Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Analysis and Design of Interfacing Inverter Output Virtual Impedance in

A Low Voltage Microgrid


Jinwei He Yun Wei Li
Student member Member
Department of Electrical and Computer Engineering, University of Alberta, Edmonton, Canada, T6G 2V4
E-mail: hjinwei@ualberta.ca, yunwei.li@ece.ualberta.ca

Abstract -- This paper presents the control strategies and through controlling the DG interfacing converters. For
output (virtual) impedance design approach for parallel instance, the microgrid can operate in both grid-connected
operated distributed generation (DG) units in a low voltage (LV) mode and islanding mode, which provides enhanced
microgrid system. To achieve communicationless control among reliability and power quality. To facilitate the smooth
DG interfacing converters, virtual impedances are placed operation mode transitions, voltage control method in both
between the interfacing converter outputs and the grid to
improve system stability and mitigate coupling between real and
the grid-connected mode and islanding mode is an attractive
reactive power. In order to properly design the virtual option for a microgrid [3, 4, 6].
impedance, small-signal models of a microgrid in different The control of DG in a microgrid can be realized by
operation modes are developed using complex number matrix, measuring local available signals without communications
which is flexible and can be easily extended when more DG units among DG systems, through the communicationless control
are added. Based on the developed microgrid models, system or decentralized control. In this control category, the
damping and stability are evaluated. Together with the DG frequency and voltage magnitude droop control is one of the
power capacity and power decoupling requirements, the desired most popular methods for real (P) and reactive (Q) power
DG output virtual impedance can be designed. To realize the regulation [6, 7, 10-13]. However, when the traditional droop
virtual impedance, a robust implementation method that avoids
using of any high/low pass filters is presented. Moreover, to
control method is applied to a low-voltage microgrid, it is
further improve the power control performance during subject to several particular problems [6]:
transients, a transient power decoupling algorithm and an 1) The P-Q coupling is serious due to the presence of
adaptive transient impedance control scheme are proposed. nontrivial feeder resistance in a LV microgrid;
Experimental results are provided to validate the impedance 2) The DG system stability and transient performance will be
design approach, the virtual impedance implementation method affected if no additional coupling inductor is connected at the
and the proposed transient power control strategies. DG output.
To address some of the aforementioned problems, [8]
Index Terms—Microgrid; distributed generation (DG); proposed a virtual real and reactive power frame
virtual and adaptive impedance; small-signal model; power transformation method to decouple the real and reactive
coupling; droop control.
power. However, actual real and reactive power cannot be
controlled and shared directly in this method. To improve the
I. INTRODUCTION
power sharing accuracy, additional high frequency signal
In recent years, an increasing number of renewable energy injection is proposed in [9]. In this method, accurate power
resources (RES) or micro-sources such as photovoltaic, small sharing is achieved with the costs of additional voltage
wind turbines and fuel cells are incorporated into power grid distortions to the system. Another effective approach to
in the form of distributed generation (DG). DG units are improve transient and suppress power coupling is to add
normally connected to utility grid through power-electronics- virtual inductances at the DG output [6, 10, 11]. It has been
interfacing converters [1]. The stability and efficiency of DG shown that if the virtual impedance is properly considered in
systems highly depend on the control performance of the the power control and sharing scheme, both transient and
interfacing converters. steady-state power control and sharing performance of the
According to current grid standards [2], DG units need to droop method can be improved. However, if poorly designed
be disconnected from the point of common coupling (PCC) or implemented, the virtual impedance method may induce
when utility is not available. However, with the increasing output current distortions and therefore adversely affect the
penetration of DG systems, intentional islanding operation system stability and dynamics.
will be an important function in the near future. In contrast to To ensure effective virtual impedance aided DG inverter
conventional current controlled method, the DG output control in a LV microgrid, this paper presents a DG output
voltage control method shall be adopted to provide voltage impedance design approach and a robust implementation
support to loads when islanding operation is allowed. In method. The impedance design is based on the developed
addition, with the development of DG, the concept of microgrid models in both grid-connected and islanding
microgrid which contains a number of systematically modes, with the considerations of system transient and
organized DG units has been proposed to achieve better stability performances, DG power capacities and P-Q
operation of DG systems. Microgrid provides more flexibility

978-1-4244-5287-3/10/$26.00 ©2010 IEEE 2857


j pcc
decoupling requirements. The virtual impedance Y1e j1 Y2e j2 Y3e j3 Yt e jt E pcc e
implementation scheme is based on the fundamental
component control without using any differentiators or high
(low) power filters.

