Proceedings of the 2008 International Conference on Electrical Machines

Paper ID 852

A Flywheel Energy Storage System with Matrix

Converter Controlled Permanent Magnet
Synchronous Motor
Paulo Gamba (1,2,4), S. Ferreira Pinto (3,4), J. Fernando Silva (3,4) Elmano Margato (1,2,4)
Instituto Superior de Engenharia de Lisboa, DEEA, Rua Cons. Emdio Navarro 1, 1950-062 Lisboa, Portugal
(2)
Centro de Electrotecnia e Electrnica Industrial, Rua Cons. Emdio Navarro 1, 1950-062 Lisboa, Portugal
(3)
DEEC, Instituto Superior Tcnico, TU Lisbon
(4)
Center for Innovation in Electrical and Energy Engineering, Av. Rovisco Pais 1049-001 Lisboa, Portugal
E-mail: pjgamboa@deea.isel.ipl.pt, soniafp@ist.utl.pt, fernandos@ist.utl.pt, efmargato@deea.isel.ipl.pt

(1)

Abstract - This
paper
presents
an
experimental
characterization of a flywheel energy storage system. The device is
based on steel seamless tube mounted as a vertical axis flywheel
storing kinetic energy. The motor/generator is a Permanent
Magnet Synchronous Machine controlled by an AC-AC Matrix
Converter. The matrix control method uses a discrete-time model
of the converter system to predict the future values of the input
currents for all the 27 possible output voltage vectors generated
by the matrix converter. An optimal controller minimizes current
errors using a weighted cost functional. This flywheel and control
process is intended for dynamic voltage restorer (DVR) to
mitigate voltage sags.

based on a weighted cost functional, selects the optimal output

voltage vector together with the input current vector.
In this paper, a description and characterization of an
experimental kinetic energy accumulator, based on FES is
presented. It is also shown that predictive optimal control can
enhance the controllability of matrix converters by optimizing
the vector choice. In the near future FES with matrix
converters will be used as a dynamic voltage restorer (DVR) to
mitigate voltage sags at critical loads (Fig. 1, 2).
II. ELECTROMECHANICAL FLYWHEEL ENERGY STORAGE

I.

INTRODUCTION

In the last years, Flywheel Energy Storage (FES) systems

have been rediscovered by industry due to their advantages in
comparison with other energy storage systems [1], [2] , [3].
FES can be designed to protect critical loads against voltage
sags, by using a Permanent Magnet Synchronous Motor
(PMSM) and a matrix converter to control the power transfer
process between the flywheel and the critical load. This
arrangement eliminates the dc link capacitor needed in
conventional inverter rectifier systems.

A. The concept of Flywheel Energy Storage

An experimental kinetic energy storage (Fig. 3) was designed
and built to have a moment of inertia of 4,2kgm2 and a
maximum rotating speed of 2500 rpm.
A PMSM (2.9kW, 3000rpm) was selected as the
electromechanical energy transfer device. The steel seamless
tube (fig. 3) was selected as a rotating mass, given its
availability and ease of assembly. This part of the device is
characterized by parameters which depend on material (in this
case steel), and parameters and dimensions of the barrel
(Fig.4).
The volume of the steel barrel is:

[ ]

(1)

V = 2 r22 r12 h = 0,0062 m 3

Where r1 , r2 are respectively the internal and external radius of

the barrel. Its mass is:
m = V = 70,0[kg ]

Where = 7,8 10 [kg / m ] is the steel density.

3

Fig. 1. The electromechanical energy accumulator.

The predictive discrete-time control of power converters has

been outlined in [4], [5], [6], [7], and recently applied to matrix
converters [8]. A detailed dynamic model of the matrix
converter and PMSM is used to predict the future values of the
input currents for all the 27 possible output voltage vectors
generated by the matrix converter. A minimization procedure,

(2)

The moment of inertia is given as:

J=

Using

1
m r22 + r12 = 4,2 kgm 2
2

r1=250mm,

r2=240mm,

(3)
m=70kg,

and

max = 2500 2 60 [rad s ] as the maximum angular speed of the

barrel, the maximum energy stored in the FES is:

Proceedings of the 2008 International Conference on Electrical Machines

Ek max =

1
2
J max
= 144[kJ ]
2

Paper ID 852

(4)

This energy cannot be completely extracted from the FES, as

the energy conversation becomes inefficient when the angular
speed drops below a certain value (roughly 50% of the rated
speed) [11]. Supposing this limit equals 50% of the maximum
speed, the amount of useful energy is 75% of the total energy
and, in this case, 0.75 144[kJ ] = 108[kJ ] . Given this energy,
Figure 2 relates the maximum power available from the
flywheel for a given time, for example, if we wish to have
power available during 0.5s, we can supply nearly 200kW.

