This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2018.2817738, IEEE
Transactions on Smart Grid
1
Two-Stage Load Shedding for Secondary Control
in Hierarchical Operation of Islanded Microgrids
1
Quan Zhou, Student Member, IEEE, Zhiyi Li, Member, IEEE, Qiuwei Wu, Senior Member, IEEE and
Mohammad Shahidehpour, Fellow, IEEE
Abstract—A two-stage load shedding scheme is presented to
cope with severe power deficits caused by microgrid islanding.
Coordinated with the fast response of inverter-based grid-forming
distributed energy resources (DERs), the proposed approach
estimates the load shedding at each stage and the resulting power
flow redistribution. The first stage of load shedding will cease any
rapid frequency decline in which the locally measured frequency
deviation is employed to guide the load shedding level and process.
Once a new steady state is reached at the end of first stage, the
second stage is activated, which performs load shedding according
to load priorities. The effectiveness of the proposed scheme is
verified through time-domain simulations in PSCAD/EMTDC
based on a scaled-down microgrid system.
Index Terms—Microgrid islanding, secondary control,
controllable loads, renewable energy resources, load shedding.
M
I. INTRODUCTION
can operate in grid-connected and island
modes to ensure the continuous supply of power [1]. In
island mode, a microgrid will operate as a self-controlled entity
to regulate its frequency and voltage by dispatching distributed
energy resources (DERs) including wind turbine generator
(WTG), photovoltaic (PV), battery energy storage system
(BESS), and controllable loads [2]. Accordingly, when the
microgrid load is larger than its total generation capacity, the
frequency will drop rapidly due to the low effective inertia and
the DER power output will increase to supply local loads.
Appropriate load shedding procedure would be inevitable if the
supply continues to be less than demand.
In such circumstances, the power deficit which is determined
by the power measurement at the point of common coupling
(PCC) can be mitigated by a combination of fast responses of
local generators and under-frequency load shedding (UFLS)
that can stabilize the islanded microgrid to prevent any
microgrid collapse [3]-[6]. The UFLS settings are predefined,
including frequency threshold values, percentage of shedding at
each step, delay time, etc. However, load shedding schemes
could result in excessive load curtailments as they do not
estimate the power deficiency at each step under prevailing
operating conditions [3]. Adaptive load shedding schemes,
which measure frequency derivatives to estimate power
imbalances, could be utilized [4],[5]. Such schemes monitor the
rate of frequency changes in the equivalent center of inertia and
devise an adaptive load shedding at each step corresponding to
the primary frequency control response. However, presumed
delays in microgrid islanding detections and the oscillatory
nature of frequency derivatives may pose instantaneous
measure errors are considered unacceptable for islanded
ICROGRIDS
Q. Zhou, Z. Li, and M. Shahidehpour are with the Galvin Center for
Electricity Innovation, Illinois Institute of Technology, Chicago, IL 60616
USA (e-mail: qzhou15@hawk.iit.edu; zhiyi.li@hawk.iit.edu; ms@iit.edu). Q.
Wu is with the Center for Electric Power and Energy (CEE), Department of
Electrical Engineering, Technical University of Denmark (DTU), 2800, Kgs.
Lyngby, Denmark (e-mail: qw@elektro.dtu.dk).
microgrids facing rapid frequency drops.
In addition, a very fast primary control response of local
inverter-based DERs, such as WTGs with a BESS, can sustain
the microgrid frequency [7],[8]. In under-frequency conditions,
the primary control response of DERs depends largely on DER
capacity rather than response speed.
This paper focuses on circumstances in which a microgrid is
suddenly disconnected from the utility grid and an optimal level
of load curtailment is required in order to restore the microgrid
frequency expeditiously [9]. The load shedding level and the
related power flow adjustments are adjusted in a two-stage
solution which considers the fast primary frequency control
response of inverter-based DERs for achieving the stated speed
and quantity objectives. The first stage stabilizes the microgrid
system frequency quickly. The proposed scheme considers the
location and the quantity of load shedding for enhancing the
microgrid system dynamics [10],[11]. The frequency deviation
selects loads and estimates the first stage of load shedding
quantity [12]. The estimated curtailment will determine the
corresponding changes in the microgrid frequency and power
flow. The second stage will restore the steady-state frequency
considering load supply priorities [13]. This stage provides an
adequate frequency margin for applying the secondary control
in a microgrid.
Considering the power sharing in the hierarchical control of
microgrids, the secondary control in an island mode will restore
the rated frequency and voltage corresponding to changes in
local demand and generation. However, the secondary control
may fail if the power balance cannot be maintained or the
frequency deviation due to the low effective system inertia is
significant. In such circumstances, sufficient curtailments
within the available time may cease the rapid frequency decline
and prevent the frequency collapse in the islanded microgrid.
Even though the frequency collapse can be averted using a
heuristic method, operating a microgrid at low frequency may
pose a threat to its frequency stability. The microgrid will be
vulnerable to additional disturbances in such occasions because
the key microgrid parameters are designed based on the rated
operating states. If the frequency falls below its set point for a
while, some DERs could stop working resulting in additional
power deficit. The frequency set point defined as a safe
frequency is adjustable. Therefore, load shedding should be
sufficient to recover the frequency to its safe value quickly in
order to prevent undesirable tripping of microgrid generators
[14]. Then the secondary control, which is followed and
implemented in the span of seconds or minutes, restores the
rated microgrid frequency.