II. SMALL SIGNAL MODEL OF MICROGRID Ypcc e


j pcc

In this section, the operation principle of voltage droop E1e j1 E2e j2 E3e j3 Et e jt

control is presented at first. In addition, the small signal


Fig.1. DG units in a single common bus islanding microgrid
models of a droop controlled microgrid in different operation
modes are developed, which are used to investigate the The grid voltage (PCC voltage), inverter voltage and
system dynamics and stability later. Note that the complex line conductance can be expressed in complex numbers
number matrix based model is flexible and can be easily
as , and . The apparent power can
extended when the microgrid expands with more DG systems
then be expressed in (5) as
included.
· · · (5)
A. Droop control scheme If there is a small disturbance at the equilibrium point,
For voltage controlled DG systems, the flow of real and small signal equations can be shown as
reactive power from DG terminal to utility grid through ∆ω K P · ∆PLPF (6)
distributed line (conductance ) can be expressed in ∆ / ·∆ (7)
(1) and (2) as ∆ 1/ 1 ·∆ (8)
(1) ∆ 1/ 1 ·∆ (9)
(2) ∆ ∆ / ∆ / ∆ (10)
where and are the voltage magnitudes of DG and PCC, ∆ ∆ / ∆ / ∆ (11)
is the phase angle difference between the two voltages, and
where is the time constant of low pass filter, ∆ represents
are the magnitude and phase angle of line conductance
the small disturbance around the equilibrium point, Re and
between the two voltages. When distributed line is mainly
inductive and is very small, the flow of real power is Im denote the real and imaginary part of complex numbers.
proportional to the angle difference , and the reactive power Considering ∆ω s∆θ, the characteristic equation in (12)
flow is proportional to the voltage magnitude difference can be obtained based on (5)-(11). Subsequently, the closed-
( ). Therefore, the droop controller can be adopted for loop dynamic of the system can be evaluated using the
real and reactive power control. However, this conclusion is eigenvalues of (12), which is obtained as shown in (13)
not always valid especially in a LV microgrid due to the · 0 (12)
effects of resistive distribution feeders and presence of local 0 (13)
loads. A new method to address the problems is recently where
proposed [6], where virtual impedances are added at the DG 1 0
,
outputs to make the equivalent impedance inductive. Also the 0 1
voltage magnitude droop control is replaced by PI control in 0
grid connected mode, while the voltage droop slope is ,
0
determined by estimating and compensating the impedance X= ∆ ∆ ,
voltage drops and local load effects in islanding mode to / /
improve the power control and sharing accuracy.
/ /
B. Grid-connected mode model 2
In grid-connected mode, the real and reactive power is .
2
regulated by controlling the DG output voltage frequency and
magnitude with droop and PI controller as shown in (3) and C. Islanding mode model
(4) In intentional islanding operation, the loads in a microgrid
(3) are supported by the DG units, and the power flow from DG
/ 4 units to PCC is controlled by the droop control method. Fig. 1
where and are the measured real power and shows the equivalent circuits of a typical single common bus
reactive power after a first-order low pass filter, and microgrid. The parallel DG interfacing converters are
are reference power, is the droop slope of real power connected to the PCC with distribution feeders, and a lumped
PCC load is considered in this model. All DG units are
controller, and are the parameters of PI controller for
reactive power in grid-connected mode. modeled as ideal voltage sources due to the much higher

