TECHNICAL PAPER

Development of Compact Flywheel Energy Storage System (ComFESS)

A. KUBO H. KAMENO

R. TAKAHATA

The objective of this research program is to develop a flywheel energy storage system that can replace the lead battery,
which has inherent problems in regard to maintenance and environmental protection. The basic development target has been
a "1 kWh class system (ComFESS) capable of providing a backup power supply of 300 watts for approximately 3 hours," for
which challenges were reduction of windage loss and rotation loss. Therefore, the flywheel is housed in a vacuum container,
and an active magnetic bearing has been adopted to support the flywheel without contact.
This paper concerns the watt loss as related to the specifications of the active magnetic bearing (AMB) and
motor/generator. Specifically, discussion will be focused on the approaches employed to cut down on overall watt loss of the
flywheel system to approximately a quarter of that involved in the conventional system. This improvement stems mainly from
the reduction of AMB loss achieved by means of the zero-power nonlinear control.
This zero-power nonlinear control, being able to dispense with the bias current, enables both AMB loss and power
consumption for AMB control to be minimized.
Key Words: flywheel energy storage system, ComFESS, active magnetic bearings zero-power nonlinear control,
hysteresis motor

1. Introduction

2. Target of Loss Reduction

Along with the advance of the IT revolution, the impact of

disrupted communications due to blackouts and the like on
social and personal life has become extremely serious. That is
why many lead batteries are currently used as the backup
power sources, and rapid expansion of their usage is expected.
However, the lead batteries have service life of only several
years (3~4 years) and require maintenance. To make matters
worse, as they contain chemical substances and heavy metals
that are hazardous to the human body and plants, the batteries
cannot help facing the environmental problems.
Under such circumstances, clean, high-performance energy
storage systems that can replace the lead batteries having
maintenance and environmental problems are demanded. The
flywheel energy storage system is one of the candidates,
regarding which development has been progressing
worldwide1).
This research program has been aimed at development of
elemental technology that would comprise a compact flywheel
energy storage system (hereinafter referred to as
"ComFESS"), which would be a 1 kWh class system capable
of supplying 300 W power for approximately 3 hours,
replacing the lead batteries. In order to ensure 3 hours backup
power supply, it is necessary to minimize the windage loss
and bearing loss of the ComFESS. Therefore, the flywheel is
located in a vacuum chamber and the flywheel rotor is
supported by active magnetic bearings (hereinafter referred to
as "AMB") that provide non-contact support with minimum
loss.

With a view to constructing a system of 300 W 3 hours, a

conceptual system design was first worked out2). The structure
of the ComFESS is shown in Fig. 1.

Koyo Engineering Journal English Edition No.167E (2005)

Main shaft upper side

Flywheel

Hub

RaAMB unit

Vacuum
container
Axial AMB
Lower side RaAMB unit

Motor / Generator

Fig. 1 3D drawing of ComFESS

To make the system compact, we incorporated a flywheel

440mm in diameter made of CFRP (carbon fiber reinforced
plastic). Inside the bore of the flywheel is mounted the rotor
unit comprised of the hub and the shaft. The rotor unit is
accelerated up to 24 000 min1, and the electric power is taken
out while the unit is subsequently decelerated. Here, however,
a part of the energy stored in the system is consumed as the
driving power for the AMB controller and the inverter. Taking
this into account and assuming that the power of 300 W 3
hours is taken out, calculation is made to approximate the
rotational decay of the rotor as shown in Fig. 2. In this figure,
the "+" marked curve represents the rotational decay pattern

29

Development of Compact Flywheel Energy Storage System (ComFESS)

for the case of conventional AMB technology and the

motor/generator. Specifically, the AMB relies on the linear
3)
control method consuming bias current , whereas the
motor/generator is such that the induction motor is driven by a
conventional electric converter. Thus, the figure indicates that
the conventional system can take out the power only for 1.5
hours. Table 1 summarizes various losses estimated for the
1
rotational speed condition of 24 000 min . The values in the
"conventional" column are for the conventional technology
while those in the "Target" column are the values at which the
new system is aimed.
25 000

15 000

3. 1 Motor/Generator & Converter

As motor/generator candidates for this system, a hysteresis
motor and permanent magnet motor (hereinafter referred to as
"PM") were comparatively appraised. As a result, the
hysterisis motor was selected because of its having the
following features:
Simple structure with excellent cost performance
Smaller loss (PM motor has 4~5 times greater loss)
No need to develop converter (IGBT type converter
available in the market was selected)

3. 2 Magnetic Bearing (AMB)

10 000
5 000
0

In an attempt to achieve the above target, we worked on

design of ComFESS. The results are as follows.