Fig. 4. Dimension of the steel barrel.

B. Flywheel dynamics
The Flywheel dynamics can be given as:
d
T=J
+ K D + K C
dt

(5)

Where J is combined inertia of rotor PMSM and Flywheel, KD

is the friction coefficient, KC is the Coulomb friction and T is
the resultant torque. To obtain parameters KD and KC an
experimental deceleration (from 1500rpm to 0rpm) test was
made (Fig. 5).
From fig. 5, the flywheel parameters K D = 0.01Nms and
K C = 1.04 Nm can be obtained.

Fig. 2. Output power of the flywheel energy storage.

Fig. 5. Experimental results of flywheel deceleration.

III. MATRIX CONVERTERS

Matrix converters are based upon an association of nine bidirectional switches with turn-off capability, which allow the
connection of each one of the three output phases to any one of
the three input phases (Fig. 6). The 3-phase matrix converter
presents 27 switching combinations, or vectors [9].

Fig. 3. Flywheel energy storage prototype.

Proceedings of the 2008 International Conference on Electrical Machines

Paper ID 852

The AC-AC Matrix Converter, supplying the machine, was

using the Eupec ECONOMAC matrix module (Fig. 7). As an
example, Fig. 8 shows one of the IGBTs drivers.

Fig. 6. Matrix Converter topology.

Fig. 7. The Eupec ECONOMAC matrix module.

A nineelement matrix with elements Sij representing the

state of each bi-directional switch (if switch Sij is off then Sij=0,
else Sij=1), can be used to represent the output voltages (vA, vB,
vC) as functions of the input voltages (va, vb, vc):
v A S11
v = S
B 21
vC S 31

S12
S 22
S 32

S13 van
S 23 vbn
S 33 vcn

(6)
Fig. 8. View of the gate driver printed circuit board.

The input phase currents (ia, ib, ic) can be related to the output
phase currents (iA, iB, iC) by:
ia S11
i = S
b 12
ic S13

S 21
S 22
S 23

S 31 i A
S 32 iB
S 33 iC

(7)

The 27 switching combinations of the nine bi-directional

switches Sij, can be represented as voltage vectors, numbered
from 1 to 27 and given in table I as a function of each Sij state.
TABLE I: VOLTAGES VECTORS GENERATED BY THE MATRIX CONVERTER
N.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27

S11
1
0
0
1
0
0
1
0
0
0
0
1
0
1
0
0
1
0
0
1
0
0
1
0
1
0
0

S12
0
1
0
0
1
0
0
1
1
0
0
0
1
0
0
1
0
0
1
0
0
1
0
0
0
1
0

S13
0
0
1
0
0
1
0
0
0
1
1
0
0
0
1
0
0
1
0
0
1
0
0
1
0
0
1

S21
0
0
1
0
1
0
0
1
0
0
1
0
1
0
0
0
0
1
0
1
0
0
1
0
1
0
0

S22
1
0
0
0
0
1
1
0
0
1
0
0
0
1
1
0
0
0
1
0
0
1
0
0
0
1
0

S23
0
1
0
1
0
0
0
0
1
0
0
1
0
0
0
1
1
0
0
0
1
0
0
1
0
0
1

S31
0
1
0
0
0
1
0
1
0
0
1
0
0
1
0
0
1
0
1
0
0
0
0
1
1
0
0

S32
0
0
1
1
0
0
1
0
0
1
0
0
1
0
0
1
0
0
0
1
1
0
0
0
0
1
0

S33
1
0
0
0
1
0
0
0
1
0
0
1
0
0
1
0
0
1
0
0
0
1
1
0
0
0
1

vA(t)
va(t)
vb(t)
vc(t)
va(t)
vb(t)
vc(t)
va(t)
vb(t)
vb(t)
vc(t)
vc(t)
va(t)
vb(t)
va(t)
vc(t)
vb(t)
va(t)
vc(t)
vb(t)
va(t)
vc(t)
vb(t)
va(t)
vc(t)
va(t)
vb(t)
vc(t)

vB(t)
vb(t)
vc(t)
va(t)
vc(t)
va(t)
vb(t)
vb(t)
va(t)
vc(t)
vb(t)
va(t)
vc(t)
va(t)
vb(t)
vb(t)
vc(t)
vc(t)
va(t)
vb(t)
va(t)
vc(t)
vb(t)
va(t)
vc(t)
va(t)
vb(t)
vc(t)

vC(t)
vc(t)
va(t)
vb(t)
vb(t)
vc(t)
va(t)
vb(t)
va(t)
vc(t)
vb(t)
va(t)
vc(t)
vb(t)
va(t)
vc(t)
vb(t)
va(t)
vc(t)
va(t)
vb(t)
vb(t)
vc(t)
vc(t)
va(t)
va(t)
vb(t)
vc(t)

