Simulation and Analysis of High-speed Modular Flywheel Energy Storage
Systems Using MATLAB/Simulink
Parag Upadhyay, Member IEEE and Ned Mohan, Fellow IEEE
upadh008@umn.edu
Department of Electrical Engineering, University of Minnesota,
Minneapolis, USA 55455
May 30, 2009
Keywords: Storage system, Flywheel energy storage
system, High-speed drives, PM motor
Abstract:
Storage is an extremely important area of research and
has several applications, including potential of furthering
the integration of renewable in the grid. An efficient and
cost-effective electric storage is a transformative
technology and benefits the environment and decreases the
reliance on conventional energy sources. Flywheel energy
storage systems, unlike chemical batteries of around 75%
efficiency, have the potential of much higher cycle-life and
round-trip efficiency (RTE), without recycling battery
chemicals at life-end. Determination of RTE of a storage
system requires multidiscipline system modeling and
simulations. The modeling and simulation presented in this
paper determines the RTE of the flywheel storage system.
The losses in the converter, magnetic bearings, and the
machine losses (copper and iron losses) are considered for
calculation of RTE. The loss variance due to power and
speed change during the charging and discharging cycles
are incorporated in the model.
1. INTRODUCTION
Lots of research efforts are needed at addressing the
challenges to greater use of renewable energy. The major
challenge to using them as a source of power is that they
are intermittent and they do not always blow when
electricity is needed. Also, not all amount of source can be
harnessed to meet the timing of electricity demands. The
batteries are not the right solution as storage because of;
lower round-trip efficiency, chemically hazardous in
nature, larger volume, weight and cost (about 25
cents/kWh) [1]. The higher storage cost discourages user
or utility to employ renewables. Energy storage systems
have not been given equal importance as generation,
transmission and distribution technologies. Flywheel
energy storage system (FESS) is environment friendly and
can be a best fit solution for renewables storage by
addressing the challenges of; (a) making it cost effective
and (b) improving the round-trip efficiency (RTE) up to
90 %.
A modular FESS located at load end not only helps in
storing renewables but also enhance transmission and
distribution efficiency by load equalization. Determination
of RTE of a storage system requires multidiscipline
system modeling and simulations. The modeling and
simulation presented in this paper determines the RTE of
the modular FESS. The losses in the converter, magnetic
mohan@umn.edu
bearings, and the machine losses (copper and iron losses)
are considered for calculation of RTE.
Figure 1. Flywheel Energy Storage System Layout
2. FLYWHEEL ENERGY STORAGE SYSTEM
The layout of 10 kWh, 36 krpm FESS is shown in
Fig(1). A 2.5kW, 24 krpm, Surface Mounted Permanent
Magnet Motor is suitable for 10kWh storage having
efficiency of 97.7 percent. The speed drop from 36 to 24
krpm is considered for an energy cycle of 10kWh, which
is advantageous to the power electronics converter and the
motor drive designs. It is considered that the composite-
fiber flywheel, magnetic bearings and PM machine
(motor/generator) are integrated into a vacuum chamber.
Only the solid conductor leads go through the vacuum
chamber. The power electronic controller plays a vital role
in this system. When the renewable source generates the
power, controller sets power flow from renewable to the
motor and the motor accelerates the flywheel. The speed
is increased from 24 krpm. The controller sense the speed
of the flywheel and it stops the power flow when flywheel
reaches its maximum speed of 36 krpm. When plug-In
hybrid is connected for the battery charging, controller
revert power flow from PM generator to the Plug-In
Hybrid. This time, it senses the speed and stops supplying
power when speed touches predefined speed which is
higher than 24 krpm. During the interval between the two
processes, the system observes free-running. In free-
running flywheel speed drops due to the mechanical losses
and iron losses. As the motor is operated in vacuum, rotor
losses can be dissipated only through radiation hence it
increases the temperature of the rotor. Efforts are being
made to reduce the rotor temperature-rise by reducing the
rotor losses so it is essential to incorporate a sub-system
which can determine the rotor temperature-rise during the
process. The complete FESS simulation model is the
integration of different subsystems essential for the
analysis as discussed in following sections.
74
2.1. Generation Through Renewables on time and speed. The model of this sub-system is shown
The renewable energy sources are variable in nature in Fig(4). A threshold speed is also set for PV and PHEV.
hence a variable generating source is considered in this
simulation model. For the illustration, solar PV generation
with clear sky condition is considered. The solar PV
generation is proportional to the sine of solar angle with
the earth considering the clear sky [1].The time duration
considered is 8 hours (i.e. the renewable charges the
flywheel from 24 krpm to 36 krpm in 8 hours). The
charging process repeated after every 24 hours, hence in
the model shown in Fig(2), two sources with different
time duration are considered. PV generation is not speed
dependent hence the speed feedback is not given to the
blocks. In case if the generation is depending on the
flywheel speed, speed feedback can be given to this
subsystem.