Time, hours

Fig. 2 Estimated rotational decay of the rotor

Table 1 Classification of ComFESS loss

@24 000min1

Conventional

Target

[W]

[W]

500

100

50

50

Loss of motor / generator

200

95

Total

750

245

Loss of AMB
Windage loss (at 1Pa)

The target value for loss of AMB is set around 100 W,

which is expected to be attained by adopting the non-linear
control (zero-power control)4) and other improvements. The
windage loss of 50 W is based on the pressure of 2 Pa, which
cannot be further improved. On the other hand, the loss of the
motor/generator is targeted for around 95 W by adopting the
motor with smaller loss than the induction motor. The
rotational decay curve expected if the targets in Table 1 are
achieved is plotted in Fig. 2 with "" marks. This result
indicates that the taking power of 300 W 3 hours would be
attainable5).

1) PM Biased axial magnetic bearing (AxAMB)

In this project, a PM biased axial magnet (PM biased
AxStator) that allows PM attracting force was adopted, which
can substitute for the magnetic force required to sustain the
dead weight of the rotor unit.
2) Radial active magnetic bearing (RaAMB)
A pair of radial active magnetic bearings provided on each
of the upper and the lower parts of the spindle respectively are
also required to save energy consumption. In this project,
therefore, the "zero-power non-linear control method" was
adopted to minimize the bias current of RaAMB (total current
supplied to RaAMB, I = bias current, I0 + control current, Ic).

3. 3 Rotor Unit
The rotor unit of ComFESS is comprised of the shaft, the
hub and the flywheel. Its natural frequency was calculated as
shown in Fig. 3. The main specifications of the ComFESS are
summarized in Table 26).
200

# : Bearing
: Sensor

150
Rotor unit radius, mm

Rotational speed, min

: Target
: Conventional

20 000

3. Design of ComFESS

100

f2B : 1 185Hz (1st Bending mode of shaft)

50
#

50
f1B : 145Hz (1st Bending mode of hub)

100
150
200
0

100

200

300

400

500

Rotor unit length, mm

Fig. 3 Result of mode shape simulation

30

Koyo Engineering Journal English Edition No.167E (2005)

Development of Compact Flywheel Energy Storage System (ComFESS)

Table 2 Main specifications of ComFESS

Rotor
Total mass of rotor:

75kg

Main shaft:

Outer dia. 75mm

controller had a loss of 20 W, which clearly was far smaller

than the target of 70 W. Thanks to this, the total measured loss
on ComFESS BB was 300 W, which met the target of 345 W.
Table 3 Measurement of ComFESS BB loss
@24 000min1

Length 258mm
Flywheel:

Outer dia. 440mm

Target

Inner dia. 300mm

Length 260mm
Energy storage capacity

900Wh (effective)
1

Kinetic energy of the rotor 1.6kWh (24 000min )

0.1kWh (5 000min1)
Power capacity
AxAMB

ComFESS BB
Measurement

Motor / Generator (W)

25

30

Motor / Generator controller (W)

70

20

Motor / Generator total loss (W) A

Ball bearing (W)
Windage loss (W)

AMB (1 DOF)

Rotational loss (W)

PM biased AxAMB

Total loss (W)

95

50

200

200

50

50

250

250

A+B

345

300

Control method:

Nonlinear control

RaAMB

AMB (4 DOF)

4. 2 Test on PM Biased Axial Magnetic Bearing

Control method:

Linear control

Electromagnets:

Hetero-polar

Rotor lamination:

Silicon steel plate

The rotor unit of the ComFESS was successfully levitated

stably. At that moment, the waveform of electric current was
measured with the results as shown in Fig. 4.

Touch down bearings

Emergency support for Radial

5.86

Upper side:

direction

3.91

Lower side:

Radial direction and axial

Power supply for magnets:

80~150DCV

Max. current for AMB:

40A (total)
8A (max.) for each electromagnet

Amplifier:

PWM

Total current, A

direction
AMB controller

Mean : 0.22A
1.95
0
1.95
3.91
5.86

0.2

0.4

0.6

0.8

Time, sec

4. Basic Element Tests and Results

(a) PM biased AxAMB

4. 1 Measurement of Loss
Each category of loss measured on the ComFESS BB at
24 000 min 1 is shown in Table 3. The motor/generator

Koyo Engineering Journal English Edition No.167E (2005)

5.86
N = 0 [rpm]

4.88
Total current, A

Basic element tests were carried out to confirm the

performance of each element with basic design. Therefore,
three kinds of test machines were designed and manufactured.
The first one was ComFESS BB. The flywheel was
comprised of CFRP and the main shaft including the flywheel
and hub was supported only by ball bearings. This test
machine was used for testing such elements as the motor,
flywheel and vacuum chamber.
The second test machine was a dummy ComFESS using a
bakelite flywheel and active magnetic bearing. It was used for
element tests of AMB and the natural frequencies of the rotor.
The third test machine was a flywheel AMB system
(AMX095) designed to evaluate the zero power control.
After all element tests, ComFESS with AMB and CFRP
flywheel was assembled by using some parts of Dummy
ComFESS and ComFESS BB as shown in Fig. 3.
The results of the basic element tests are as follows.

