1
Abstract--Under normal operating conditions, a MicroGrid
(MG) is interconnected with the Medium Voltage (MV) network.
However, planned or unplanned events like maintenance or
faults in the MV network, respectively, may lead to MG
islanding. In order to deal with islanded operation and even
black start following a general blackout, an emergency operation
mode must be envisaged. Two possible control strategies were
investigated and are described in this paper in order to operate a
MG under emergency mode. A sequence of actions for a well
succeeded black start procedure, involving microgeneration
units, has also been identified contributing for an increase in
distribution network reliability.
Index Terms-- Dynamic response, energy storage, frequency
control, microgrid, dynamic stability, power system restoration.
I. INTRODUCTION
ONNECTING to Low Voltage (LV) networks small
generation units – the microsources (MS) – with power
ratings less than a few tens of kilowatts may increase
reliability to final consumers and bring additional benefits for
global system operation and planning. In this context, a MG
can be defined as an LV network, plus its loads and several
small modular generation systems connected to it, providing
both power and heat to local loads (Combined Heat and Power
– CHP) [1]. MG flexibility can be achieved by allowing its
operation under two different conditions:
• Normal Interconnected Mode – the MG is connected to a
main Medium Voltage (MV) grid being either partially
supplied from it or injecting some amount of power into
it;
• Emergency Mode – the MG operates autonomously (as
in physical islands) when the disconnection from the
upstream MV network occurs.
The successful design and operation of a MG requires
solving a number of demanding technical and non-technical
issues, in particular related to system functions and controls
[2-3]. The presence of power electronic interfaces in fuel cells,
photovoltaic panels, microturbines or storage devices
This work was supported by the European Commission (EU) within the
framework of EU Project MicroGrids, Contract No. ENK-CT-2002-00610.
J. A. Peças Lopes, C. L. Moreira, A. G. Madureira and F. O. Resende are
with INESC Porto – Instituto de Engenharia de Sistemas e Computadores do
Porto, Porto, Portugal (e-mail: jpl@fe.up.pt).
X. Wu, N. Jayawarna, Y. Zhang and N. Jenkins are with the University of
Manchester, Manchester, United Kingdom (e-mail: n.jenkins@umist.ac.uk).
F. Kanellos and N. Hatziargyriou are with NTUA – National Technical
University of Athens, Athens, Greece (email: nh@power.ece.ntua.gr).
characterizes a new type of power system when compared
with conventional systems using synchronous generators. The
dynamic behaviour of a system with low global inertia,
comprising some MS with slow responses to control systems,
is also quite different from traditional power systems.
Furthermore, classic power systems have the possibility of
storing energy on the rotating masses of synchronous
generators, which provides energy balance in the moments
subsequent to a load connection. A MG requires some form of
energy storage (batteries or flywheels) in order to be able to
face transients during islanded operation [2], [4].
The MG islanding process may result from an intentional
disconnection from the MV grid (due to maintenance needs)
or from a forced disconnection (due to a fault in the MV
network)
. MG dynamic behaviour in islanded operation needs
to be analysed under different load conditions. Since there is
little inertia in a MG, load-shedding strategies and storage
devices must be used for an efficient frequency control
scheme [2].
Exploiting MG capabilities to provide fast Black Start (BS)
at the LV level is an innovative aspect that has been developed
and tested in this research [3]. Such an approach will enable
fast restoration times to final consumers, thus improving
reliability. In large conventional systems, tasks related to
power restoration are usually carried out manually by system
operators, according to predefined guidelines. These tasks
must be completed speedily, in real-time basis and under
extreme stressed conditions. In an MG, the whole restoration
procedure is much simpler because of the small number of
control variables (loads, switches and MS). However, specific
characteristics of most MS (such as primary energy source
response time constants, intermittency, technical limits) and
control characteristics of power electronic interfaces require
the identification of specialized restoration sequences [3].
II. M
ICROGRID ARCHITECTURE
The MG architecture developed within the EU R&D
Microgrids project [1] is presented in Fig. 1; it comprises a LV
network, loads (some of them interruptible), both controllable
and non-controllable MS, storage devices and a hierarchical-
type management and control scheme supported by a
communication infrastructure used to monitor and control MS
and loads.