2858
control bandwidth associated with the interfacing inverter Table I. Microgrid System parameters
voltage control loop compared to the bandwidth of power Single DG Parameters Values
control loop. In Fig. 1, variables Y e and Y e Existing Line Impedance 1Ω,0.25mH
represent the total loads at PCC ( ) and distribution feeder PCC Line Voltage(RMS) 104V
line conductance (Y ) for DG m, and E e represents the P /Q 500W/300Var
complex number output voltage of DG m. P /Q 250W/150Var
Assuming the number of DG unit in a microgrid is t, PCC PCC Loads 300W/150Var
voltage can be obtained as in (14): DG Frequency 59.5Hz-60.5Hz
∑ · / ∑ (14)
described by equation (21). It is important to note that
If (∑ is replaced by a new symbol , characteristic equations have similar structure in grid
the apparent power generated by DG m can be obtained as connected and islanding mode, and it is very easy to update
S E Y ·e E Y /Y e · the matrix when the number of DG units increases. Local
∑K EK · YK · e (15) loads are not considered here since the offset effects can be
Expanding the expression of apparent power at the compensated with accurate droop control method in [6].
equilibrium point, we can have
∆ ∑ / ∆ ∑ / ∆ (16) III. DG OUTPUT (VIRTUAL) IMPEDANCE DESIGN
∆ ∑ / ∆ ∑ / ∆ (17) The microgrid models developed in the previous section
The detailed expressions of (16) and (17) are attached in are used here for the design of the DG output virtual
the bottom of this page. impedance. The following subsections present the design
For DG unit m, the small signal droop controllers are considerations related to DG power capacity, output real and
shown in (18) and (19) reactive power decoupling, system stability and transient
∆ ·∆ , (18) performances in both grid-connected mode and islanding
∆ ·∆ , (19) operation modes of a microgrid.
where , are the droop gain of real and reactive power
A. DG power capacity and P-Q decoupling
controller, , , , are the measured power. and
To maintain power quality of microgrid loads, the DG
denote the angle frequency and magnitude of voltage.
output voltage magnitude should be confined to certain range
Similarly, the small signal model in islanding mode can be
as E εV (where V is the nominal value of PCC
constructed by matrix equation shown in (20), where the
eigenvalues are determined by (21): voltage, ε is selected as 1.1 in this work,), and angle
· 0 (20) difference between DG voltage and PCC voltage θ should
· 0 (21) be limited at the linear region as θ θ [11] (θ
1 , 0 10 is considered here). To meet the power generation
where , I , is the identity capacity requirements, when E and θ are regulated within
0 1 ,
given range, the maximum real power and reactive power of
matrix,
and ∆Ѳ , ∆Ѳ , … … ∆Ѳ , ∆ , ∆ , … … ∆ . each DG shall be no less than the power demand. Note that
the DG systems work at power generation mode in this
The elements of matrix B s and C s are shown as
discussion. If power flows from PCC to the DG units, the
,1 ;
lower constraints of voltage magnitude and angle difference
, 1 2 ; 0,
can be considered in a similar way.
/ Ѳ ,1 ,1 ; A second consideration for the DG output power is the
/ ,1 , 1 2 ; real and reactive power (P-Q) decoupling. Conventionally,
/ Ѳ , 1 2 ,1 ; the P-Q coupling during transient can be effectively
suppressed when line impedance has a high reactance to
/ , 1 2 , 1 2 . resistance (X/R) ratio. The power decoupling features are
The free motion of islanding operation microgrid with shown in (22) and (23), where k1 is the P-Q decoupling
small disturbances around the equilibrium point can be coefficient.
____________________
/
∂S / ∂θ Y E /Y · e · EK YK · e K K Y E /Y · e
K
/
∂S / ∂θ Y E /Y · e ·E Y ·e
∂S / ∂E 2Y E · e Y /Y · e · EK YK · e Y E /Y · e
K
∂S / ∂E Y Y E /Y · e

2859
3.5 the allowable microgrid voltage range. To address this
Real Power Decoupling Boundary (Eq.23)
3 Real Power Capacity Boundary
conflict, the negative resistive virtual impedance can be
Reactive Power Decoupling Boundary (Eq.22)
implemented as will be discussed in Section IV and the
Line Impedance: Imag(Z)

2.5
experiments. In addition, a relatively lower 1 shall be
2 considered in a LV micorgrid application as the effects of line
1.5
resistance can be compensated with proposed voltage feed-
forward control, which will be discussed latter. In this work,
1
Reactive Power Capacity Boundary
1 1 is considered.
0.5
Finally, considering both the power capacity and power
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
decoupling requirements, the DG output impedance boundary
Line Impedance: Real(Z) plot is shown in Fig. 2, where the lateral axis is the line
Fig.2. Impedance range considering DG power capacity resistance, the vertical axis is line reactance, and the shaded
and P-Q coupling region is the desired impedance range. Other system
40 parameters can be found in Table I.
30
2 B. Stability and transient in grid-connected mode
20
Root Locus (Imaginary Axis)

