides replacing the Li-ion battery that gets it back to new.

Pretty sure "charge directly off the grid" is the optimal option for "supercharger like charging stations" (perhaps not for the grid operator), but if you want 600+KW per charging station, and the grid cannot deliver that (economically) where you need it, I'd be surprised to find flywheels would come out something like an order of magnitude cheaper to operate long term than Li-ion battery storage.

(But I'm certainly not a "Former flywheel energy storage startup engineer" - I'd love to know where I screwed up my calculations to indicate and order-of-magnitude benefit that _probably_ doesn't exist???)

Your numbers are reasonable. 5-10x as many cycles on a flywheel versus best-available Li-ion tech was what we figured, too. The initial construction cost per kWh is much higher for a flywheel, though, basically enough to wipe out that advantage. Also, even if a flywheel is cheaper in the long run, it's a tradeoff of upfront capital cost vs long-term maintenance costs - and when you refurbish your battery in 10 years, cells are probably going to cost 20-30% less than they do today.

The other problem we had was that we were making 10s of flywheels per year and competing against Samsung and LG's battery manufacturing efficiencies. And buying an ultra-low-volume product from a startup that might not be around to maintain it in 20 years is also a tough sell in the risk-averse power industry.


As usual - the actual problem isn't reflected in "assume a perfectly spherical cow of uniform density" physics, or in the "Hey, I know a tiny bit about a related problem to this, I'll just extrapolate from there, WCPGW?" analysis.

I reach towards both those oversimplifications way to often.

Glad to see the guestimates/calculations I did bares at least an order-of-magnitude level of correctness. I _mostly_ do these quick calculations so I can rule out thinking harder about things that are 3 or 4 orders of magnitude away from possible.

"Yeah boss, we can do that, it'll take <scrible scrible, estimate, google, calculate, double check> something like $800k of Amazon resource per month, maybe only half a mil if we commit to 12 months up front. How much did you say we could sell this for? To how many customers?"

How did you feel working on a project you obviously have a lot of knowledge in but don’t believe in long term?

Happy that I was learning a lot by getting to do hands-on product design but constantly checking in with my professional contacts to see if there would be somewhere for me to land if it all went belly-up. ;)

>You also need uninterruptible power supply to maintain safety and control systems when grid power is unavailable since you've still gotten a huge spinning mass that you can't slow down without somewhere to send the energy (it's possible to use braking resistors, but it's another cost).

The braking resistors are however a much lower cost, which for this (infrequently used) application is mostly what matters.

It seems like spending >10x more on a contingency system just to save a few hundred kWh (costing a few tens of dollars) every few years seems suboptimal. Especially since those batteries could be in a daily cycling installation, so the opportunity cost (compared to using those batteries elsewhere) is very high.

Why buy X kWh of batteries to sit idle 24/7/364, plus X kWh of flywheel storage? Why not A) eliminate the flywheel, use the battery, and be done with it? Or B) use the braking resistor? It seems like either A or B should always be preferable to a "hybrid" given reasonable assumptions.

Favoring energy recapture over resistor heat dump seems like very suboptimal high-level design coming from a flywheel storage engineer, so what am I missing here?

It's not X kWh of batteries for X kWh of flywheels, it's 0.01 kWh of batteries to provide enough power to run the flywheel's protection and control systems while it 'freewheels'. Ideally, you don't want to brake the flywheel while the grid is down, because then you're losing all the storage energy and need several minutes to spin back up when the grid is back.

Surely you can power the control electronics from the flywheel, right? That should always work except when the thing is stopped, which is a safe state.

in failsafe systems it's best not to depend on the thing you expect to fail to behave in a particular way when it... fails, redundancy is the key.

The grid is the primary power source. Flywheel when no grid. Battery would be third. It might be enough to have a passive resistor load for the third option. But at the scale of these things, a battery to run a few hours of controller is not a big deal.

See my comment above on this: https://news.ycombinator.com/item?id=17071303

Some kind of frictionless magnetic bearing is not possible?

It is, and there is actually a really neat passive and stable solution to this (circular Halbach array with compensation coils running from one side to the other).

"Electrodynamic bearings" is the keyword.

It's possible and it exists in other applications but my understanding is that the costs were prohibitive for our product (~500 kWh flywheels).

I saw a startup, Velkess with a flexible lasso type flywheel (safety), I think 15 kWh. It folded though.

Do you know any companies doing in the 15kWh ballpark type flywheel batteries? It's an idea I really like - I wanted to pair it with a small hydropower installation.

I am a member of a forum for professional car mechanics. You would not believe the number of people who abuse their car by driving with brake rotors down to ridiculous thicknesses, or only changing their oil once in 40,000 miles.

Consumers are bad at maintaining mechanical things.

Think about this for a moment...

Now Google "flywheel explosion".

We now have government-mandated maintenance standards for some types of power utility equipment (mostly protection systems) in North America because it turns out even big companies with hundreds of engineers on staff sometimes aren't very good at maintaining stuff.

You seem to be the correct person to ask what are the do's and dont's regarding lithium battery maintenance and to increase device (say mobile phone) battery capacity?

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact