Sie sind auf Seite 1von 7

Operating MicroGrid Energy Storage Control during

Network Faults
Nilanga Jayawarna
School of Electrical and Electronic Engineering
University of Manchester, Manchester, MI 7AB, UK

Catherine Jones
School of Electrical and Electronic Engineering
University of Manchester, Manchester, MI 7AB, UK
C.jon (mncbester.ac.uk

Mike Barnes
School of Electrical and Electronic Engineering
University of Manchester, Manchester, MI 7AB, UK

Nick Jenkins
School of Mechanical, Aerospace and Civil

N.Jayawarnagmancbester.ac.uk

M.Bamnes(~manchester.ac.uk

Abstract - A MicroGrid is expected to operate both as a


sub system connected to the main grid or as an islanded
system. However the provision of fault currents, in an
islanded MicroGrid consisting only of micro-generation
interfaced with relatively low-current power electronics, is
a serious system protection issue. This paper presents the
novel concept of using the central energy storage system
(flywheel) as the main fault current source in islanded
mode. The three-phase MicroGrid test rig used at
University of Manchester, and the flywheel control system
are described. The importance of accurate systems
modeling of the whole microgrid and energy storage unit is
shown. A fault study is carried out on the test rig and in
PSCAD. The flywheel inverter system is shown to
contribute enoughfault currentfor a sufficient duration to
cause the system protective device to clear the fault.

Keywords: MicroGrid Energy Storage Control Fault


currents,
DC link modeling

Introduction

Formation of MicroGrids is becoming an increasingly


attractive option because they have the potential to increase
the use of renewable generation and micro-CHP. A
MicroGrid consists of a cluster of micro sources, energy
storage systems (e.g. flywheel) and loads, operating as a
single controllable system providing both heat and power

to its local loads [1-2]. A MicroGrid may operate

connected to the main grid or it may disconnect from the


main grid, in the event of an upstream fault, and operate as
an islanded system. Energy storage is a key element in the
MicroGrid concept in order to facilitate these two modes of
operation. Apart from correct operation in these two steady
states, the energy storage control should help facilitate the
islanding and re-synchronizing of the MicroGrid to the
main grid. A unified control concept embracing all these
requirements is presented in [3]. In this paper we extend
the concept to include fault behaviour,

1-4244-1 160-2/07/$25 .OO 2007 IEEE.

Engineering
University of Manchester, Manchester, MI 7AB, UK
N

Jchester.ac.uk

In grid-connected mode, the main grid will provide


high fault currents to activate fault protection devices such
as fuses or circuit breakers (typically 3 to 6 times normal
rated current is required). However an islanded MicroGrid
is dominated by power electronic interfaces and these
inverters will have a maximum current rating of perhaps
one-and-a-half times normal rated current, unless extra cost
is added to the interfaces. Very limited research has been
published on MicroGrid protection schemes with alternate
fault detection methods to date [4-6]. However none of
these protection schemes have yet been successfully
demonstrated on an actual MicroGrid study system or a test
rig. This paper proposes to use the energy storage system
to supply fault currents in islanded mode. It has already
been established [7] that typical fault current in a
residential MicroGrid would be in the order of 450-560A
for a network feeder fault. Central energy storage units for
microgrids are typically rated in the range of 1O's of kW to
200kW. With careful design, the required fault current
would be within the capability of a large central energy
storage unit, without any additional expenditure on
MicroGrid hardware above and beyond what would be
required for basic MicroGrid operation. This also allows
the use of well established conventional overcurrent
protection. The practical implications of this concept were
investigated by conducting a fault study on a hardware test
rig representing a small-scale MicroGrid system.

The MicroGrid Test Rig


A medium power (2OkVA), three- phase test rig
a simple MicroGrid system had been

representing

constructed at the University of Manchester [3, 8]. Figure 1


shows a picture of the test rig in the laboratory and Figure
2 shows a simplified single line circuit diagram of the rig.
The 0.4kV/5OHz mains supply of the laboratory is
considered as the main grid and the MicroGrid feeder is
connected to the mains supply through contactors 1 and 2.

being run). A set of 12kW discharge resistors on the


DC link are used to dissipate excess energy and for quick
discharging of the DC link.

not

Central Energy Storage Control

3
-,ln

The /Wlcrogrld Tesit Peig

Figure 1. The laboratory test rig


1

Laboratory

Supply

The Mcrogrid Test Rig


2

Coupling

Inductance, X-CG
Microgridside
Inverter (MSI)

Drve

MGI12kW
M

12kW

Dump

ResistorS

Flywheel

passive
load

A MicroGrid may operate either connected to the


main grid or disconnected from it. There are two steady
states of operation, grid-connected and islanded. There are
two transient modes of operation, islanding and resynchronizing. A unified control capable of operating in all
these four modes has been developed for the MSI of the
central energy storage system [6-7]. This control concept is
based on local measurements only and is illustrated in
Figure 3. The MSI control during islanded mode is
encompassed in 'Islanded and Synchronization control'
block. The flywheel inverter system sets the voltage and
frequency of the islanded MicroGrid and maintains them
within acceptable limits by injecting or absorbing active
power and reactive power as required (based on active
power, P, vs. frequency and reactive power, Q, vs. voltage
droop lines). As soon as the presence of mains is detected,
Dthe MicroGrid control system uses feedback information
from the mains voltage to adjust the energy storage unit
voltage and frequency control loops to synchronize the
MicroGrid voltage with the mains voltage.

ContIoI
~~~~~~~~~~~~~~~~~~~~~~~~~Syichror saton

f4 Vcontdhuh
VariacMosrdMcgd

Onverge-et
gI

*~~~~~~~~~~~~~~~~~~~
a

Figure 2. Single line diagram of the test rig


A lOOkW/14MJ flywheel unit connected through a
100kW inverter represents the central energy storage
system. In the laboratory MicroGrid, the flywheel unit
power has been limited to 20kW in its software. The
flywheel system is connected to the MicroGrid through the
MicroGrid Side Inverter (MSI) and 9mH per phase
inductance formed by the coupling reactance and
transformer. A synchronous generator and an induction
motor coupled together (MGset) acts as the micro-source.
The induction motor drive operates in open loop constant
volts/hertz (V/Hz) mode. Many MicroGrids assume that all
micro-sources are interfaced with power electronics.
However this MGset was connected to the test rig to study
the interaction between a rotating machine generationsource with converter based generation, such as might be
found in some more conventional micro-generation. A

three-phase balanced load of 12kW was connected at the


end of the feeder.
The 22kW/600V rectifier acts as a supplementary
source to supply the DC link voltage (when the flywheel is

SW

zo

l9Rwnw1
j

larolS

vbniveZ
Med Mins

VqRe
s to

COCIVe
Mea D

voltageIS*Fn
s

Mea

red X

RSt

Figure 3. The Control of central energy storage system

Once the MicroGrid voltage is locked to the


mains voltage, contactor 2 can be closed to enter gridconnected mode. In this mode, the MSI control swaps over
from droop-line control to control of MSI real and reactive
pow er. Full details are given in [3].

(LPF) through which VdCis passed, Figure 6. This was


subsequentl assessed.
P [k]
20 ..............................

4 Flywheel Drive Control


The DC link for the MSI is supplied by the flywheel
interfaced through the flywheel drive (FWD). The FWD is
controlled by commercial software. This system is
illustrated in Figure 4.
P and Q from
MS] to Mlc gridl

Grid
Connected
conrol

Control
' ignal

DC Link F0

Voltag

Cont rol

Control
Signal

Figure 5. DC link voltage- real power profile of FWD

_
DC Link

Connection to the

Microgrid

microgrid-side

lnverter

490:

Fl w eel

Drive

DC Link FIywh~~~el

Voltage

Low
Low
PasLvl
Filter

Voltage

Le

Limiter

Voltae
Pro iler

Generation
Generato
of Control

Signals

~~~~~~~~~~~~~~~~~~~~~~~~~
flywheel
Drive

Figure 6. The flywheel drive feedback control path


Figure 4. The flywheel system and its control
The FWD maintains the DC link voltage at a set level, 5 Simulation and Experimental Studies
by charging or discharging the flywheel as necessary. The
A parameterized simulation model of the hardware rig
MSI uses a voltage source inverter. Two control structures had been developed in EMTDC/PSCAD to verify the
are available for such inverters: an amplitude modulation developed control prior to test-rig implementation. The
signal to create an ac voltage output which is a function of flywheel system was modeled with a controllable voltage
the dc link voltage, or a current command which uses a fast source with a power-voltage droop line and a low pass
feedback loop to apply the dc link voltage to the ac phases filter (see above). The DC link (including losses) was
and force the ac currents to follow some predefined included along with the MSI control. The synchronous
trajectory. There is clearly strong level of coupling generator had been modelled with its AVR and exciter
between the MSI and FWD control as a function of the dc based on manufacturer data sheets. The open loop V/Hz
link voltage (VdC). Measuring the dc link voltage to apply control of the induction motor drive was also set according
command feedforward and compensate for dc link voltage to data sheet values. Hardware propagation delay,
variation is sometimes used to decouple these control sampling, measuring and calculation delays were assumed
small enough the
to be ntrliler
negligibleSmal
when cpule
compared width
with
..control. ~to
thebebandwidthgo
loops. However sampling delays in the digital
systems limit the available response of such a system. m ut (Pw Msti.ofthe MSimwas not model
was
switching
frequency
nou aoteto
Observers would allow a zero-lag assessment of dc link because the
voltage in response to disturbances. However observers
require an accurate systems model. The approach followed interfere with the control.
was to attempt to develop an appropriate model to allow
The effect of coupling between the MSI and FWD
tuning of the FWD control loop to limit interaction control on the system was investigated on the test rig and
between the MSI and FWD. The model could then be used,
with
the simulation model by changing the response of the
so that an observer could be formulated if necessary.
FWD control. This was achieved by varying the time
Based on discussion with the flywheel manufacturers, constant of the LPF. The system was run in islanded mode
then synchronized and connected to the main grid.
it was theorized that the response of the flywheel unit in and
The
in the DC link voltage when
the MicroGrid
MirCrd
t
nh,
d
to
t
the,
grdcnetdwsosre
lretm
the MicroGrid system can principally be reducedl to two beamchange
.
Intal.
system elements, 1) the non-linear real power-voltage (P- cosatwsuetoflrotdcinnie.Fge7
o-asfle
V) droop line, which determines the amount of P to be sosteD ikvlaebhvorwt
absorbed or injected by the flywheel according to VdC, see time constant of 200ms. Clearly the response causes the
Figure 5, and 2) the time constant of the low pass filter
r

rig to trip out -the MSI and FWD loops


decoupled.

test

560 /-

are

not____
!558

here

Tv

552

~DC Link Voltage [V]

-------

----

LikVoltage [V

540

a--n-grid

__-__n ga r- d

here----- ---------_

h e re

568

6.2

6.4

535
53

6.

Tme [s]

7.2

7.4

------2--- 28Figure 10. DC link voltage response, T=OOlIs


(experiment)

------

Time [s]

Figure 7. DC link voltage response, T =0.2s (experiment)


558&0

The test was repeated twice, with the time constant


for the low pass filter on the flywheel drive control being
set to Ol s in the first case (Figure 8 and 9) and O.Ols in the
second case (Figure 10 and 11).

UVdcLink

DC LinkVoltaae

56
e

5520

5500

__DC Link Voltage [V]

556

54&0

--------

556 --

------

--

546.0

252I

tilfl

9360

KO

94140

94.80

95.20

Figure 1. DC link voltage response, T -0.OlIs (simulation)

~~~~~~~~~~~~Agreement

between the simulation results and the


hardware results is good but not perfect. The response of
simulation results is slower, by about 600o when
~~~the
54E 7 __ __ 7.41.-IL
76 78 8 L'82
6.4 66 8
7.2
T=0.Ols and about 3O0o when T=0.ls. In addition
the final
Time hsi
steady-state value varies between simulation and
The fundamental difference between
experiment.
Figue 8DClnk oltge rspose, =0 is(exprimnt)
simulation and experiment is that in its final steady-state
~~the model draws only 2kVA whereas the experimental
DC LibkVdWdd
draws 5kVA. This is an error of only 1.500 of the
DLnkVfaesystem
560.0 *9Vdctindk
name-plate rating of the energy storage unit, but at low
546

Figre

. olagerepone, =.
C
lnk

558.....................0....................................

556.0

552.0
550.0
54& 0

(epeimet)

ow ers

th is

differen ce

is

sign

ificant.

he

steady

-state

however depends not only on the FWD and MSI but also
the
MicroGrid response. To a very significant extent the
~~~' 554.0 ~~~~~~~~~difference
is due to limitations of the MG set model, which
apred very sensitive to the variation between nominal
typical control parameters. This will be discussed
~~~~~~~~~~and
further in section 7.

control to prevent unwanted MSI and FWD interaction and


for zero-lag command feed-forward (if necessary).

Fault Study

fault. Figure 13 shows the hardware results of the fault


current contributions from the flywheel inverter and the
MGset. Figure 14 shows the PSCAD simulation results of
the fault currents for this case.

In islanded mode, the flywheel unit is the main fault


current source in this system. Fault studies have been
carried out to determine the fault current distribution in an
islanded MicroGrid system [7].

Single-phase-to-earth faults are the most common


faults occurring in any power system. The earthing
arrangement in the laboratory test rig is a so-called TN-C-S
system. In a TN system, a single-phase-to-neutral fault is
equivalent to a single-phase-to-earth fault, and thus a
single-phase-to-neutral fault was used for this test-rig
study. A fault circuit was designed and connected parallel
to the 12kW load bank of the existing MicroGrid test rig,
Figure 12. A fault impedance of 14Q was selected. The
current flowing with the fault circuit switched on is
approximately twice the normal load current. Two per unit
fault current was considered adequate since the flywheel
system behavior and control are being studied. Larger fault
currents have the principle effect of requiring higher rated
devices but otherwise would not affect the flywheel system
control or its response.
Neutral

CB 2

CB 6

14W5

20

-201

5o

10

15

20

25

10

15

20

25

#1 :1

#1 :3

#1 :2

Figure 13. Current from the flywheel inverter (top graph)


and the MGset (bottom graph) - experiment (x-axes,
time/s, y-axes, current/A)

C45 2

Phase A

Phase B

-#1:1#1:3

14~5

Curret f brom the FVM1 anid the MGet~


'mIbFA1 inIcVW

UIaF~M

Phase C
30B5

124g'<8

Fault

creation
Flyvheel system

Micro-source

50

"IbMG

nIcMG

Figure 12. Test System for the unbalanced fault study


A 5 Amp fuse was selected as the protection device
for the fault circuit. The fault was created by connecting
the fault impedance (14Q) between phase C and the neutral
(by closing Contactor CB7). The Contactor CB2 was kept
open to ensure islanded operation.

7 Fault Study Results


7.1

Parallel operation of MGset and flywheel

In this study, the flywheel inverter and the MGset


were connected in parallel with the 12kW regular load. The

stator frequency of the induction motor drive was set such


that the MGset supplies the total regular load in normal
operation prior to the fault. The fault was created just
before 10 seconds by switching on the fault circuit and the
fuse blew approximately one second later, clearing the

p....ase.......an........neu....
2S.0 30.0 35.0 40.0 45.

4............

50.0

Figure 14. Current from the flywheel inverter (IxFWI) and


the MGset (IxMG) - simulation

In both the experiment and the simulation, the MGset


supplies most of the load current prior to the fault with the
flywheel inverter contributing only a small fraction. During
the fault both the sources supply fault currents, which blow
the 5A fuse within one second. In the experiment, the
MGset current, which is approximately 25A, peak, prefault, increases to about 30A, peak, during the fault. The
fault current contribution from the flywheel inverter is

-30|_~ ~ ain:Grphjs.

circa 15A, peak. However in the simulation results, the


flywheel inverter unit supplies most of the fault current
whereas the MGset contribution is negligible. This is
attributed to the limitations of the MGset model used in the
PSCAD simulation. This is illustrated more clearly when
we look at the power flows from the flywheel inverter and
the MGset during the fault. Figure 15 shows the hardware
results of the active power (P) and reactive power (Q) from
the two sources for the unbalanced fault while Figure 16
shows the software results.
4000

During this single-phase-unbalanced fault, the voltage


of the faulty phase drops slightly. The Automatic Voltage
Regulator (AVR) on the synchronous generator operates on
a very steep Q-V droop line, which results in large Q flows
even for a slight change in voltage as it tries to correct its
terminal voltage by drawing in reactive power. Hence the
MGset draws in a significant amount of reactive power
during the fault, which in turn is supplied by the flywheel
inverter in the experiment. However this MGset Q
behavior is not replicated in the simulation - the settings of
the experimental MGset being slightly different to their
nominal datasheet values.

With only the flywheel inverter

7.2

2000

#1:2

40

20

15000X

10

20

15

25

60

5000o

#1_ 1:3

-I2

10000

-5000

10

15

01

20

25

Figure 15. Active power and Reactive power from the


flywheel inverter (top graph) and the MGset (bottom

graph) - experiment (x-axes time/s, y-axes power/VA)

1.0
is

~~

300

35.0

40.0

12

#:

-10000

-20000

LI

Figure 17. Experimental results of current (top graph) and

g
~~~~~~~~~~~~~~~~~~h
Qe01
O.h
EiFVA
IbF.
IcFV
60 .............

a,

O S I S ~~~~~~~~~~~~~~~PFVVP1.

ISOOk "O

power/VA)

rnet1
PMQmoe Pr

12 5k-III

25.0

Q~~~_JF___VA

P and Q (bottom graph) from the flywheel inverter (xaxes, time/s, y-axes: top - current/A, bottom graph

Real and R6eidiV6 Pbower Flbow feron the FM ahd the MG Se


MP
FVMW

-1

20000

o#0

time[01

-20-..

- 40

I 0000 si;

I~

1:2

45.0

*O>FWAr)II

50.0

Figure 16. Simulation results of P and Q from the flywheel


inverter (top graph) and the MGset (bottom graph)

10.0
4

12.0

14.0

16.0

16.0

20.0

Figure 18. Simulation results of current (top graph) and P


and Q (bottom graph) from the flywheel inverter
To confirm this, a test was carried out where the MGset
was not connected and the flywheel inverter only supplies
a wholly passive load. This is the worst case scenario in an
islanded MicroGrid since only the flywheel inverter unit is
available to supply fault currents to operate the protection
devices. Figure 17 shows the hardware traces of current
and power from the flywheel inverter. These waveforms
were captured 2 seconds before the occurrence of the fault
(-2 seconds). The fault occurs at zero seconds and it was
cleared after approximately 0.85 seconds. The software
results are shown in Figure 18 and they are in excellent
agreement with hardware results. These results illustrate
that the flywheel inverter system and its control system are
capable of reacting nearly instantaneously to a fault and
supplying the required total fault current.

The results also indicate that significant discrepancies


can arise in model of the energy storage unit behavior if the
model of any other significant element is not extremely
accurate. Nominal datasheet values were initially used for
the model of the MG set (which was about one year old).
Similar responses were achieved in simulation and
experiment, if off-nominal, modified controller settings
were used. Clearly factory settings can drift with time.
However the energy storage unit must be designed to
ensure appropriate MicroGrid behavior at all times. Given
that the behavior of the micro-sources and loads may
change over time and as the MicroGrid expands or changes
topology, achieving adequate MSI behavior for all cases is
thus made extremely challenging.
8

Conclusions

The MicroGrid test rig and the control of the central


energy storage system were discussed. It was shown that
the DC link cannot be assumed to be constant. Good
system models are necessary to achieve control decoupling.
An unbalanced fault study was carried out on the test rig at
the University of Manchester. The flywheel inverter
control demonstrated that it is capable of providing desired
amounts of fault currents, either connected in parallel with
another source or on its own.
Precise models of system behavior are required to
achieve adequate system tuning. However precise
knowledge of system parameters is required. Variations in
standard component "factory" settings from nominal can be
substantial and may drift with time. There is still some way
to go to make the MSI system robust to such parameter
variations.

Acknowledgements

The authors wish to acknowledge EPSRC for funding


this work under project "UK-Microgrids", EP/C00177X/1.
The authors also wish to thank Dr Mike Richardson
formerly of UPT, for his help.
References

[1] R. Lassetter, A. Akhil, C. Marnay, et al, "The CERTS


MicroGrid Concept", White paper on Integration of
Distributed Energy Resources, CERTS, April 2002.
[2] European Research Project MicroGrids [Online].
Available:http://microgrids.power.ece.ntua.gr/

[3] C. Jones, C. Fitzer, M. Barnes, "Unified Control and


Operation of a 2OkVA Laboratory MicroGrid
incorporating Flywheel Energy Storage", submitted for
publication in the International Journal of Distributed
Energy Resources.
[4] H. Al-Nasseri, M. A. Redfern, R. O'Gorman,
"Protecting MicroGrid systems containing solid-state
converter generation", Proceedings of International
Conference on Future Power Systems, pp 1-5, Nov 2005

[5] J. L. Pereira da Silva, C. Moreira, "Evaluation of the


discriminating capability of zero sequence protection",
Internal report for EU MicroGrids project, INESC Porto,
February 2004.
[6] "MicroGrid protection and control - summary",
Accessed at http://eppe.tamuedu/pserc/summaa/gridpdf
[7]

N.

Jayawarna, N. Jenkins, M. Barnes, M. Lorentzou,

S. Papathanassiou, N. Hatziagyriou, "Safety Analysis of a


MicroGrid", International Journal of Distributed Energy
Resources, Volume 2, Number 4, pages 261 - 278, October
- December 2006
[8] C. Jones, C. Fitzer, M. Barnes, "Investigation of
MicroGrids ", Proceedings of International Conference on
Power Electronics, Machines and Drives, pp 510-514,
April 2006

Das könnte Ihnen auch gefallen