The small signal model derived in Section II is applied


10
1 4 here to extract the family of root loci of a DG system in grid
0
connected operation. Fig. 3 shows such an example root locus
-10
plot where the DG output resistance is fixed (1Ω) and line
-20 inductance increases from 0.0001mH to 13mH. As shown,
-30
3 there are four root loci, and compared to the conventional
-40
analysis in [12], the 4th root locus is produced by the reactive
-70 -60 -50 -40 -30 -20 -10 0 10
Root Locus (Real Axis) power integral controller. It can be seen that there are two
Fig.3. Root locus with variable inductance in grid- poles located at the right half plane when line inductance is
connected mode (arrow direction: from 0.0001mH to very small, and the system is unstable. When the output
13mH). inductance increases, the two unstable poles (pole2, pole3)
3.5 move to the right half plane with nonzero imaginary part,
3
indicating an oscillatory response of DG output power
Desired Dampling Boundary control. Further increasing of the DG output inductance
2.5
Line Impedance: Imag(Z)

causes the nonzero imaginary poles to the real axis, making


2 the system over damped. On the other hand, the additional
1.5 pole (pole4) located in the real axis goes to the origin of
coordinates when inductance increases, making the system
1

Desired Stability Boundary less stable.


0.5
Critical Stability Boundary
It is not straightforward to give a rigorous evaluation of
0
0 0.5 1 1.5 2 2.5
the 4th order system. Instead, the performance of power
Line Impedance: Real(Z) tacking can be evaluated in terms of dominant poles. E.g. the
Fig.4. Impedance range considering stability and system damping mainly depends on nonzero imaginary
damping requirements in grid-connected mode. conjugate poles, and the system stabilities are guaranteed if
all poles are in the left half plane. Therefore, to maintain
⁄ Ѳ 1 ⁄ Ѳ (22)
properly stability and transient performance, the DG output
⁄ E 1 ⁄ E (23)
impedance needs to meet the following constraints:
A sufficiently high 1 will ensure that the real and
 The real part of all poles should be less than k to
reactive power can be controlled with minimum coupling. In
guarantee good stability performance as
a LV microgrid, the DG P-Q decoupling coefficient is
significantly lower than synchronous generators if there is no Max Real Poles k (24)
additional inductance added at the DG output. To satisfy the  As system damping is mainly related to the nonzero
P-Q decoupling constraints, DG output virtual impedance imaginary conjugate poles, damping ratio can be
should be designed with a high coefficient 1. confined by the following relation
Further considering that in a LV microgrid with nontrivial K Imag Poles /Real Poles K (25)
line resistance, a high P-Q decoupling coefficient leads to a where , and are the parameters of the constraints,
large reactance value, which will cause significant line and they are selected as 7, 0.8 and 1.5 in
voltage drop and may affect the DG power capacity due to this work. Note that (25) is valid only for the nonzero

2860
imaginary conjugate poles (pole2, pole3). With the
50
consideration of stability and transient constraints in grid-
40
connected operation mode, the DG output impedance plot is 30
2

illustrated in Fig. 4, where the desired stability boundary is

Root Locus (Imaginary Axis)


20
related to the constraint in (24) and the desired damping 10
boundary is determined by (25). In Fig. 4, the critical stability 0
1 4 5

boundary is the theoretical stability limit with 0 in -10

(24). -20

-30
C. Stability and transient in intentional islanding mode 3
-40

Similarly, the microgrid dynamics and stability in -50


-100 -80 -60 -40 -20 0 20
islanding operation can be evaluated with the developed Root Locus (Real Axis)

islanding mode small signal model. Fig.5. Root locus with variable inductance in
To simplify the discussion, a simple microgrid which islanding mode (arrow direction: from 0.0001mH to
consists of two identical DG systems and a lumped PCC load 13mH).
is considered. It is important to note that with the developed 3.5

microgrid model in Section II, the analysis can be easily 3

extended to a microgrid with more DG units at different

Line Impedance: Imag(Z)


2.5
ratings. Desired Damping Boundary

Similarly, Fig. 5 shows the family of root loci when the 2

DG output resistance is fixed (1Ω) and inductance increases 1.5

from 0.0001mH to 13mH. It can be seen that the two DG 1


systems have five nonzero eigenvalues and one zero Desired Stability Boundary

eigenvalue, and only the nonzero eignevalues are valid for 0.5
Critical Stability Boundary
stability and dynamic studies [13]. As shown, the system is 0
0 0.5 1 1.5 2 2.5

unstable when the line inductance is too small. On the other Line Impedance: Real(Z)