IV. PERMANENT MAGNET SYNCHRONOUS MOTOR MODEL

A. Machine Model
The mathematical model of a PMSM in the d-q frame can be
given by following differential equations [10]:
d
id r Lqiq
dt

(8)

d
iq + r Ld id + r f
dt

(9)

ud = Rid + Ld
uq = Riq + Lq

(10)
r = n ppm
Where ud and uq are the d and q axis voltages of the
synchronously rotating reference, R is the stator resistance , id
and iq are the stator currents, r is the electrical rotor frequency
in rad/s, Ld and Lq are the rotating field inductance. The rotor
mechanical angular speed is m and the machine pole-pair
number is represented by npp.
The electromagnetic torque is:

Te = n pp f iq + (Ld Lq )id iq

(11)

Where f is amplitude of the flux induced by the permanent

magnets of the rotor in the stator phases.

Proceedings of the 2008 International Conference on Electrical Machines

B. Control strategy
The control strategy is developed using the rotor field
orientation technique. The goal is to apply appropriate machine
currents. The adequate machine phase currents are enforced by
the matrix converter, to follow phase reference currents i*s1,
i*s2, i*s3. These phase reference currents are determined using
related dq reference current components i*sd and i*sq (as zero)
and Park transformation. The estimation of the dq reference
frame position r is obtained from (12), using the measured
rotor electrical angular speed r:
t

r (t ) = r + r dt

Paper ID 852

minimizes output current errors using a weighted cost

functional (17). The controller block diagram is depicted in
Figure 9.

ia

ib

c
v

S
11
"
S
33

RT
T
LT
isq (k + 1) = 1 s isq (k ) + s uq (k + 1) d s r (k )isd (k )

(14)
L
L
Lq
q
q

T
f s r (k )
Lq

B. Quadratic cost functional of the output current errors

Applying the Park transformation in dq coordinates, the errors
esd (15) and esq (16) are the errors between references and
output stator currents isd and isq respectively.
esd (k + 1) = isd (k ) isd (k + 1)

(15)

esq (k + 1) = isq (k ) isq (k + 1)

(16)

The cost functional must enhance larger errors, while accepting

small errors. Thus, a suitable cost evaluator is a quadratic cost
functional F (17), which computes the norm of the weighted
output currents errors.

i
s3

m (k )

r (k )

V. PREDICTIVE OPTIMAL CONTROL OF OUTPUT MATRIX

CURRENTS

Ts Lq
T R
T
r (k )isq (k ) (13)
isd (k + 1) = 1 s isd (k ) + s ud (k + 1) +
Ld
Ld
Ld

i
s2

S """" S
11
33

A. Discrete-time model for machine model

For the prediction of the dynamics of the PMSM, discretetime equations for sampling time Ts are used.
The equations for the stator current prediction are:

i
s1

(12)

F = esd2 ( k +1) kd + esq2 ( k +1) kq

ic

r (k )

e (k + 1)
sq

e (k + 1)
sd

i* (k )
sd

i (k )
sd

isq (k )

i (k + 1)
sd

isq (k + 1)

i = 0
sd

T * (k )

=
isq
n

T*
f

pp

i*
sq (k )

Fig. 9. Predictive current controller.

VI. SIMULATIONS
The matrix converter with input filter, PMSM and the real
time predictive optimal controller were simulated in the
Matlab/Simulink environment to evaluate the control
performance. A sampling period Ts=50s was used in the
equations to predict output currents.
A. Matrix Converter
The matrix converter has the following parameters
and
ra = rb = rc = 25 ,
la = lb = lc = l = 6.5mH
Cab = Cbc = Cca = 6.6 F . The three-phase system frequency is
i = 2f i with fi=50Hz and Vi=230V(rms).

B. PMSM parameters
The PMSM machine has the following parameters
R = 0.83 ,
Ld = 6.5mH ,
Lq = 6.5mH ,
n pp = 3 ,
J PMSM = 16.1 104 kgm 2 and torque constant KT = 1.39 Nm A .