The main contributions of the paper are listed as follows:
(1) The correlation between frequency deviation and the
integral of active power deficit is utilized to select proper load
curtailment locations and quantities in an islanded microgrid.
(2) The proposed two-stage load shedding scheme coordinates
1949-3053 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2018.2817738, IEEE
Transactions on Smart Grid
2
load curtailments with fast responses of grid-forming DERs for
improving the steady and the economic operation of an islanded
microgrid.
The remainder of this paper is organized as follows. Section
II discusses the steady state frequency decline and recovery
based on the combined droop characteristics. Section III
illustrates the relation between locally measured frequency
deviations and the integral of local power deficits. A two-stage
load shedding scheme is presented in Section IV. Case studies
are presented in Section V to verify the effectiveness of the
proposed scheme. Section VI concludes the paper.
If the islanded power deficit Pdef measured at PCC exceeds
the frequency regulation range stated in the droop curve,
microgrid DERs will operate at their maximum capacities.
Accordingly, the load curtailment carried out by the primary
frequency control of grid-forming DERs for restoring f restored is
expressed as:
 nG

f  f
gross
Pshed
 Pdef    Pi  k 1
Pk 

f
i
k
1


k


A loss factor  , which is the ratio of initial load to initial
generation, is introduced for considering network losses:
gross
Pshed  1    Pshed
II. FREQUENCY RESPONSE OF AN ISLANDED MICROGRID
Microgrids adopt a droop-based control strategy in island
mode in which DERs are classified as grid-forming and
grid-following control resources. The grid-forming DERs (e.g.,
natural gas turbines, energy storage devices) are operated as
voltage sources to regulate the microgrid frequency and
voltages according to the microgrid demand. The set points of
these dispatchable DERs are determined according to their
interactions with other local grid-forming DERs. The
grid-following DERs (e.g., WTGs, PVs), which are operated as
current sources, follow microgrid frequency and voltage
settings of local grid-forming DERs.
In island mode, the active power deficit can be reduced by
either increasing the DER generation by primary frequency
control or decreasing the system demand by load shedding. So,
load shedding quantities and DER responses are integrated for
frequency restoration in which the corresponding values are
calculated based on steady-state characteristics of microgrids.
In an island mode, the microgrid DER supply will increase
when the utility supply is interrupted, resulting in the frequency
drop. The microgrid load shedding will restore the system
frequency to its pre-islanding values. The range of frequency
regulation is determined by DER droop coefficients and
operating state. If DERs are numbered according to their
descending range of frequency regulations:
m p ,1Pmax,1  m p ,2 Pmax,2  ...  m p , n Pmax, n
(1)
G
G
where m p ,1 is the active power droop coefficient of DER1, and
Pmax,1 is the maximum power regulation capacity of DER1, and
nG is the number of DERs. Fig. 1 shows the aggregated droop
characteristics of microgrid DERs, in which the microgrid
generation response to the frequency deviation in each segment
is expressed as:
k
1
Pk  
 f k
i 1 m p , i
(2)
where Pk is the incremental DER supply in segment k which
corresponds with f k (i.e., a frequency drop from f k 1 to f k ).
In Fig. 1, the restoration of nominal frequency f  in segment
k will determine the required load curtailment. Accordingly, if
the islanded microgrid frequency has dropped to f  , and
f  and f  are located in segments j and k, respectively, the
required load curtailment required for restoring f  is given as:
gross
Pshed

f
j 1
 f  
f j
k 1
Pj   Pi 
i  j 1
 f   f k  P
f k
k
(4)
(3)
(5)
Taking into account the fast response of inverter-based
DERs, it is feasible to use the steady state frequency response to
determine the load curtailment for frequency restoration. Used
in this way, any load curtailment after islanding will be in
conjunction with the primary frequency control of DERs for
frequency regulation.
f
f rated
f nG
f nG 1
f nG
Segment nG
Segment nG  1
f nG 1
..
.
f
f2
f1
Segment k
..
.
f1 P
nG
P1 Segment 1
PnG 1
Remaining
Shedding
P
Fig. 1 Aggregated droop characteristic in a microgrid.
III. CORRELATION OF POWER DEFICIT AND FREQUENCY
DEVIATIONS FOR INVERTER-BASED DERS
In Fig. 2, the primary control structure for a grid-forming
DER includes: (i) power sharing control (lower half of Fig. 2),
(ii) inverter control (upper half of Fig. 2). The power sharing
control will share active and reactive power demands among
DERs in a microgrid according to their droop settings. The
inverter control in Fig. 2, which consists of an outer loop for
voltage control and an inner loop for current control, is aimed at
regulating the inverter voltage and current [15].
Lf
Vi
Vb
Io
Il
DC
Source
Lc
Vo
Cf
f
ref
Viref
Voltage
Loop
Microgrid
Distribution
Network
Il
Vo
Ilref
Viref
Current
Loop
Vo,n
Voref
Load
Qn
Voref
Q
P/Q
Calculation
P
f ref
Vo
Io
P rated
f rated
Fig. 2. Primary control of DER.
1949-3053 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2018.2817738, IEEE
Transactions on Smart Grid
3
The droop control characteristic of DER i is expressed as:
rated
f inv ,i  f invrated
(6)
, i  mP , i   Pinv , i  Pinv , i 
rated
Vinv ,i  Vinvrated
, i  nQ , i   Qinv , i  Qinv , i 
(7)
where mP ,i and nQ ,i are droop coefficients of active and reactive
power for DER i, respectively, f inv ,i and Vinv ,i are the frequency
rated
and the voltage of DER i, respectively, f invrated
, i and Vinv , i are the
rated frequency and voltage of DER i, respectively, Pinv ,i and
Qinv ,i are the active and the reactive power of DER i,
rated
respectively, Pinvrated
, i and Qinv , i are the rated active and reactive
power of DER i, respectively.
The instantaneous active pinv ,i and reactive power qinv ,i of
DER for power sharing control are calculated as a function of
output voltage and current, i.e.,
(8)
pinv ,i  Vinvd ,i  I invd ,i  Vinvq ,i  I invq ,i
qinv ,i  Vinvq ,i  I invd ,i  Vinvd ,i  I invq ,i
(9)
where Vinvd ,i and Vinvq ,i are the d-axis and q-axis voltage
components of DER i, I invd ,i and I invq ,i are the d-axis and q-axis
current components of DER i.
To eliminate high-frequency harmonics, the instantaneous
active and reactive power are passed through a low-pass filter,
Pinv ,i 
fc
fc
pinv ,i , Qinv ,i 
qinv ,i
s  fc
s  fc
(10)
where f c is the cut-off frequency of the low-pass filter.
Equivalently,
dPinv ,i
dt
dQinv ,i
dt
 f c Pinv ,i  f c Vinvd ,i  I invd ,i  Vinvq ,i  I invq ,i 
(11)
 f cQinv ,i  f c Vinvq ,i  I invd , i  Vinvd ,i  I invq , i 
(12)
We define the instantaneous power demand from the
distribution network Pe ,i as:
Pe ,i  Vinvd ,i  I invd ,i  Vinvq ,i  I invq ,i
(13)
Then the relationship between frequency derivative and local
power deficit (analogous a swing equation) is stated as:
1 df inv ,i
 Pinv ,i  Pe ,i  Pdef , i
mP , i f c dt
microgrids with low system inertia. However, the relationship
between local frequency deviations and the integral of power
deficit is stated as:
f i  mP ,i f c   Pdef ,i  t dt
(15)
Hence, a local frequency deviation distinguishes microgrid
locations with higher power deficits. The frequency deviation
among microgrid locations will continue until a new steady
state frequency is reached. In (15), the frequency deviation and
the integral of local power deficit must shrink after islanding as
shown in our simulation studies. Therefore, local frequency
deviations and their relative differences in various locations
facilitate load shedding decisions in low inertia microgrids.
IV. PROPOSED TWO-STAGE LOAD SHEDDING SCHEME
The proposed two-stage load shedding scheme addresses the
power imbalance caused by microgrid islanding. The first stage
is to cease rapid frequency drop and allows the frequency to
settle at a temporary settling point. Further load shedding is
triggered at the second stage to recover the system frequency to
a predetermined safe value.
In Fig. 3, the microgrid frequency will deviate from its rated
Point A and settle in Point B as load shedding occurs in island
mode. At times, the frequency might continue to drop below
Point B before it rebounds, which could be due to a delay in the
frequency support provided by load shedding or a slow
response of certain types of DERs, like hydro turbines [16].
Within a time limit, additional load shedding is activated to
recover the frequency to a predetermined Point C. At this time,
the secondary control is initiated to restore the rated frequency
at Point D.
f
60 Hz
Ceasing
Remediation
Restoration
A
D
59.5 Hz
f safe
C
B
f temp
(14)
If a microgrid power deficit occurs due to a sudden change in
its operating condition (e.g., islanding, generator unit tripping,
load change, etc.), the power deficit will be distributed among
participating DERs. Accordingly, the magnitude of power
deficit at different microgrid locations will depend largely on
individual DER responses and the transient DER response will
depend on the microgrid damping conditions. In such
circumstances, the microgrid frequency derivative is an
instantaneous indicator of local power deficit, which is very
sensitive to pertinent transients and difficult to measure [8].
When facing a power deficit, the microgrid frequency could
typically decline at the rate of 10 Hz/second, due to low system
inertia. Accordingly, the available load shedding period for
lowering the power deficit is very limited. Although frequency
derivative could be a proper index for improving the load
shedding performance, the index calculation would be a
prohibitive task due to the oscillatory nature of the index in
Stage 1
Stage 2
Secondary Control
Time t
Fig. 3. Frequency response during load shedding process.
The proposed flowchart is depicted in Fig. 4 and the steps are
further analyzed as follows.
A. First Stage of Load Shedding Scheme
At the first stage shown in Fig. 4, the objective is to retain a
normal operation in a microgrid after it faces severe power
deficits caused by islanding. The loads designated for
curtailment should be selected based on microgrid system
dynamics rather than economics. Hence, locally measured
frequencies are utilized after islanding to distinguish most
effective loads for quickly ceasing and stabilizing the system
frequency.
We consider the following steps in order to optimize the
quantity and the location of load shedding at the first stage:
1949-3053 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2018.2817738, IEEE
Transactions on Smart Grid
4
1) Load Shedding Quantities at the First Stage
According to the first stage objective, load shedding can be
divided into two parts:
a) Direct part: Eliminates the mismatch between available
power generation and load demand, namely,
nD
nG
j 1
i 1
Pmis   PD , j   PGmax
,i
(16)
where nD is the number of loads, PD , j is the demand of load j,
and PGmax
, i is the maximum capacity of DER i. However, this
provision may fail to provide a sufficient frequency support.
b) Marginal part: Restores a temporary frequency (i.e., Point B
in Fig. 3) for providing a certain margin for frequency stability
in the islanded microgrid. The corresponding load curtailment
is obtained by (3).
