Flywheel Energy Storage

A battery without chemistry

A chemical battery stores energy by holding a difference in chemical potential between two electrodes; you charge it by forcing ions one way, you discharge it by letting them go back. The whole thing is fundamentally a kinetic problem dressed up in electrochemistry — and as a kinetic problem it inherits the constraints of chemistry: temperature sensitivity, electrode wear, electrolyte breakdown, finite cycle life, slow charge rates because the ions have to physically move through a membrane.

A flywheel sidesteps all of that. Instead of converting electricity into chemistry and back, it converts electricity into rotation and back. A motor accelerates a heavy spinning rotor; the rotor's kinetic energy is the stored energy; when you need it back, the same machine runs as a generator and the rotor slows down. There is no chemistry to age, no electrodes to corrode, no temperature window to maintain. The only constraints are mechanical — how fast can you spin a chunk of mass before it tears itself apart, and how do you keep it spinning without losing energy to friction.

That mechanical character gives flywheels their distinctive performance envelope. They charge and discharge in milliseconds because nothing has to diffuse anywhere. They survive a million cycles because no material is being repeatedly chemically reconstituted. They achieve 90+ percent round-trip efficiency because the only loss is electrical-to-mechanical conversion, which is well-understood and very efficient. And they are deeply indifferent to ambient temperature — a flywheel cares whether its bearings are lubricated, not whether the air around it is at -20 or +60 degrees Celsius.

The governing equation

The kinetic energy of a rigid body rotating about a fixed axis is

E = ½ I ω²

where I is the moment of inertia and ω the angular velocity. The moment of inertia is the rotational analogue of mass: it measures how much of the body's mass is far from the axis. For a uniform solid disk of mass m and radius R, I = ½ m R². For a thin-walled cylindrical shell — closer to a real composite flywheel rotor — I = m R². So for a thin-walled rotor

E = ½ m R² ω² = ½ m v_tip²        (v_tip = ω R)

The specific energy ε = E/m therefore equals one half the square of the tip speed. The whole game of flywheel engineering is to push v_tip as high as the material will allow without bursting.

Concretely: a 1100 kg composite rotor running at 16,000 rpm with an effective radius of 0.4 m has a tip speed v_tip = ω R = (16000 × 2π/60) × 0.4 ≈ 670 m/s. Specific energy ε = ½ × 670² ≈ 2.24 × 10⁵ J/kg ≈ 62 Wh/kg. Total stored energy ≈ 68 kWh. The Beacon Power production unit advertises 25 kWh because it cycles between roughly 16,000 rpm (full) and 8,000 rpm (empty) — the bottom quarter of the energy is structurally locked in to keep the rotor under tension. You only get to use E_full − E_empty = (3/4) × E_full.

Why tip speed is the real currency

The bursting load on a spinning rotor is the hoop stress in the rim. For a thin rotating ring of density ρ at radius r spinning at angular velocity ω, the hoop stress is

σ_hoop = ρ ω² r² = ρ v_tip²

The rotor fails when σ_hoop exceeds the material's ultimate tensile strength σ_max. Setting σ_hoop = σ_max gives the upper limit on tip speed:

v_tip,max = √(σ_max / ρ)        (specific-strength bound)

The quantity σ_max/ρ — specific strength — is therefore the figure of merit for a flywheel material. The specific energy a rotor can store scales as ε ∝ v_tip² ∝ σ/ρ. This is why no amount of mechanical cleverness lets a steel rotor catch a composite rotor: it is bounded by physics.

Rotor material	Density ρ (kg/m³)	σ_ult (MPa)	σ/ρ (kJ/kg)	Practical ε (Wh/kg)
Mild steel (AISI 1020)	7850	400	51	~ 5
Maraging steel (300)	8100	2000	247	~ 30
Titanium 6Al-4V	4430	1170	264	~ 35
S2-glass / epoxy (filament-wound)	1900	1700	895	~ 60
Carbon fiber T700 / epoxy	1550	2400	1548	~ 130
Carbon fiber T1000 / epoxy	1600	3500	2188	~ 200

The "practical" column applies a derating factor of roughly 0.25 to (σ/ρ) — accounting for the difference between ultimate strength and design allowable, the fact that real rotors are not infinitely thin rings, and a safety margin against fatigue. Each step down the table represents a real engineering generation: industrial steel flywheels of the 1960s, the maraging-steel rotors of early aerospace systems, glass-fiber rotors of the 1980s American Flywheel Systems era, and the carbon-fiber composite rotors that dominate today's commercial offerings.

How a composite rotor is built

A real flywheel rotor is not a homogeneous ring of carbon. It is a filament-wound composite, in which thousands of high-strength fibres are wound circumferentially around a mandrel and impregnated with epoxy. The geometry exploits the directional strength of the fibre: hoop stress, the dominant loading, pulls each fibre along its strong axis (tension), not across it (much weaker). Radial stress — the secondary loading that pulls layers apart — is small in a thin ring but rises quickly in a thick rotor; designers manage it by tapering the rotor, using lower-density fillers in the inner core, or building the rotor as a stack of thin-walled shells press-fitted together with controlled radial pre-stress.

The benign failure mode that makes composites safe in this application is a direct consequence of construction. When a steel rotor over-stresses, a crack propagates and the rotor parts into a few hardened-steel pieces that retain almost all their kinetic energy. When a filament-wound composite rotor over-stresses, the fibres separate one at a time — the result is a cloud of fibre fragments and shredded epoxy that the containment can absorb. The energy is the same; the partitioning is what matters. The hazard analysis for a flywheel facility hinges on this distinction, and regulators in some jurisdictions (notably the EU and California) effectively require composite rotors for systems above a few kWh.

Magnetic bearings and the vacuum housing

A flywheel that loses 10 percent of its energy per hour to bearing friction and air drag is not useful for grid storage. Two engineering systems bring idle losses down to about 1 percent per hour.

Active magnetic bearings (AMBs) levitate the rotor electromagnetically, with no physical contact between the rotor and any bearing race. A set of electromagnets is controlled in feedback by position sensors that detect the rotor's deviation from centre; the controller drives the magnets to restore it. There are no mechanical losses because there is no contact; the only loss is small eddy-current drag in the rotor's iron laminations. Commercial AMBs hold position to ~10 µm and tolerate the heaviest aerospace rotors. Touchdown bearings — a set of mechanical bearings sitting a small clearance away — catch the rotor only if the AMBs fail.

The vacuum housing addresses aerodynamic drag. At 670 m/s tip speed in atmosphere, the rotor would be doing supersonic work on the surrounding air — the drag would be enormous and the heat generated unmanageable. Pumping the housing down to ~10⁻³ Pa (a millionth of atmospheric) reduces aerodynamic power loss by roughly six orders of magnitude. The vacuum also eliminates oxidation of the composite at the running temperature. A single roughing pump plus a turbomolecular pump maintains the vacuum against the small outgassing rate of the composite and epoxy.

Charge and discharge — how it actually moves power

The same machine that spins the rotor up draws it back down. A permanent-magnet synchronous machine (PMSM) mounted coaxially with the rotor acts as a motor during charging and a generator during discharge. A bidirectional power electronics converter (typically an IGBT-based active front-end) sits between the machine and the grid bus; it controls torque on the machine and shapes the grid-side current to whatever frequency and phase the grid demands.

The power available is limited not by the rotor but by this converter — that is why flywheels can respond in milliseconds. The rotor itself can ramp from 5 MW out to 5 MW in (a swing of 10 MW) in the time the converter takes to flip its IGBT polarity, well under a second. Compare this with a chemical battery, where the maximum C-rate is bounded by ion mobility through the electrolyte. For grid frequency regulation — where the system is correcting tens-of-megawatts of imbalance every few seconds — this responsiveness is the killer property.