hand, when inductance keeps increasing, the system becomes Fig.6. Impedance range considering stability and
over damped, and one pole at the left half of real axis is very damping requirements in islanding mode.
close to the origin of coordinates. Therefore, the desired 3.5

impedance range can be derived considering damping 3

response and stability requirements using (24) and (25), 2.5


Line Impedance: Imag(Z)

with K 10, K 0.8 and K 1.5. The range of Desired Overlap Region

desired impedance in islanding mode is shown in Fig.6. 2

Similarly, the critical stability boundary is corresponding to 1.5

(24) with K 0. Note that a tighter stability constraint with 1

smaller K is used in islanding mode, as generally the


0.5
islanding mode has better stability compared to grid-
connected mode (due to the integral reactive power control 0
0 0.5 1 1.5 2 2.5
Line Impedance: Real(Z)
used in grid-connected operation).
Fig.7. Designed impedance range (overlap region
Finally, to ensure optimal operation in both grid-connected
between grid-connected and islanding modes).
mode and islanding mode, all the constraints mentioned
above should be satisfied for the DG output impedance
design. Therefore, by integrating Fig. 2 (power capacity and IV. VOLTAGE CONTROL AND VIRTUAL IMPEDANCE
P-Q decoupling constraints), Fig. 4 (stability and transient IMPLEMENTATION
constraints in grid-connected mode) and Fig. 6 (stability and Once the desired value (range) of the DG output
transient constraints in islanding mode), the desired DG impedance is determined, the next step is to implement the
output impedance is determined as shown in the shaded impedance through the control of the DG interfacing inverter.
region of Fig. 7. For the example DG system as described in To properly implement the virtual impedance control, a few
Table I, the desired output impedance can be approximately factors need to be considered: i) the existing feeder
determined as 1 , 1Ω 1.5 . Note that impedance, ii) the DG output impedance produced by the DG
and include the actual feeder impedance, the DG voltage control, and iii) the virtual impedance control
output impedance introduced by DG voltage controller, and scheme.
the virtual impedance to be discussed in the next section. For a DG unit with nontrivial line resistance, the desired
total output impedance contains three parts:

2861
Table II. Voltage Control Scheme Parameters proposed as shown in (27) and (28):
Parameters Values VV, s RV I s ω LV I s (27)
LC filter Filter Inductor ( / ) L:5mH/R:0.2Ω VV, s RV I s ω LV I s (28)
Filter Capacitor ( ) 40uF where R V and LV are virtual resistance and inductance,
Outer loop Cutoff Frequency 4Rad/s , and , , and are voltage drop and
Proportional gain ( ) 0.22 line current in the stationary frame. Note that in (27)
Resonant gain ( _ ) 25 and (28) can be positive or negative, depending on the
Resonant frequency 0 120 Rad/s existing feeder resistance. As mentioned, in a LV microgrid,
Inner loop Inner Loop Gain 22 it is very possible that the feeder resistance is higher than the
desired range, and therefore negative will be required for
better transient performance.
Fig. 8 shows the overall voltage control system of a DG
interfacing inverter. As shown, the reference voltage comes
from power control loop, and the reference voltage variation
induced by the virtual impedance is incorporated into voltage
reference. For the DG system with output LC filter, a multi-
loop voltage control scheme is adopted. In this voltage
controller, the output filter capacitor voltage is used as outer
loop feedback variable, and a practical form P+Resonant
controller is implemented for outer loop voltage
tracking [6]. For the inner current loop, a proportional
controller is selected for fast tracking dynamics.
Fig.8. Voltage control and virtual impedance implementation
To evaluate the closed-loop dynamic behavior of voltage
8

7
controller, the interfacing converter can be modeled by the
0.08
6
Thevenin equivalent circuit as shown in (29)
Magnitude

5
0.06
4

3
0.04
(29)
2

0
59 60 61
where and are DG output voltage and voltage
50 100 150 200 250 300 350 400 450 500
reference, is DG line current, is the voltage gain.