Start
Islanding Detection
Estimation of shedding amount of
stage1
the first stagePshed
stage1
Pshed
0
N
Y
Load shedding selection based on
decision criteria
I
Stop criterion
N
df dt  0
Y
Estimation of shedding amount of
stage 2
the second stage Pshed
stage 2
Pshed
0
N
Y
Load shedding allocation based on
frequency deviations and
redistributed power flow
Y
Load shedding allocation based on
load priority
stage1
Shed Pshed
of loads
Shed Pshed of loads
increments and the system frequency to loading increments are
calculated using an efficient power flow tracing method. Such
sensitivities are computed based on steady state droop
characteristics in an islanded microgrid. Accordingly, buses
with faster frequency drops will demonstrate higher sensitivity
values, which can participate effectively in load shedding.
In island mode, local frequencies may start to deviate from
the rated value of 60Hz with a given threshold value of 59.5Hz
as shown in Fig. 3. Once the local frequency deviates below
59.5 Hz, the corresponding loads will be subject to curtailment.
The stopping criterion for the load curtailment selection is
given as:
stage1
(17)
 PDnet, j  Pshed
iS
where S is the set of selected loads, PDnet, j is the net power
stage1
is the estimated load
injection at load bus j, and Pshed
shedding at the first stage. Once the stopping criterion is
satisfied, the selection process will stop.
The curtailment usually occurs where local and neighboring
DERs are insufficient for supplying the local demand. When
facing a severe power deficit, a higher frequency threshold will
be set so that the load shedding scheme can be triggered more
quickly.
3) Load Shedding Allocation at the First Stage
II
stage 2
End
Fig. 4. Flowchart of the secondary control with two-stage load shedding.
2) Load Shedding Locations at the First Stage
The first stage is to rescue the microgrid frequency drop in
islanding mode. So, the proposed approach will quickly
identify several effective loads for curtailment considering the
frequency response delay, especially in large microgrids with a
large number of loads. It is essential to ensure continuous
supply of critical loads (e.g., hospitals, police stations, data
centers, major traffic lights, etc.) and these critical loads must
be excluded from load shedding. In addition, curtailment
selection should exclude specific loads whose curtailment in
islanded microgrid may have detrimental effects on the
frequency or threaten the microgrid operation [16],[17].
A microgrid may experience temporary network
reconfiguration among local feeders during the transient
process resulting in a redistribution of microgrid power flow.
For grid-forming DERs, the redistribution of power flow after
islanding may result in an instantaneous demand fluctuation. In
addition, as discussed in Section III, the severity of local power
deficits is manifested in frequency deviations among
grid-forming DERs which will determine the frequencies for
local grid-following DERs and loads.
The objective of the proposed load shedding scheme and the
subsequent secondary control is to restore the rated frequency.
In [18],[19], the sensitivities of generator output to loading
After selecting load curtailment locations, the exact load
curtailment is determined using frequency deviations and the
estimated redistribution of microgrid power flow, which would
be different from the initial load and power flow distribution
[19]. Accordingly, a load bus with a larger frequency deviation
and heavier net load will have a larger share of curtailment.
Some DERs located close to load buses may be able to
respond quickly to supply the corresponding loads. After
islanding, any load curtailments at such buses may lead to
additional frequency fluctuations. Hence, local frequency
deviations along with redistributed power flows determine the
load shedding allocations as:

stage1
net
Pshed
, j   f D , j  PD , j

nS
 f
j 1
D, j

stage1
 PDnet, j   Pshed

(18)
stage1
net
where Pshed
, j is the curtailment at load j in the first stage, PD , j is
the net power injection at load bus j after islanding, and f D , j is
the locally measured frequency of load j.
Accordingly, the selection and allocation of load shedding at
this stage are determined by temporary and final redistribution
of power flows. This first stage process ought to be
implemented expeditiously after islanding; otherwise, power
imbalances will lead to severe microgrid transients and even
system collapse.
B. Second Stage of Load Shedding Scheme
For avoiding undesirable generator tripping, the second stage
of load shedding will recover the microgrid frequency from its
temporary setting to its safe value (i.e., shifting from Point B to
Point C in Fig. 3) which will be followed by secondary control
for restoring the rated frequency. The secondary control would
shift each DER’s operating point to mitigate frequency and
voltage deviations caused by the primary droop-based control.
1949-3053 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2018.2817738, IEEE
Transactions on Smart Grid
5
However, the secondary control might result in a lower
damping which could lead to oscillatory responses [20].
Therefore, the objective of second stage load shedding is to
provide an adequate frequency margin for the restoration of
rated frequency. The additional curtailment at the second stage
is calculated according to the steady state frequency deviation
at the end of the first stage and recovery characteristics
embedded in the islanded microgrid.
After the first stage load shedding, the power imbalance will
be alleviated as the microgrid reaches a temporary steady state.
The delay time at the second stage is not so critical. The load
curtailment selection at this stage is determined by load
priorities rather than the microgrid stability. At the second
stage, remaining loads are classified according to their
priorities and those with lowest priorities are curtailed first.
The second stage is designed to recover the frequency from
the temporary settling frequency f1 to the safe value
f safe ( f r  f safe  f r 1 ). In Fig. 5, the loads are designated as Type
1 to Type r-1 for curtailment, which can shift the microgrid
frequency from f1 to f r . Furthermore, additional curtailments
are required to increase the frequency from f r to f safe . This
provision is proportionally shared among nr loads of Type r.