Round-trip efficiency, end to end, is the product of the conversion efficiencies on the way in and on the way out: AC grid → converter (~98%) → PMSM as motor (~96%) → kinetic energy → PMSM as generator (~96%) → converter (~98%) → AC grid. The product is roughly 0.88 over a full cycle if the rotor sits idle for one hour; closer to 0.92 if the cycle is sub-second. There are no chemistry losses — every joule that comes out of the converter on charge goes back into the rotor as kinetic energy. The losses are all in the electrical conversion stages and in slow leakage from the magnetic bearings and residual aerodynamic drag.

Where flywheels show up

• Grid frequency regulation. Beacon Power's Stephentown, NY plant operated 200 composite-rotor flywheels providing 20 MW of ancillary services to the NYISO from 2011, with a second 20 MW plant in Hazle Township, PA. Each rotor: ~1100 kg, 16,000 rpm, 25 kWh. The plants earn "regulation" revenue by responding to NYISO's 4-second control signal — the kind of millisecond ramping that is uneconomic for spinning gas turbines.
• Motorsport regenerative braking. Formula 1's KERS regulations (2009-2013) allowed teams to store 400 kJ per lap and discharge 60 kW for 6.7 seconds. Williams Hybrid Power built a 60,000 rpm composite flywheel KERS for Porsche's 911 GT3 R Hybrid and Audi's R18 e-tron quattro at Le Mans 2012. The unit weighed roughly 25 kg and stored ~120 Wh — small numbers, but at the power densities racing demands, no chemical battery could match it.
• Rail brake-energy recovery. Trackside flywheels (e.g. Vycon REGEN units on the LIRR, Adgero on London Underground extensions) capture braking energy from approaching trains and dump it back into the catenary as the next train accelerates. Cycle life of 10⁶ is the killer feature: trains stop and start hundreds of times a day, year after year.
• Uninterruptible power supply. Active Power and Piller Power Bridge sell flywheel UPS units that bridge the 5-15 seconds between a utility outage and a diesel genset coming online. The flywheel replaces a battery bank that would otherwise need replacement every five years and a climate-controlled room.
• Pulsed-power experiments. The JET tokamak (UK) used a pair of 500-ton flywheel-motor-generator sets to charge its toroidal field coils; ITER uses the same architecture at higher capacity. The flywheel decouples a 400 MW pulsed load from a grid that could not supply it directly.
• Spacecraft attitude control. Every "reaction wheel" on a satellite is a small flywheel. By spinning it up or down, the spacecraft develops a reaction torque that rotates the body without expending propellant. NASA's G2 program at Glenn Research Center demonstrated using larger flywheels for combined energy storage and attitude control on the ISS, exploiting the fact that the same hardware can do both.

Flywheel versus battery — the trade table

Property	Composite flywheel	Lithium-ion battery
Energy density (cell)	100 – 200 Wh/kg	200 – 300 Wh/kg
Power density	5 – 10 kW/kg	0.5 – 3 kW/kg
Round-trip efficiency	> 90 %	85 – 95 %
Cycle life	~ 10⁶	10³ – 10⁴
Calendar life	20 + years	5 – 15 years
Self-discharge	~ 1 % / hr (huge)	~ 5 % / month
Operating temperature	−40 to +60 °C	+10 to +40 °C optimal
Response time	< 5 ms	10 – 100 ms
Failure mode	Composite shred (benign)	Thermal runaway (severe)
Best at	Power, short duration, many cycles	Energy, long duration, few cycles

Read it from the bottom row: a flywheel is the right answer when you need many megawatts for a few seconds many times a day; a lithium battery is the right answer when you need many kilowatt-hours over many hours, fewer times per day. Grid frequency regulation, regenerative braking, and UPS bridging sit firmly in the first regime. Solar load-shifting and electric-vehicle range sit firmly in the second. Hybrid systems exist (a lithium pack plus a flywheel buffer) but the technologies are complementary, not competitive.