1
80

60
Phase (Degree)

40

20

0
50
1
0
-20
(30)
-40
-50
-60

-80
59 60 61 where 1⁄ , 1⁄ are the transfer
functions of LC filters.
0 50 100 150 200 250 300 350 400 450 500

Frequency (Hz)

Fig.9. Output impedance of the closed-loop voltage controller. Fig. 9 illustrates the frequency domain response of output
impedance with the voltage control parameters listed in Table
II. The frequency dependent output impedance must be
Z s Z s Z s Z s (26)
carefully considered because it will contribute to total DG
where is the existing feeder impedance, Z s is
output impedance. This is particularly true when considering
output impedance of voltage controller, and is the
that the droop control will cause frequency variations. To
virtual DG impedance to be realized by control. Therefore, if
avoid this complication, the magnitude of desired total
the desired total impedance and existing impedance
impedance shall be 10 times larger than the maximum
(including line and voltage control impedances) are known,
the value of virtual impedance can be determined
accordingly.
To avoid the drawbacks of differential computations
(which amplify high frequency noises) or the using of high
pass filters (which introduce phase shift, delays), a robust
implementation of virtual inductor was proposed based on
fundamental power control [6]. Further considering that the
feeder resistance in a LV microgrid may be higher than the
desired total resistance range illustrated in Fig. 7, a more
general approach to get the virtual impedance voltage drop is Fig.10. Adaptive virtual impedance control scheme during a
reactive step increase.

2862
impedance introduced by the voltage controller within the 160

possible microgrid frequency variation range (59.5-60.5Hz). 140

As shown in Fig. 9, the maximum impedance is about 120

0.055Ω, which indicates that the total DG output impedance 100


P-a P-b
P-c

Power (P:W/Q:Var)
should be higher than 0.55Ω. The desired impedance range 80

60
determined in Section III sufficiently satisfies this Q-a Q-b
40
requirement. Note that with the increase of DG ratings, the 20
Q-c

requirement on the DG output impedance produced by 0

voltage control will be stricter. -20

-40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
V. FURTHER TRANSIENT PERFORMANCE IMPROVEMENT Time (Sec)

Fig.11. Transient response to real power step


A. Further transient P-Q decoupling algorithm increase. (Subscript: a - proposed method, b - with
Although a dominant DG output inductance can suppress negative resistance, c - conventional method)
the P-Q coupling, a transient change of real power flow may 180
P-a
still cause a considerable reactive power disturbance. This is 160

because the real power flow change will introduce a change 140 P-b

of the impedance voltage drop, and this coupling voltage drop 120

Power (P:W/Q:Var)
has a direct impact on the reactive power control. Therefore, 100

a better way to prevent P-Q coupling is to compensate the 80


Q-a

coupling voltage drop caused by real power flow. 60


Q-b
During islanding operation, the DG output reactive power 40

is determined by the load demands. Therefore, the dynamics 20

of reactive power tracking is not a real concern in islanding 0


0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
mode, and steady-state sharing accuracy can be improved if Time (Sec)

this coupling voltage drop due to steady state real power flow Fig.12. Transient response to a reactive power step
can be compensated [6]. To the contrary, in grid-connected increase. (Subscript: a – with adaptive impedance
control; b - conventional method)
operation, the PCC voltage is stiff and therefore any
additional voltage drop due to real power flow will affect the 700

transient reactive power output of a DG unit. The integral 600


Virtual Impedance Control Disabled
control will improve the reactive power control accuracy in 500

steady state, but it does not help much for the transient due to 400
Power (P:W/Q:Var)

the P-Q coupling. For this reason, the P-Q coupling is 300
P-a P-b
200
especially serious in grid-connected mode. 100
Based on previous considerations, a better way to suppress 0

power coupling through the combination of virtual -100


Q-a Q-b

impedance method and real power feed-forward control is -200

proposed for the DG voltage control to further improve the 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85
Time (Sec)
transient performance of a DG unit in grid-connected mode.
The expression of voltage magnitude control is derived as Fig.13.Adaptive virtual impedance variation during
transient.
∆VDG K E K I /s ∆Q ∆Q ⁄V ∆P (31)
By using (31), the DG output voltage is adjusted by both scheme corresponding to a step increase of reactive power
the reactive power PI controller and the transient real power reference. As shown, the virtual impedance is kept as a
feed-forward control. constant in steady state. When there is a step increase of
reactive power reference, the reactive power feedback is
B. Adaptive transient virtual impedance
compared with reference value to produce the power
The proposed adaptive virtual impedance method provides difference. If the power difference ∆ is larger than a
one more tunable variable for DG output power control, threshold 2 , which means an output power transient is
which can improve the reactive power control performance anticipated, and the virtual inductance is decreased
during a transient. When the reactive power stabilizes after a automatically using an integrator. This process is shown as
transient, the virtual impedance is gradually controlled back the “adaptive transient impedance” region in Fig. 10. As
to the originally designed value. Therefore, the static mentioned, the adaptive impedance is valid only during a
damping and stability of DGs are maintained. transient. When the transient completes, the virtual
Fig. 10 shows the adaptive virtual impedance control