The MG is controlled and managed by a MicroGrid Central
Controller (MGCC) installed at the MV/LV substation. The
MGCC includes several key functions (such as economic
Control Strategies for MicroGrids Emergency
Operation
J. A. Peças Lopes, Senior Member, IEEE, C. L. Moreira, A. G. Madureira, F. O. Resende, X. Wu, N.
Jayawarna, Y. Zhang, N. Jenkins, Senior Member, IEEE, F. Kanellos and N. Hatziargyriou, Senior
Member, IEEE
C
2
managing functions and control functionalities) and heads the
hierarchical control system. At a second hierarchical control
level, controllers located at loads or at groups of loads (Load
Controllers – LC) and controllers located at MS (Microsource
Controllers – MC) exchange information with the MGCC and
control local devices. LC serve as interfaces to control loads
through the application of an interruptibility concept, and MC
control microgeneration units, for example in terms of active
and reactive power production levels.
Fig. 1. MG architecture, comprising MS, loads and control devices
III. MICROGRID CONTROL FOR ISLANDED OPERATION
In this section different approaches to deal with MG
islanded operation are described. In the first approach, the
main concern is related to inverter control modes. As the MG
is an inverter dominated grid, frequency and voltage control
during islanded operation is performed through inverters. In
this case, the main issue is how to get a voltage and frequency
reference in the islanded MG. The other approach closely
follows concepts related to conventional synchronous machine
control.
A. MicroGrid Operation Regarding Inverters Control Modes
The approach focused on inverter control modes required
the modelling of MS and storage devices, as well as inverters.
MG loads are modelled as a combination of impedance type
and induction motor type loads. Load shedding mechanisms
based on MG frequency deviation were also considered to be
implemented in the LC.
1) Microsource and Storage Devices Modelling
Several MS models have been developed including fuel-
cells, microturbines, wind generators and photovoltaic arrays
[5]. A Solid Oxide Fuel-Cell (SOFC) was used in this
research. Its model includes a Fuel Processor, which converts
fuels like natural gas to hydrogen, a Power Section, where
chemical reactions take place, and a Power Conditioner that
converts DC to AC power. More details about the used
dynamic model of the SOFC can be found in [6] and [7].
The GAST dynamic model [6] was adopted for the primary
unit of microturbines, since they are small simple-cycle gas
turbines. Both high-speed single-shaft units (with a
synchronous machine) and split-shaft units (using a power
turbine rotating at 3000 rpm and a conventional induction
generator connected via a gearbox) were modelled. The
single-shaft unit requires an AC/DC/AC converter for grid
connection. The wind generator is considered an induction
machine directly connected to the network. Concerning the
PV generator, it was assumed that the array is always working
at its maximum power level for a given temperature and
irradiance. Basically, it is an empirical model based on
experimental results as described in [5], where a detailed
description on MS modelling adopted in the Microgrids
project can also be found.
Considering the time period under analysis, storage
devices, such as flywheels and batteries, are modelled as
constant DC voltage sources using power electronic interfaces
to be coupled with the electrical network (AC/DC/AC
converters for flywheels and DC/AC inverters for batteries).
2) Inverter Modelling
Two kinds of control strategies may be used to operate an
inverter [8]. The inverter model is derived according to the
control strategy followed:
• PQ inverter control: the inverter is used to supply a given
active and reactive power set-point;
• Voltage Source Inverter control logic: the inverter is
controlled to “feed” the load with pre-defined values for
voltage and frequency. Depending on the load, the
Voltage Source Inverter (VSI) real and reactive power
output is defined.
The PQ inverter injects the power available at its input into
the grid. The reactive power injected corresponds to a pre-
specified value, defined locally (using a local control loop) or
centrally from the MGCC.
dc
dc ref
act
react
act
react
react
grid ref
ref
Fig. 2. PQ inverter control system
The control principle of a PQ controlled inverter is shown
in Fig. 2. It is based on the computation of the normalized
active and reactive current components [2] that are used to
control active and reactive output powers of the inverter,
respectively.
The Voltage Source Inverter (VSI) emulates the behaviour
of a synchronous machine, thus controlling voltage and
frequency on the AC system by using droop concepts as in
Fig. 3. Frequency variation in the MG provides an adequate
way to define power sharing among VSI [9-10]. A three-phase
balanced model of a VSI including the droop concepts was
derived from a single-phase version presented in [10]. More
details on VSI operation can be found in [2-3].