3.91
2.93
1.95
2.50A

0.98
0

0.2

0.4

0.6

0.8

Time, sec
(b) Conventional AxAMB

Fig. 4 Measured results of PM biased AxAMB

The total current supplied to the PM biased AxAMB was

significantly influenced by the noise. Nevertheless, the
average current was held around 0.22 A which was far lower
than approx. 2.50 A required with the conventional AxAMB,
confirming the energy saving effect of the PM biased
AxAMB.

31

Development of Compact Flywheel Energy Storage System (ComFESS)

4. 3 Element Test on Zero-Power Control of Radial

7)
Magnetic Bearing

x=110Hz

0.5
yl, m

0.5
1

x=110Hz

10

0.5
yu, m

1) Flywheel AMB system (AMX095)

Figure 5 shows the modeling simulation and figure of the
flywheel AMB system used for the experiment. This flywheel
AMB system consisted of a CFRP flywheel AMB, a control
unit and a high-frequency inverter. The parameters for this
flywheel AMB system are shown in Table 4.

10

0.5
0.5

0
xu, m

0.5

1
4
10

0.5

0.5
1
0
4
xl, m
10

(b) Lower magnetic bearing

(a) Upper magnetic bearing

Fig. 6 Orbits of rotor (6 600min1)

Magnet3

f3

y hy
Magnet7
i7
x0
U7
f7

x=100Hz

i1

x0

U1

f1
Magnet1
Magnet5 i5
x0

U5

Sensor2
(a) Model of system

(b) Picture of system

Fig. 5 Flywheel-AMB system (AMX095)

Table 4 Parameters of rotor-AMB system

Symbol

Value

Unit

13.672

kg

Ir

0.173

kgm

Ia

0.186

kgm

Lu

0.0499

Ll

0.1676

L1

0.02535

Ku

4.47106

Nm2/A2

Kl

4.47106

Nm2/A2

X0, Y0

0.2510

2) Test results
For this test, a feed back control system was constructed by
means of a digital signal processor (DSP). This control system
receives input of 4-direction displacement information from 4position sensors and gives output control current to 8electromagnets, i.e. a 4-input-8-output system. The control
output is supplied to the electromagnet via a D/A converter
and power amplifier.
The control performance was evaluated by the orbit
waveforms of the rotor, of which examples recorded at 6 600
min1 for upper and lower magnetic bearings were shown in
Fig. 6. Figure 7 shows the waveforms of the control current
on two electromagnets facing each other at the lower position
of the spindle. These waveforms demonstrate that at each
moment only one of the two electromagnets is energized in
accordance with the displacement of the rotor4), or that the
basic control action is properly performed.

32

1.6

1.6

1.2

1.2

0.8

0.8

0.4

0.4

f5

x=100Hz

i7, A

L1
L2

L1 Sensor1
i3
x0
U3

i5, A

Lu

0 0.02 0.04 0.06 0.08 0.1

Time, s

0 0.02 0.04 0.06 0.08 0.1

Time, s

Fig. 7 Control current (6 600min1)

However, the measured current value was still high due to

unbalance of the rotor. It is possible to reduce this current by
additionally applying the unbalance force rejection control
(UFRC) which rotates the rotor around the center of inertia
without controlling rotation synchronous component.
An experimental attempt to perform the zero-power nonlinear control utilizing the conventional magnetic bearing
mechanism for the linear control was successful. As shown in
Table 5, the energy consumption was reduced to 60% of that
required for conventional magnetic bearing control. This
technology will make it possible to realize a highly efficient
flywheel energy storage system that has not yet been achieved.
It is also confirmed that the new technology can be applied to
such other systems that involve high gyroscopic action or an
elastic rotor. As further improvement is expected at higher
rotating speed, it is conceivable to realize perfect zero-power
control.
Table 5 Rotational speed vs. power consumption
Rotational speed (min1 )
Zero-bias method (W)
Bias, method (W)
Energy reduction (%)

600

1 200

1 800

64.8

64.4

64.4

70.9

112.2

111.7

112.1

112.1

48

47

48

41

Koyo Engineering Journal English Edition No.167E (2005)