Figure 3. Simulation Model for the Plug-In Hybrid

Battery charging

Figure 2. Simulation Model for the Renewable Source

2.2. Plug-In Hybrid Battery Charging

Once the flywheel is charged up to 36 krpm, it
releases the energy to charge a plug-in hybrid EV in the
night. The charging rate can be decided by the controller.
Constant current charging is efficient as well as faster. To
set with the constant current charging, constant torque
operation is desirable for the motor. The average
consumption of the plug-in hybrids is 10 kWh and battery
charging process usually takes 4-5 hours [2]. For the
simulation of this paper, PM motor charges the battery in
4 hour at constant torque. The variable power for the
battery charging is developed by the speed feedback and
pre-set torque value as shown in Fig(3).
2.3. Controller
The controller is the brain of the FESS made up with
power electronic devices, memory, gate-drives, sensors
etc. The role of controller for this simulation is to control
the power flow in the system. The sub-system receives
signals from all sub-systems and set the power flow based
Figure 4. Simulation Sub-system for the Controller
2.4. Permanent Magnet Motor
A Permanent magnet Synchronous machine is
designed and analyzed using FE analysis. The results of
analysis are used for modeling the PM motor sub-system
as shown in Fig(5). The motor dynamics are related to the
power flow from controller to the flywheel or the other
way. The sub-system calculates the losses in the PM
machine and the rotating system. As losses are frequency,
speed and load dependent speed and power inputs are
given to sub-system to calculate the losses. The PM
Synchronous machine is rated as 3-phase 2.5 kW, 36000
rpm, 230 V. Typical values of losses for the designed
motor are given in table 1. The iron losses are calculated
based on FE analysis, Copper losses are determined using

75
computer aided design programs, and mechanical losses used, this loss can be reduced. Composite materials, such
are considered for magnetic bearings [3]. The windage as carbon fiber, have the desired properties and result in a
losses are negligible as the system operates in vacuum. lightweight wheel that operate at high rotation speed or
have a large radius. High speed flywheels offer distinct
Table 1. Losses of designed PM motor advantages, such as high power density, compact design,
Loss (watts) 24 krpm 36 krpm very high energy efficiency, very low heat output,
Cupper losses at rated load 35.80 16.50 extremely high uptime availability and exceptionally
Core losses 21.62 45.21 infrequent low-cost maintenance. It is common practice to
Mechanical losses 06.01 09.02 wound the carbon fiber rings on a solid steel structure to
Total Losses 63.43 70.73 have maximum energy transfer from shaft to the wheel
tip. The surface speed for the carbon fiber material is
limited to 1000 m/s where as for the steel it is 250 m/s [4]
considering the tensile stress of both the materials. The
density of the carbon fiber material is 1800 kg/m3 [4].

2.6. Simulation of Losses

During this process, switching losses occur in the
controller, iron and copper losses occur in the motor and
friction and windage losses occur in the rotating parts. As
the power and speed terms are not constant, these losses
are variable and are function of the speed and the power.

40
Hysteresis Loss
35 Eddy-Current Loss
Total Iron Losses
Figure 5. Simulation Sub-system for PM Motor/Generator 30
Losses (Watts)

25
2.5. Simulation of Flywheel
20
The wheel is the energy storage component of the
FESS. The energy stored in a basic cylindrical wheel is 15

determined by the following relationship; 10

5
1 2
E= Iω (1) 0
2 300 350 400 450 500 550 600
Frequency (Hz)
2
Inertia I = mr for thin wall cylindrical shape
Figure 7. Iron Loss variations obtained using finite
1 element method
= m(r12 + r22 ) for thick wall cylinder (2)
2
where, m is mass, r is radius of thin wall cylinder, ω is
the rotation speed, ro and ri are the outer and inner radius
of the thick wall cylinder respectively.