Another important issue is the behaviour of power
electronic interfaces during short circuits [2]. Special control
3
functions and sufficient oversizing is required because, in
contrast to synchronous generators, power electronics have no
thermal short-term overloading capabilities. The current
limiting function is easily implemented in the PQ controlled
inverters by limiting the total gain of the PI controllers shown
in Fig. 2 In order to limit the output current of a VSI, a control
technique like the one presented in Fig. 2 is also used. The
main difference is that in this case the reference current has a
maximum peak value dependent on switching devices
characteristics and its frequency is imposed by the inverter
frequency/active power droop.
Fig. 3. Frequency / active power droop characteristic
3) Control Schemes for MicroGrid Islanded Operation
If a cluster of MS is operated within a MG and the main
power supply (the MV network) is available, all the inverters
can be operated in PQ mode, because there are voltage and
frequency references. However, a VSI can be used to provide
a reference for frequency and it will be possible to operate the
MG in islanded mode and to smoothly move to islanded
operation without changing the control mode of any inverter
[2-3]. After identifying the key solution for MG islanded
operation, two main control strategies are possible:
• Single Master Operation (SMO): A VSI is used as
voltage reference when the main power supply is lost (in
order to balance local load and generation); all other
inverters can then be operated in the PQ mode;
• Multi Master Operation (MMO): More than one inverter
is operated as a VSI. However, other PQ controlled
inverters may coexist.
During islanded operation, the power injected by the
storage devices is proportional to MG frequency deviation.
Therefore, correcting permanent frequency deviation during
islanded operation should be considered a key objective in any
control strategy in order to avoid storage devices to keep
injecting (or absorbing) active power whenever MG frequency
deviation differs from zero [2-3]. The combination of primary
frequency regulation provided by storage devices, load
shedding schemes for less important loads and secondary load
frequency control are the key for successful MG islanded
operation.
B. MicroGrid Operation Regarding Primary Energy Source
Control
In this representation, the MS and storage device (flywheel)
can be represented by synchronous generators or by
STATCOM Battery Energy Storage (STATCOM-BES). In
grid-connected mode, the frequency of the MG is maintained
within a tight range. However, following a disturbance, the
frequency of the MG may change rapidly due to the low
inertia present in the MG. The control of the MS and storage
devices (flywheel) is very important in order to maintain the
frequency of the MG during islanded operation. The
controllers of MS and flywheel inverters respond in
milliseconds. For basic operation of the MG, the controllers
should use only local information to control the flywheel and
MS.
1) Local Frequency Control Strategies
The possible control strategies of the MS and the storage
device may be:
• PQ control (fixed power control);
• Droop control;
• Frequency/Voltage control.
PQ control is adopted so that the MS and the flywheel run
on constant power output. The electricity generated by the MS
may have to be constant because of the needs of the related
thermal loads. In addition, the power output of the flywheel
may be fixed at zero when the MG is operated in grid-
connected mode. As PQ control delivers a fixed power output,
it makes no contribution to local frequency control of the MG.
Therefore, the control scheme of the flywheel has to be
changed from PQ control to droop control or
frequency/voltage control during islanded operation. Droop
control is similar to the function of primary frequency control
in a conventional synchronous generator. The frequency of the
MG can be restored to a steady-state value determined by the
droop characteristic. Frequency/voltage control is similar to
the function of secondary frequency control in the
conventional synchronous generator. The power output of the
flywheel is regulated according to predetermined droop
characteristics.
With droop control action, a load change in the MG will
result in steady-state frequency and voltage deviations,
depending on the droop characteristics and frequency/voltage
sensitivity of the load. The flywheel will contribute to the
overall change in generation. Restoration of the
frequency/voltage of the MG to their normal values requires a
supplementary secondary frequency control action to adjust
automatically the output of the flywheel.
IV. M
ICROGRID BLACK START
The MicroGrid Black Start functionalities were developed
using the control strategy described in section III.A. During
normal operation, the MGCC periodically receives
information from the LC and MC about consumption and
power generation levels, storing this information in a database.