(17)

In (17), kd , kq are the weighting constants.

C. Predictive control block diagram
The matrix predictive control method uses the discrete-time
model (13,14) to predict the future values of the input currents
and their errors (15,16) for all the 27 possible output voltage
vectors of the matrix converter. An optimal controller

C. PMSM currents
In this work a reference torque of 2.6Nm was used to test the
current predictive controller.
Figure 10, present the reference (i*s1, i*s2, i*s3) and simulated
stator currents (is1, is2, is3) without steady state or dynamic
errors (apart from the converter ripples). The speed is
1500rpm.

Proceedings of the 2008 International Conference on Electrical Machines

Paper ID 852
VIII.

Fig. 10. Reference (i*s1, i*s2, i*s3) and simulated stator currents (is1, is2, is3).

VII. DYNAMIC VOLTAGE RESTORER (DVR)

For DVR operation the matrix converter can be connected
between the PMSM motor and the 3-phase grid using series
connected star-delta transformers and lCr filters (Fig. 11, 12).
The DVR outputs the series voltages VTi, upon detection of
voltage sags, aiming to protect critical loads.

Fig. 11. DVR with series connected topology

Fig. 12. The electromechanical energy accumulator used for DVR operation.

CONCLUSIONS

This paper presented an experimental characterization of a

flywheel energy storage system, including a Permanent Magnet
Synchronous Machine controlled by an AC-AC Matrix
Converter. The matrix converter was controlled using a
predictive optimum method based on discrete-time model of
the converter and load.
Predictive optimal control is a powerful control technique, that
will open new possibilities in the control of permanent magnet
synchronous motors, and in the control of the matrix currents
since it allows the control designer to overcome the matrix
converter input/output coupling, easing the use of all the
available voltage vectors.
The next step in this investigation is to derive the prediction
equations for DVR operation and to add strong experimental
support.
REFERENCES
[1]

R. Hebner, J. Beno, A. Walls, Flywheel batteries come around again,

IEEE Spectrum, April 2002, pp. 46-51.
[2] T. Siostrzonek, S. Pirg, The Flywheel Energy Storage with Brushless
DC Motor - the Practical Results, EPE-PEMC 2006 September 2006,
Portoroz, Slovenia.
[3] B. Wang, G. Venkataramanan, Dynamic Voltage Restorer Utilizing a
Matrix Converter and Flywheel Energy Storage, Industry Applications
Conference, 2007. 42nd IAS Annual Meeting. Conference Record of the
2007 IEEE, Volume , Issue , 23-27 Sept. 2007 Page(s):208215.
[4] J. Silva, B. Borges,. Anunciada, Improving Control Strategies for HF
Resonant Link Converters: the Current Mode Predictive Modulator,
Power Electronics Specialists Conference, IEEE PESC91, ISSN 02759306, pp 268-274, USA, Junho 1991.
[5] J. Rodriguez, J. Pontt, C. Silva, P. Corts, and S. Rees, Predictive
Current Control of a Voltage Source Inverter, 35th Annual IEEE Power
Electronics Specialists Conference. Aachen, Germany, 2004.
[6] J. Barros, J. Silva, S. Pinto, Real Time Based Optimal Controller For
Four Multilevel Converters, Proc. IEEE CEE05 International
Conference on Electrical Engineering CEE05, CD-ROM paper 3w55,
Outubro 2005, Coimbra, Portugal.
[7] E. Wiechmann, A. Garcia, L. Salazar and J. Rodriguez, HighPerformance Direct-Frequency Converters Controlled by PredictiveCurrent Loop IEEE Transactions on Power Electronics, Vol. 12, No. 3,
May 1997.
[8] P. Gamba, S. Pinto, J. Silva, E. Margato, Predictive Optimal Control of
Unity-Power-Factor Matrix Converters used in Field Oriented Controlled
Induction Motor Drives, International Symposium on Industrial
Electronics, Vigo, Spain, 4-7 June 2007.
[9] L. Huber, D. Borojevic, N. Burany, Analysis, Design and
Implementation of the Space-Vector Modulator for Forced-Commutated
Cycloconverters IEE Proceedings-B, Vol. 139, no 2, March 1992, pp.
103-113.
[10] J. Palma, Accionamentos Electromecnicos de Velocidade Varivel
Fundao Calouste Gulbenkian, Portugal, 1999.
[11] M. Bollen, Understanding Power Quality Problems, WileyInterscience, ISBN 0-7803-4713-7.