So, the recovery gain of load j is defined as:
(19)
mDr , j   f r 1  f r  PDr , j
where 1  j  nr and PDr , j is the initial demand of load j of Type r.
According to (19), the curtailment of each load of Type r is
expressed as:
Pshed , j   f safe  f r  mDr , j
(20)
The designated load curtailment is calculated as:
r 1
f safe  f r
l 1
f r 1  f r
stage 2
Pshed
  PDl 
 PDr
(21)
nl
where PDl is the total demand of Type l, i.e., PDl   PDl ,i ,
i 1
and PDr is the total initial demand of Type r.
f
.
.
.
fr+1
V. CASE STUDIES
A. Microgrid System
In Fig. 6, the microgrid is a scaled-down three-phase
balanced microgrid system with five DERs and six aggregate
loads [20]. DER 1-4 are interactive grid-forming DERs
whereas DER 5 is a non-interactive grid-following DER. The
microgrid is connected to the utility grid at the PCC through a
circuit breaker, representing a common microgrid structure.
The detailed technical parameters of the microgrid are shown in
Table I. Since DERs are larger than 30kW, the safe frequency is
set at 59Hz and the corresponding time limit is 5s [14].
The total delay for activating relays and load shedding is
50ms (3 cycles). The communication delay and calculation
time required by microgrid master controller will take 50ms at
the first stage. For comparison, we consider a UFLS scheme
with a 4-step load shedding of 35-30-20-15% at threshold
frequencies, 59, 58.8, 58.4 and 58Hz. These frequency
thresholds and load shedding allocations at each step are
deliberately chosen to minimize the load curtailment at higher
levels of power deficit [3].
We consider the following assumptions for simulation:
1) Grid-following DERs are WTGs, and grid-forming DERs
are a combination of WTGs and BESS;
2) Grid-following DERs are modeled as PQ-controlled voltage
source converters for maintaining pre-determined active and
reactive power outputs;
3) Grid-forming DERs are modeled as droop-controlled voltage
source converters, which interact with other DERs for sharing
the supply of loads autonomously;
4) DERs are connected to the microgrid distribution network
through identical LCL filters;
5) Loads are controllable. If a load is curtailed, its active and
reactive power consumptions (determined by the load power
factor) will both be adjusted;
6) For improving the system dynamic performance, a fraction
of L6, which is directly supplied by a grid-following DER, does
not participate in load shedding unless this DER is tripped.
The proposed load shedding scheme is tested in the
PSCAD/EMTDC platform using this microgrid system.
mDr , j
.
.
.
Utility
Grid
f safe
r
D ,1
m
mDr ,nr
fr
f2
.
.
.
m1D ,1
DER1
PD1 ,1
PDr ,1
Pshed , j
Bus 2
DC/AC
L1
PDr ,nr
Pshed
Fig. 5. Frequency recovery corresponding to load shedding levels.
At this stage, load curtailment is determined based on
individual load priorities, ensuring that loads with low priority
are curtailed first and critical loads are preserved.
DC/AC
L2
Type 1
PD1 ,n1
DER2
DER4
L6
Bus 3
DC/AC
Bus 4
f1
Bus 0
Line 3
Line 2
Line 1
Bus 1
m1D ,n1
DC/AC
PCC
Type r
.
.
.
DER5
DC/AC
L3
DER3
Line 4
Bus 5
Line 5
L5
L4
Fig. 6. A schematic diagram of the studied microgrid system
Table I Technical specifications of the studied microgrid
DER
Parameter
mP (Hz/W)
nQ (V/Var)
Prated (kW)
Pmax (kW)
DER 1
5e-2
1e-4
10
30
DER 2
3.75e-2
1e-4
10
50
DER 3
2.5e-2
1e-4
10
90
1949-3053 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2018.2817738, IEEE
Transactions on Smart Grid
6
DER 4
DER 5
2.5e-2
-
Parameter
Line 1, 2, 3
Line 4, 5
Parameter
Load 1
Load 2
Load 3
Load 4
Load 5
Load 6
1e-4
Network
Resistance R()
0.1
0.05
Loads
Priority
Critical
Non-critical
Non- critical
Semi- critical
Semi- critical
Semi- critical
10
60
110
60
Reactance X(mH)
0.5
0.25
Active Power (kW)
60
80
80
20
20
100
B. Simulation Cases
Three simulation cases are carried out:
Case 1: Microgrid is islanded from the utility grid;
Case 2: Microgrid islanding with a lower load demand;
Case 3: Microgrid islanding with an outage of DER;
The microgrid islanding is simulated by opening the circuit
breaker at t=1s. The simulation is recorded for 15s.
Case 1: Microgrid Islanded from the Utility Grid
Before islanding, the total load is 360kW and the microgrid
network loss is 15kW. The supply from the utility grid and
DERs are 275kW and 100kW, respectively. So, the loss factor
is 4.2%. The islanding interrupts the utility supply and causes
severe active power deficit. The DERs increase their supplies in
the islanded microgrid. However, the power deficit will
continue due to the limited DER capacity.