Information about technical characteristics of the different MS
in operation is also stored. MG Black Start involves the
identification of a set of rules and conditions to be checked
during the restoration stage, which should be identified in
advance and embedded in the MGCC software. The
implementation of a BS procedure requires the availability of
some MS with BS capability, which involves an autonomous
local power supply to feed local auxiliary control systems and
launch MS generation. During the restoration of the LV
4
network, the storage will also contribute to face load-tracking
problems, since some microgenerators (fuel-cells,
microturbines) have a slow response and are inertia-less. For
the development of the BS procedure it was assumed that MS
with BS capability are the single-shaft microturbine (SSMT),
the SOFC and the storage device. MS with BS capability
should have batteries in their DC bus. It was also assumed
that, at least during the first stages of this sequence, a multi
master control approach is adopted, being switched to SMO in
the final stages of the BS procedure. The strategies to be
followed make use of the hierarchical control system of the
MG, namely of LC, MC and the MGCC.
A. Sequence of Actions for MicroGrid Black Start
After a system blackout, the MGCC will try to restore the
last MG load scenario. The main problems to deal with during
the restoration procedure include building the LV network,
connecting microgenerators, controlling voltage, controlling
frequency and connecting controllable loads. Considering
these problems the following sequence of actions for MG
restoration should be carried out [3]:
- Disconnection of all loads
in order to avoid large frequency
and voltage deviations when energizing the network. The MG
should also be sectionalized around each MS with BS
capability in order to allow it to feed its own (protected) loads.
- Building the LV network
. The inverter associated with the
storage device will be responsible for LV and Distribution
Transformer (DT) energization. In order to follow the earthing
LV protection guidelines, the MG should keep the earth
reference, available in the earth connection of the neutral of
the DT. Therefore, when building the LV network it is
necessary to energize the DT as soon as possible. When
energizing the DT by the LV side, a large inrush current is
experienced. In order to overcome this problem, transformer
energization should be performed using a ramp-wise voltage
wave form.
- Small islands synchronization
. MS already in operation in
stand alone mode should be synchronized with the LV
network. The synchronization conditions (phase sequence,
frequency and voltage differences) should be verified by local
MC in order to avoid large transient currents.
- Connection of controllable loads to the LV network
is
performed if the MS running in the LV network have the
capacity to supply these loads taking into account the available
energy storage.
- Connection of non-controllable MS
or MS without BS
capability, like PV and wind generators.
- Load increase
. In order to feed as much load as possible,
depending on production capability, other loads can now be
connected.
- Change the control scheme of the inverters
: after service
restoration on the MG, the control schemes of the SSMT and
SOFC inverters are changed from VSI to PQ control. This is
required because batteries that are assumed to be installed in
the DC link of these MS are not suitable to respond to frequent
load variations, since charge and discharge cycles reduce
significantly their life-cycle. On the other hand, flywheel life
is almost independent of the depth of discharge. Flywheel
storage systems can operate equally well on frequent shallow
discharges and on very deep discharges.
- MG synchronization with the MV network
when it becomes
available. The synchronization conditions should be verified
again.
National Technical University of Athens (NTUA)
developed an approach for MG service restoration based on
similar concepts and making use of a Multi Agent System
(MAS). An important issue is that both procedures have a
common rational due to the specific nature of the MS and to
the dimension of the MG. The general idea is that the agents
will execute all the necessary actions (decided in advance)
without human interaction. This approach has two main
advantages. The first advantage is that the computational
demand during the critical event is limited and this is very
important considering that the time limits are very strict and
the processors that will be used in a future MG are not
powerful supercomputers, as in large centralized power
systems. The second advantage lies in the fact that during the
black out there are several communication problems that can
be avoided by using a MAS. Therefore the data exchange
should be even more limited.
V. M
ICROGRID SIMULATION PLATFORMS
A simulation platform under the Matlab/Simulink
environment was developed to study the dynamic behaviour of
several MS operating together in a LV network and controlled
according to what was described in section III.A. The fast
transients associated with the initial moments of the MG
restoration procedure were studied in another simulation
platform developed in EMTP-RV, where the switching details
of the power electronic interfaces were included. The longer
term dynamic behavior of the MG during the restoration
procedure was also evaluated using the Matlab/Simulink
platform. As an illustrative example, Fig. 4 shows the study
case LV network in the Matlab/Simulink simulation platform.