Figure 6. Simulation Sub-system for Flywheel

Figure 8. Simulation Sub-system for the Losses
Due to the dependence of energy storage on both
mass and speed, it is inevitable that high bearing loss will The mechanical losses and iron losses are function of
occur. If materials with high strength to weight ratios are speed/frequency therefore the model needs the speed input

76
to determine these losses. The copper losses decrease and
iron losses increase with the speed, changing from 24000
to 36000 rpm. An iron loss characteristic obtained using
finite element method is shown in Fig(7).
This variation can be easily modeled in
MATLAB/Simulink as a sub-system shown in Fig(8) to
calculate losses at any instantaneous speed. The copper
losses are function of current therefore the power and
speed are the inputs to the subsystem.
2.7. Simulation of Temperature-Rise
3. ANALYSIS OF STORAGE SYSTEM
The flywheel energy storage system shown in Fig(1)
can be simulated by a Simulink model shown in Fig(10).
The simulation model deals with various aspects the
system: power flow, electromechanical conversion,
dynamics of flywheel, and temperature-rise of the rotor.
The subsystems described in section II are integrated to
analyze the FESS. As discuss earlier that the renewable
source of energy is variable with time. A complete FESS
cycle has four components namely; (a)charging,
(b)freewheeling at 36 krpm, (c)discharging and
(d)freewheeling at 24 krpm. As the losses are variable
with speed and load conditions, the RTE depends on the
duration of all above components.

Figure 9. Simulation Sub-system for temperature-rise

The heat generated in the rotor of the motor/generator

will dissipate only through radiation in vacuum. The
Energy radiated per second is given by Stefan-Boltzmann Figure 10. Simulink Model for Flywheel Storage System
of radiation as;
H = εsAT4 (3)
where, ε is emissivity (0-1), s is Stefan-Boltzmann
constant (5.67 x 10-8 J/s-m2-K4), A is the surface area of
object and T is Kelvin temperature. The relation of the
established heat and the temperature-rise can be given as;
Rate of Heat developed = (rate of temperature-rise of the
rotor body) + (Heat dissipated through radiation) (4)
which gives the power relationships as;
Protor loss = mc∆T/dt + εsA(Tf4 –T04) (5)
where, Protor loss is the rotor losses, m is mass of the rotor
body, c is specific heat coefficient of the rotor material,
∆T/dt is the rate of temperature-rise, Tf is the final steady Figure 11. Power flow curve of Flywheel Storage system
temperature of the rotor and T0 is the ambient temperature.
The simulation model for the temperature-rise is shown in In this simulation the discharge period is extended in
Fig(9). such a way that the FESS speed at the end of FESS cycle