The load curtailment at the first stage is determined by (4)
and (5) as 52.8kW. According to the locally measured
frequencies, shown in Fig. 7, L6 and L2 are selected
successively to participate in the first stage load shedding (LS)
with the corresponding allocations determined by (18). The
proposed load shedding provides a sufficient support for
system frequency, which is ceased at 58Hz. At t=3s, the second
stage load shedding is activated and shared by L2 and L3 due to
their lower priorities. The corresponding curtailment is
calculated by (21) as 89.6kW. Accordingly, the frequency is
recovered to the safe value 59Hz. At t=10s, the secondary
control is initiated which restores the rated 60Hz frequency.
The microgrid frequency response and active power adjustment
are depicted in Figs. 7 and 8, respectively. For the 4-step UFLS
scheme, the first three load shedding steps are triggered at
t=1.22s, 1.25s and 1.35s, respectively. In the first step, only L2
and L3 are selected due to their low priorities. Then all the other
loads, except for the critical load L1, will be involved in load
shedding. Fig. 9 shows the comparison of system frequency
responses in the two schemes, with the corresponding
information on activated load shedding for each scheme
presented in Table II.
In the 4-step UFLS scheme, there is a delay in the frequency
support provided by load shedding, resulting in the
over-curtailment of loads. Although both schemes have limited
the subsequent frequency alterations to 58Hz after islanding,
the level of curtailment and the number of participating loads
are much fewer in the proposed scheme.
Table II Load shedding from each load in the two schemes for Case 1(kW)
Proposed Load Shedding Scheme
Frequency [Hz]
Stage
Activation
Time (s)
L1
L2
1
1.27
2
3.00
Load Shedding (kW)
0
0
0
22.63
37.42
60.05
Load (kW)
L3
L4
0
52.18
52.18
0
0
0
L5
L6
Sum
0
0
0
30.17
0
30.17
52.8
89.6
142.4
L5
L6
Sum
0
2.9
13.2
0
16.1
0
5.8
26.4
0
32.2
92.4
79.2
52.8
0
224.4
Conventional 4-Step UFLS Scheme
Step
Fig. 7. Frequency response using the proposed LS scheme in Case 1
Activation
Time (s)
L1
L2
L3
Load (kW)
L4
1
1.22
2
1.25
3
1.35
4
Load Shedding (kW)
0
0
0
0
0
46.2
33.8
0
0
80
46.2
33.8
0
0
80
0
2.9
13.2
0
16.1
Active Power [kW]
Case 2: Islanding with a Lower Load Demand
Frequency [Hz]
Frequency [Hz]
Fig. 8. Active power of DERs and utility grid in Case 1
In this case, the islanding and partial outage of L3 by 60kW
occur simultaneously as depicted in Fig. 10 in which the lower
power deficit by the proposed scheme is successfully
recognized. By (4) and (5), there are no loads to be curtailed at
the first stage. At t=3s, the new steady state is reached and the
second stage will curtail L2 and L3 by 65.92kW and 16.48kW,
respectively, as determined by (21). Since the curtailment does
not occur at stage one, the final curtailment of each load is only
decided by their priorities.
Fig. 9. Comparison of system frequency response in Case 1
Fig. 10. Frequency response using the proposed LS scheme in Case 2
1949-3053 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2018.2817738, IEEE
Transactions on Smart Grid
7
UFLS may result in additional load shedding. Comparatively,
the proposed load shedding scheme can successfully rescue the
frequency deviations using a smaller level of curtailment.
Active Power [kW]
For this case, only two shedding steps are triggered in the
4-step UFLS scheme since the initial power deficit is lower.
The curtailment is reduced to 134.16kW. But, there are still five
loads involved in the load shedding procedure. Table III
presents the curtailment of each load in each scheme. In Fig. 11,
the frequency performance of the 4-step scheme is better than
that that in Case 1. When facing lower power deficits, the
proposed scheme still has a better performance than that of the
4-step scheme concerning the level of curtailment and the
number of involved loads.
Frequency [Hz]
Frequency [Hz]
Fig. 12. Active power of DERs and utility grid in Case 3
Fig. 11. Comparison of system frequency response in Case 2
Table III Load shedding from each load in the two schemes for Case 2(kW)
Activation
Time (s)
1
2
3.00
Load Shedding
Step
Activation
Time (s)
1
1.27
2
1.32
3
4
Loads Shedding
Proposed Load Shedding Scheme
Load
L1
L2
L3
L4
L5
0
0
0
0
0
0
65.92
16.48
0
0
0
65.92
16.48
0
0
Conventional 4-Step UFLS Scheme
Load
L1
L2
L3
L4
L5
0
57.79
14.45
0
0
0
22.21
5.55
8.54
8.54
0
0
0
0
0
0
0
0
0
0
0
80.00
20.00
8.54
8.54
L6
0
0
0
Sum
0
82.40
82.40
L6
0
17.08
0
0
17.08
Sum
72.24
61.92
0
0
134.16
Case 3: Islanding with an Outage of DER
The DER 5 is operated as a non-interactive grid-following
WTG. However, if the wind speed exceeds its cut-out value,
this DER will be tripped to avoid damages to WTG.
In this case, we assume the power deficit is caused by the
simultaneous microgrid islanding and the tripping of DER 5.
Compared to Case 1, there is an additional power deficit
resulted by the DER 5 outage. Accordingly, the active power
responses of DERs and the utility grid are also changed as
depicted in Fig. 12. In the proposed load shedding scheme, L6
and L2 are selected to share the curtailment at the first stage to
112.8kW at t=1.24s. The minimum frequency is 57.5Hz and the
temporary frequency is set at 58Hz which is reached at t=3s. At
the second stage, the curtailment is 89.6kW, which is allocated
among L2 and L3 according to their available capacities. At
t=10s, the frequency is gradually restored by the subsequent
secondary control. Fig. 13 depicts the frequency response of the
islanded microgrid.