Another simulation tool was also developed in
PSCAD/EMTDC to evaluate MG control functionalities
described in section III.B. Two situations were assumed:
• MS and the storage device represented by synchronous
generators;
• MS and storage represented by STATCOM-BES
devices.
Again, the purpose of this approach was to evaluate the
robustness of the control schemes adopted.
VI. R
ESULTS AND DISCUSSION
This section includes results showing the dynamic
behaviour of the MG (using the different control approaches
previously described) during and subsequent to the islanding
process. Results describing the fast and long term dynamic
behaviour obtained in the MG during the adopted BS
sequence are also described next.
A. Moving to Islanded Operation
Disconnection from the upstream MV network and load-
following in islanded operation was simulated using a SMO
control strategy described in the section III.A. The scenario is
characterized by a local load of 80 kW (65% of impedance
type and 35% of induction motor type) and a local generation
of 50 kW. A fault occurred at t=10s in the MV network
5
followed by MG islanding, 100 milliseconds after.
Due to the large initial frequency deviation, an amount of
load was automatically shed through the activation of load
shedding relays in order to aid frequency restoration. This load
was later reconnected in small load steps (Fig. 5). MS selected
for the secondary load-frequency control (the SOFC and the
SSMT) participate in frequency restoration using a
proportional integral control strategy in the MC. The large
time constants of the MS lead to a relatively slow process for
restoring frequency to its nominal value.
From the frequency behaviour, it may be observed that MG
stability is not lost when facing the short-circuit at the MV
grid side. Speed rotation of motor loads drops considerably
during the fault, which has a great impact in the MG current
and voltage after fault elimination, as can be observed in Fig.
6. The principles presented for current limitation in VSI can
also be observed in Fig. 6.
Fig. 4. LV test network in the Matlab/Simulink simulation platform
Fig. 7 shows the response of the MG when the control
strategy described in section III.B is used and the MS are
represented as STATCOM-BES. The flywheel is using
frequency/voltage control during islanded mode. Results have
shown that after intentional disconnection of the MG from the
main network at t=10s, the output of the MS is still retained at
30 kW. However, the output of flywheel is changed from zero
to around 180kW+j120kvar. Due to the frequency/voltage
control, the frequency and voltage of the MG are restored to
the nominal values.
B. MicroGrid Black Start
In order to study this case it was assumed that a general
collapse took place and was followed by: a) the disconnection
from the MV grid of the MV/LV transformer; b) the
disconnection of loads and c) the automatic creation of islands
operating in standalone mode to supply protected loads
associated with the SSMT and the SOFC.
0 20 40 60 80 100 120 140 160
49.4
49.6
49.8
50
50.2
Frequency (Hz)
0 20 40 60 80 100 120 140 160
-10
0
10
20
30
VSI P & Q (kW / kvar)
0 20 40 60 80 100 120 140 160
0
10
20
30
40
Time (s)
P (kW)
P
Q
SOFC
MTurb
Fig. 5. MG Frequency, VSI active and reactive power and SOFC and single-
shaft microturbine active power
9.95 10 10.05 10.1 10.15 10.2 10.25 10.3 10.35 10.4
-600
-400
-200
0
200
400
600
VSI Current (A)
9.95 10 10.05 10.1 10.15 10.2 10.25 10.3 10.35 10.4
-400
-200
0
200
400
Time (s)
VSI Voltage (V)
Fig. 6. VSI current and voltage during and after the fault
Network behaviour during BS initial stages was evaluated
with the EMTP-RV platform described in [3], including in this
case the fast inverter commutation transients. The VSI shown
in Fig. 4 was selected for energizing the LV network and the
MV/LV transformer, at t=0.2s, using a voltage ramping
control during 0.5s to reduce the magnetizing current of the
DT. The inverter current thus obtained is presented in Fig. 8
where it is possible to observe that the DT magnetizing current
was kept at low values.
In order to get an extended overview of the long term
dynamic behaviour induced by the overall BS procedure, the
MatLab/Simulink simulation platform was used. The
simulations starts considering that the MS are feeding the