The bullet-times-a-thousand problem

A 100 Wh/kg rotor stores 3.6 × 10⁵ J per kilogram. A typical 9 mm rifle bullet has a muzzle energy of ~600 J. So a kilogram of fully-charged flywheel carries roughly 600× the energy of a bullet, and a 1100 kg rotor at full speed has the energy of about 700,000 bullets. If you allow that to escape uncontrolled, you have a serious industrial accident. The containment design is therefore the most safety-critical part of the system.

Containment philosophies fall into three categories. Surface installations (large industrial flywheel plants) use thick steel vessels sized to absorb the kinetic energy as plastic deformation if the rotor disintegrates; design margins of 2× to 5× the rated kinetic energy are common. Buried installations (Beacon Power's Stephentown plant) place the housing below grade, using soil and reinforced concrete as the outer containment layer; this is the cheapest option per kWh for utility-scale systems. Manned-environment installations (vehicle KERS, UPS units, satellite reaction wheels) use multi-layer composite-and-steel containment and additional administrative safety controls — interlocked enclosures, position-monitored exclusion zones during spin-up.

In practice, the safety record has been good. The most-cited failure incident — a Beacon Power rotor at Stephentown in July 2011 — was fully contained underground; the only external evidence was a thump and a vibration signal. No injuries, no released energy beyond what was absorbed by the housing and soil. The composite construction was credited with the benign outcome; a steel rotor of the same energy would have produced a far more violent containment event.

Common pitfalls and misconceptions

• Confusing power density with energy density. A flywheel beats a lithium battery on kW/kg but loses on Wh/kg at the system level once you include housing, vacuum hardware, and converter mass. Pick the metric that matches your application; do not generalise.
• Forgetting the depth-of-discharge limit. A flywheel is rarely cycled down to ω = 0. Tip-speed pre-tension and structural integrity often require staying above 30-50% of rated rpm, which means you only access (1 − 0.5²) = 75% or (1 − 0.3²) = 91% of the full E = ½Iω². Rated capacity reflects this; nameplate kWh is usable kWh, not stored kWh.
• Underestimating self-discharge. Even at 1% per hour, idle losses are catastrophic for storage of hours or days. Flywheels are not the right technology for solar load-shifting; they shine over seconds-to-minutes durations.
• Treating "no degradation" as exact. The composite epoxy does creep slowly and the rotor's modal frequencies drift over decades. Cycle life of 10⁶ is excellent, but it is not infinite. Periodic balance and modal inspections are part of operating a fleet.
• Forgetting bearings need power. Active magnetic bearings continuously draw a few hundred watts to maintain levitation. That parasitic load eats into round-trip efficiency at long storage times and is a hard floor on self-discharge.
• Mixing up "spinning reserve" and "spinning flywheel". Spinning reserve in grid jargon is unrelated — it means a partially loaded gas turbine ready to ramp up. A flywheel provides ancillary services that are functionally similar but operates on entirely different physics.

Where the technology is going

Three frontiers. First, higher-tip-speed rotors using newer fibres (T1100, M55J pitch-based) and out-of-autoclave winding processes promise composite rotors approaching 300 Wh/kg in research — competitive with the best lithium-ion at cell level and with much longer cycle life. Second, integrated-converter designs that put the power electronics in the rotor hub (not in a separate cabinet) reduce package mass and electrical losses for mobile applications. Third, hybrid storage architectures pair flywheels with lithium banks: the flywheel handles short, peaky transients and shields the lithium from cycle-aging stress, while the lithium provides hours of capacity. Several grid operators are evaluating these hybrid plants for renewable-integration roles — solar smoothing where the second-by-second variability is handled kinetically and the daily energy shift chemically.