A comparison of the proposed LS scheme with the 4-step
UFLS scheme is shown in Fig. 14. The system frequency drops
more quickly than that in Cases 1 and 2, which is because of the
additional power deficit. Although larger curtailments have
occurred during the first few steps, the frequency still falls
below 58Hz, resulting in the triggering of all four load shedding
steps. Even the critical load L1 is partly curtailed. The
excessive load shedding increases the system frequency above
60Hz in which the curtailment at each scheme is shown in
Table IV. So, when facing a higher power deficit, the 4-step
Fig. 13. Frequency response using the proposed LS scheme in Case 3
Frequency [Hz]
Stage
Fig. 14. Comparison of system frequency response in Case 3
Table IV Load shedding from each load in the two schemes for Case 3(kW)
Proposed Load Shedding Scheme
Stage
Activation
Time (s)
1
1.24
2
3.00
Load Shedding
Step
Activation
Time (s)
1
1.19
2
1.21
3
1.26
4
1.32
Load Shedding
Load
L1
L2
L3
L4
0
26.03
0
0
0
36.10
53.50
0
0
62.13
53.50
0
Conventional 4-Step UFLS Scheme
L5
L6
Sum
0
0
0
86.77
0
86.77
112.80
89.60
202.40
Load
L1
L2
L3
L4
L5
L6
Sum
0
0
0
24.03
24.03
56.70
23.30
0
0
80.00
56.70
23.30
0
0
80.00
0
7.23
9.26
3.51
20.00
0
7.23
9.26
3.51
20.00
0
36.14
46.30
17.56
100.00
113.40
97.20
64.80
48.6
324.0
In the three cases, load shedding levels at the conventional
4-step UFLS scheme are 224.4kW, 134.16kW, and 324kW,
respectively. The rapidly increasing frequency deviation, which
is due to a delay in frequency recovery, may trigger additional
load shedding steps, resulting in excessive load shedding and
even over-frequency conditions. Clearly, there are limitations
for UFLS schemes (which is triggered step by step) to deal with
the rapid frequency deviations in low inertia microgrids.
Comparatively, the proposed load shedding scheme reduces
load curtailment requirements in the three cases to 142.4kW,
82.4kW and 202.4kW, respectively. The proposed approach
makes use of primary control of microgrids by carefully tuning
load curtailment quantities and locations. Accordingly, the
proposed scheme offers a desirable frequency regulation
1949-3053 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2018.2817738, IEEE
Transactions on Smart Grid
8
performance at various power deficit levels in islanded
microgrids. In addition, the number of involved loads is also
reduced in the proposed load shedding approach.
VI. CONCLUSIONS
In order to cease the rapid frequency deviations, a load
shedding scheme may result in the curtailment of additional
loads. In addition, there could be a delay before the active
power compensation provided by load shedding is reflected in
the system frequency response. Therefore, even with carefully
adjusted load shedding parameters, such as number of steps,
frequency thresholds, and curtailments at each step, the
frequency regulation performance may still be undesirable in
different operating conditions. Ths situation will be aggravated
in islanded microgrids with a low system inertia, in which rapid
frequency deviations may easily trigger additional load
shedding steps which tend to result in excessive levels of load
curtailment in islanded operating conditions, as depicted in our
simulation results.
This paper presented a two-stage load shedding scheme to
address the severe under frequency conditions after islanding.
Considering the fast primary frequency response of
inverter-based grid-forming DERs, we adjusted load
curtailments at each stage according to the actual system
operating conditions. The first stage was responsible for
ceasing a rapid frequency deviation and rescuing the islanded
microgrid system from collapse. Effective loads were selected
for curtailment, which was based on locally measured
frequencies and the final power flow distribution. The second
stage fulfilled the steady state frequency recovery between two
settling frequencies which ensured that the following secondary
control would be able to restore the rated frequency. The
time-domain simulation case studies were presented to show
the significant performance of the proposed load shedding
scheme in managing the frequency regulation.
REFERENCES
[1] M. Shahidehpour, Z. Li, S. Bahramirad, Z. Li and W. Tian, “Networked
microgrids: exploring the possibilities of the IIT-Bronzeville grid,” IEEE
Power Energy Mag., vol. 15, no. 4, pp. 63-71, July-Aug. 2017.
[2] L. Che, M. Shahidehpour, A. Alabdulwahab and Y. Al-Turki,
“Hierarchical coordination of a community microgrid with AC and DC
microgrids,” IEEE Trans. Smart Grid, vol. 6, no. 6, pp. 3042-3051, Nov.
2015.
[3] A. Ketabi and M. H. Fini, “An underfrequency load shedding scheme for
hybrid and multiarea power systems,” IEEE Trans. Smart Grid, vol. 6, no.
1, pp. 82-91, Jan. 2015.
[4] T. Shekari, F. Aminifar and M. Sanaye-Pasand, “An analytical adaptive
load shedding scheme against severe combinational disturbances,” IEEE
Trans. Power Syst., vol. 31, no. 5, pp. 4135-4143, Sept. 2016.
[5] M. Karimi, P. Wall, H. Mokhlis and V. Terzija, “A new centralized
adaptive underfrequency load shedding controller for microgrids based on
a distribution state estimator,” IEEE Trans. Power Del., vol. 32, no. 1, pp.