77
equals the initial speed. The FESS is charged through PV
system with the variable power as shown in Fig(11).
The power developed by the PV cell is proportional
to the solar angle with PV array. The duration considered
in this case is 8 hours as can be seen in Fig(11). The
flywheel accelerates from 24 krpm to 36 krpm before 8
hours. The controller restricts the flywheel speed between
24 krpm to 36 krpm hence at 36 krpm it stops the power
flow.
Figure 12. Speed-Time curve for the Flywheel storage
system
Owing to retardation of the wheel, power still flows
to maintain the speed at 36 krpm as seen in later part of
charging. For the duration of freewheeling as indicated in
fig(11), flywheel runs at 36 krpm and retards due to no
load losses as shown in Fig(12). The duration of
freewheeling at 36 krpm can be set by user and the time
when flywheel starts discharging is decided. In this case
flywheel charges batteries at constant power mode.
10kWh energy is transferred in four hours. The controller
tracks the speed of the flywheel and sets the time when
flywheel starts freewheeling at 24 krpm such that the
speed after the complete day cycle is equals the initial
speed.
Figure 13. Total Losses variations of Flywheel storage
system
78
The total loss during the energy cycle is variable and
can be seen from Fig(13). The RTE determined from the
simulation is 86.08%. The Steady state rotor temperature
of 44°C is obtained when heat dissipation equals the heat
developed. The simulation is run for two complete FESS
cycles and the speed at the end of each cycle is observed.
Two different charging and discharging conditions are
considered as shown in fig(14). For the first day Power
generated through PV is lesser than the second day. Also,
flywheel discharges at constant torque and constant power
for the first day and the second day respectively shown in
Fig(14). The RTE for day-1 is higher than that of RTE of
day-2 because, the discharge time is very less on day-1,
hence the speed suddenly drops from 36 to 24 krpm. The
energy lost due to copper losses remains same but iron
losses for the energy cycle reduces and the RTE increases.
It can be observed from Fig (15) that, the speed of the
flywheel at the end of each day approaches the initial
speed. It can also be observed that the speed of the
flywheel is equals the initial speed irrespective of the
power flow conditions.
Figure 14. Power flow curve of Flywheel Storage system
for the two consecutive energy cycles
Figure 15. Speed-Time curve for the Flywheel storage
system for the two consecutive energy cycles
Table 2. Round-trip efficiency of FESS for the different the discharging the flywheel results in lower round-trip
duration of freewheeling at 36 krpm efficiency. There is a design trade-off for the percentage
Duration of Duration of RTE RTE of iron and copper losses for achieving higher round-trip
freewheeling Discharge for for efficiency of the storage system. The duration of charging
at 36 krpm Motor FESS and discharging also plays a role in determining the
(Hrs) Only round-trip efficiency. With typical practical loss values,
1 4 Hrs 2.6 min 87.31 85.15 the round-trip efficiency can be improved to more than
2 4 Hrs 2.1 min 87.10 84.78 90% which is much higher than battery storage system.
3 4 Hrs 1.6 min 86.89 84.41
4 4 Hrs 1.0 min 86.68 84.04
5 4 Hrs 0.5 min 86.47 83.67
6 4 Hrs 0.0 min 86.25 83.31 REFERENCES
[1]. U.S. Department of Energy, Energy Efficiency and
Simulation is run for different time duration when the Renewable Energy, www.eere.energy.gov
flywheel is discharged and results are given in table (2). [2]. Michael Kintner-Meyer, Kevin Schneider, Robert
The peak power input from PV source is considered as Pratt “Impacts Assessment of Plug-In Hybrid
2300 watts and the power output to discharge the flywheel Vehicles on Electric Utilities and Regional U.S.
to battery is considered as 2500 watts constant. A minor Power Grids Part 1: Technical Analysis”, Pacific
variation in the duration of discharge is observed and it is Northwest National Laboratory(A), November, 2007
decreased with increase in the duration of freewheeling at [3]. G. Schweitzer, Active magnetic bearings – chances
36 krpm. RTE can be improved to more that 1 percent by and limitations, in: Proceedings of the 6th
decreasing the freewheeling duration at 36 krpm. The International. IFToMM Conference on Rotor
flywheel for 10kWh storage is considered as 550 mm long Dynamics, September 30–October 3, 2002, Sydney,
having outer diameter of 250 mm. The wheel along with Australia.
motor is suspended through two radial and an axial [4]. Bitterly Jack. J. Flywheel Technology Past, Present,
magnetic bearings. Considering the estimated loss of the and 21St Century Projections, IEEE AES Systems
bearings [3], RTE can be determined for the entire FESS Magazine, May 1998.
as shown in tables 2 and 3. The RTE for the storage
system is 85.15 percent for the freewheeling duration of BIOGRAPHY
an hour. Analysis is also carried out to arrive at the Parag Upadhyay was born in India and received a
copper losses and iron losses of the machine to achieve PhD. in Electrical Engineering from Indian Institute of
the RTE to be more than 90 percent. It is also observed Technology Delhi, New Delhi. He is currently
that the achievable RTE for the entire system is 87.7 Postdoctoral Associate in the Department of Electrical
percent which is much higher than the battery storage Engineering at the University of Minnesota. His research
systems. interests include design and finite element analysis of
permanent magnet machines, electromechanical devices,
Table 3. Round-trip efficiency of FESS for the different energy systems and motor drives. He is Member, IEEE
values of per unit losses Ned Mohan is the Oscar A. Schott Professor of Power
Case Per unit Per RTE RTE Electronics at the University of Minnesota and has
Copper unit for for numerous patents and publications in this field. He is
losses Iron Motor FESS IEEE Fellow.
losses Only
A 1.0 1.0 87.10 84.78
B 0.5 1.0 87.67 85.46
C 1.0 0.5 89.23 86.73
D 0.5 0.5 89.94 87.41
E 0.6 0.4 90.26 87.70

It can be observed that reduction in copper loss will

not significantly affect the RTE. It is advisable from the
results of Table (3) that, the 40% reduction in copper loss
and 60% reduction in iron loss gives more than 90 percent
RTE. The temperature throughout the simulation is well
within 40 °C.

CONCLUSIONS
The iron loss significantly affects the round-trip
efficiency as compared to the copper losses. The delay in

79