Development of Compact Flywheel Energy Storage System (ComFESS)

5. Feasibility of 300 W 3 hours System

(ComFESS)

6. Conclusion

Discussion on the current technology level and issues to be

addressed to realize the targeted system are shown as follows:
Applying the findings of the basic element tests to the
ComFESS, rotational decay of the rotor was calculated with
the results as shown in Fig. 8. Supposing the effect of the
hysteresis motor to be added on the rotational decay curve for
the conventional technology ("+" marked), a new curve
marked with "&" can be drawn. On top of that, if the effect of
PM biased AMB and zero biased control at the elemental
development level so far are added, "(" marked and "S"
marked curves are obtained. Obviously, however, they are still
far larger than the target ("" marked). And yet, there is
potential improvement by application of zero-power control
including unbalanced force rejection control ("'" marked
curve), that would make it possible to achieve the target.
In the future, therefore, UFRC development will be pursued
to complete the ComFESS incorporating the insights obtained
in this study.
: Conventional
: Hysteresis motor
: PM bias AMB (bias control AMB)
: Zero bias AMB
: Target
: Zero bias AMB (Unbalance control)

Rotational speed, min

25 000

20 000

As a result of this research work, the following element

technology has been established.
1) Reduction of loss by optimum driving of the hysterisis
motor
2) Reduction of power consumption of the axial magnetic
bearing by utilizing the attractive force of the permanent
magnet for supporting the dead weight of the rotor unit
3) Reduction of power consumption of the radial magnetic
bearing through application of zero-power non-linear
control
On the other hand, issues needing to be addressed for
commercialization have been clarified.
4) For the zero-power-non-linear control to be applied to the
ComFESS, the unbalance force rejection control needs to be
added, that would enable further reduction of energy
consumption.
Based on the findings of this study, the development
program will be shifted to the commercialization phase,
wherein development of commercial products will be pursued.
At the same time, there has arisen a new commercial needs or
development of a power backup system that can provide
several hundred kilowatts for around a minute. To meet this
demand for high-output-short-time backup systems, further
advance of these elemental technologies will be pursued.
This research program has been performed under the
sponsorship of the NEDO (New Energy and Industrial
Technology Development Organization in Japan) as one of the
NEDO International Joint Research Projects (NEDO-Grant
Project) for these three years.

15 000

Acknowledgement
10 000

5 000

Time, hours

Fig. 8 Rotational decay of rotor

Koyo Engineering Journal English Edition No.167E (2005)

Taking this opportunity of publication, we wish to extend

our deepest gratitude to our research partners, Professor
Nonami of Chiba University, Dr. Thoolen of Centrum voor
Constructie en Mechatronica (CCN: the Netherlands),
Emeritus Professor Schloesser of Eindhoven Institute of
Technology (the Netherlands), and President Nishimoto of
NTRK Co., Ltd. for their assistance and support.

33

Development of Compact Flywheel Energy Storage System (ComFESS)

References
1)
2)
3)

4)

5)

6)

7)

R. Takahata: Tribologist, 49, 5 (2004) 416.

H. Kameno, A. Kubo, S. Gachter and R. Takahata: Koyo
Engineering Journal, 163E (2003) 44.
Y. Miyagawa et al, A 0.5 kWh Flywheel Energy Storage
System using A High-Tc Superconducting Magnetic
Bearing, IEEE Transactions on Applied Superconductivity,
9, 2 (1999) 996-999.
S. Sivrioglu, K. Nonami, R. Takahata and A. Kubo:
"Adaptive Output Backstepping Control of a Flywheel
Zero-Power AMB System with Parameter Uncertainty,"
Proceeding of 42nd IEEE Conference on Decision and
Control (CDC), Hawaii-USA (2003) 3942-3947.
H. Kameno et al.: Basic Design of 1 kWh Class Flywheel
Energy Storage System, Proceedings of the Eighth
International Symposium on Magnetic Bearings (ISMB-8),
(2002) 575-580.
A. Kubo et al.: Dynamic Analysis and Levitation Test in 1
kWh Class Flywheel Energy Storage System, Proceeding of
7th International Symposium on Magnetic Technology
(ISMST-7), (2003) 144-149.
R. Takahata, A. Kubo, F. Thoolen and K. Nonami:
Compact Flywheel Energy Storage System, Proceedings of
FY2001 International Joint Research Program (NEDO
Grant) Conference.

A. KUBO*

H. KAMENO* R. TAKAHATA**

Mechatronic Systems Research & Development Department,

Research & Development Center
** Mechatronic Systems Research & Development Department,
Research & Development Center, PhD

34

Koyo Engineering Journal English Edition No.167E (2005)