370-380, Feb. 2017.
[6] Z. Wang, B. Chen, J. Wang, and C. Chen. “Networked microgrids for
self-healing power systems,” IEEE Trans. Smart Grid, vol. 7, no. 1, pp.
310-319, Jan. 2016.
[7] Y. J. Kim, J. Wang and X. Lu, “A Framework for Load Service
Restoration Using Dynamic Change in Boundaries of Advanced
Microgrids with Synchronous-Machine DGs,” IEEE Trans. Smart Grid,
to be published. DOI: 10.1109/TSG.2016.2638854.
[8] IEEE Guide for the Application of Protective Relays Used for Abnormal
Frequency Load Shedding and Restoration, IEEE Std. C37. 117-2007,
Aug. 2007.
[9] NPCC Document A-03, Emergency Operation Criteria, Northeast Power
Coordinating Council (NPCC), Aug. 31, 2004, p. 4.
[10] F. Ceja-Gomez, S. Qadri, and F. Galiana, “Under-frequency load
shedding via integer programming,” IEEE Trans. Power Syst., vol. 27, no.
3, pp. 1387–1394, Aug. 2012.
[11] Q. C. Zhong and G. Weiss, “Synchronverters: Inverters that mimic
synchronous generators,” IEEE Trans. Ind. Electron., vol. 58, no. 4, pp.
1259-1267, April 2011.
[12] J. Tang, J. Liu, F. Ponci and A. Monti, “Adaptive load shedding based on
combined frequency and voltage stability assessment using
synchrophasor measurements,” IEEE Trans. Power Syst., vol. 28, no. 2,
pp. 2035-2047, May 2013.
[13] E. Pashajavid and A. Ghosh, “Frequency support for remote microgrid
systems with intermittent distributed energy resources—A two-level
hierarchical strategy,” IEEE Trans. Power J., to be published. DOI:
10.1109/JSYST.2017.2661743.
[14] IEEE Guide for Conducting Distribution Impact Studies for Distributed
Resource Interconnection, IEEE Std. 1547.7-2013, Feb. 28 2014.
[15] N. Pogaku, M. Prodanović, and T. C. Green, “Modeling, analysis and
testing of autonomous operation of an inverter-based microgrid,” IEEE
Trans. Power Electron., vol. 22, no. 2, pp. 613–625, 2007.
[16] D. Rodriguez Medina et al., “Fast assessment of frequency response of
cold load pickup in power system restoration,” IEEE Trans. Power Syst.,
vol. 31, no. 4, pp. 3249-3256, July 2016.
[17] Z. Li, M. Shahidehpour, F. Aminifar, A. Alabdulwahab and Y. Al-Turki,
“Networked Microgrids for Enhancing the Power System Resilience,”
Proc. IEEE, vol. 105, no. 7, pp. 1289-1310, July 2017.
[18] Y. Wang, I. R. Pordanjani, W. Li, W. Xu, and E. Vaahedi, “Strategy to
minimise the load shedding amount for voltage collapse prevention,” IET
Gener. Transm. Distrib., vol. 5, no. 3, pp. 307–313, Mar. 2011.
[19] C. P. Reddy, S. Chakrabarti and S. C. Srivastava, “A Sensitivity-Based
Method for Under-Frequency Load-Shedding,” IEEE Trans. Power Syst.,
vol. 29, no. 2, pp. 984-985, March 2014.
[20] X. Wu and C. Shen, “Distributed optimal control for stability
enhancement of microgrids with multiple distributed generators,” IEEE
Trans. Power Syst., vol. 32, no. 5, pp. 4045-4059, Sept. 2017.
BIOGRAPHIES
Quan Zhou (GSM’16) received the B.S. and M.S. degrees from the
Department of Electrical Engineering, Shanghai Jiaotong University, Shanghai,
China, in 2011 and 2016, respectively. He is currently pursuing the Ph.D.
degree in the Electrical and Computer Engineering Department, Illinois
Institute of Technology. His research interests include microgrid control and
optimization.
Zhiyi Li (GSM’14-M’17) received the B.S. degree in electrical engineering
from Xi’an Jiaotong University, Xi’an, China, in 2011 and the M.S. degree in
electrical engineering from Zhejiang University, China, in 2014. He completed
his Ph.D. degree in 2017 in the Electrical and Computer Engineering
Department at Illinois Institute of Technology. He is a visiting faculty in the
Robert W. Galvin Center for Electricity Innovation at Illinois Institute of
Technology.
Qiuwei Wu (SM’15) received the B.Eng. and M.Eng. degrees in power system
and its automation from Nanjing University of Science and Technology,
Nanjing, China, in 2000 and 2003, respectively, and the Ph.D. degree in power
system engineering from Nanyang Technological University, Singapore, in
2009. He is Editor of IEEE Transactions on Smart Grid and IEEE Power
Engineering Letters. He is also Associate Editor of International Journal of
Electrical Power and Energy Systems. His research interests are in smart grids,
wind power, electric vehicle, active, distribution networks, electricity markets,
and smart energy systems.
Mohammad Shahidehpour (F’01) received the Honorary Doctorate degree
from the Polytechnic University of Bucharest, Bucharest, Romania. He is a
University Distinguished Professor and Bodine Chair Professor and Director of
the Robert W. Galvin Center for Electricity Innovation at Illinois Institute of
Technology. He is a member of the US National Academy of Engineering and a
Fellow of the American Association for the Advancement of Science (AAAS).
1949-3053 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.