2863
impedance need to be restored to the original value. This can addition, a transient power decoupling algorithm and an
be done by detecting the reference-feedback power adaptive transient virtual impedance control method are
difference, and if it is smaller than a threshold 1 , another proposed to further improve the transient power control
integrator will slowly increase the virtual impedance back to performance. Experiments are conducted to prove the
the original value. If the upper limit of saturation block proposed design, implementation and control strategies.
shown is set to zero, the virtual impedance will be restored to
the original value smoothly. REFERENCE
Note that during this adaptive impedance control in a [1] J. M. Carrasco, L. G. Franquelo, J. T. Bialasiewwicz, E.Galvan, R. C.
transient, the desired total X/R ratio is still preferred to be P. Guisado, M. A. M. Prats, J. I. Leon, and N. Moreno-Alfonso,
constant to minimize the P-Q coupling. Therefore, the virtual “Power-Electronic systems for the grid integration of renewable energy
sources: A survey,” IEEE Trans. Power Electron., vol. 53, no. 4, pp.
resistance should also be adjusted accordingly. 1002-1016, Aug. 2006.
[2] IEEE Standard for Interconnecting Distributed Resources with Electric
VI. EXPERIMENTAL VERIFICATION Power Systems, IEEE Std. 1547, 2003.
[3] Y. W. Li, D. M. Vilathgamuwa and P. C. Loh, “Design, analysis and
To experimentally verify the proposed approach, a real-time testing of a controller for multibus microgrid System,” IEEE
hardware microgrid with two identical DG units is employed. Trans. Power Electron., vol. 19, pp. 1195-1204, Sep. 2004.
System parameters are selected to be the same as in Table I. [4] F. Z. Peng, Y. W. Li, and L. M. Tolberta, “Control and protection of
power electronics interfaced distributed generation system in a
Fig.11 shows the dynamic response of a DG unit in grid- customer-driven microgrid,” in Proc. IEEE Power & Energy Society
connected mode when the real power reference is changed General Meeting. pp. 1-8, 2009.
from 40W to 140W. The reactive power (Q-c) is affected by [5] T.Kawabata, S. Higashino, ‘‘Parallel operation of voltage source
inverters,” IEEE Trans on Industry Applications. vol. 24, pp.281-287,
the real power transient due to power coupling. As expected, Mar/Apr. 1988.
the negative resistance can increase the X/R ratio and [6] Y.W.Li, ChingNan Kao, “An accurate power control strategy for
therefore mitigate the power coupling, which leads to a less power-electronics-interfaced distributed generation units operation in a
affected reactive power (Q-b). Finally, a good reactive power low voltage multibus microgrid,” IEEE Trans. Power Electron., vol.
24, pp. 2977-2988, Dec. 2009.
disturbance rejection (Q-c) is achieved when the proposed [7] Y.Mohamed and E.Saadany, “Adaptive decentralized droop controller
transient power decoupling algorithm is implemented. Note to preserve power sharing stability of parallel inverters in distributed
that the real power control is not affected during transient. generation microgrids,” IEEE Trans. Power Electron., vol. 23, no. 6,
pp. 2806-2816, Nov. 2008.
The proposed adaptive transient impedance control is also
[8] K. D. Brabandere, B. Bolsens, J. V. D. Keybus, A. Woyte, J. Driesen,
tested in the experiments and the results are shown in Figs. 12 and R. Belmans, “A Voltage and Frequency Droop Control Method for
It can be seen that with a step up of reactive power reference, Parallel Inverters,” IEEE Trans. Power Electron., vol. 22, pp. 1107-
the dynamic response of reactive power (Q-a) is significantly 1115, Jul. 2007.
[9] A. Tuladhar, H. Jin, T. Unger, and K. Mauch, “Control of parallel
improved with the adaptive virtual impedance control. inverters in distributed AC power system with consideration of line
Moreover, the real power control is not deteriorated impedance effect,” IEEE Trans on Industry Applications. vol. 36, no. 1,
compared to the conventional method. pp.131-138, Jan/Feb. 2000.
Finally, the microgrid stability in islanding operation is [10] J. M. Guerrero, L. G. Vicuna, J. Matas, M. Castilla, and J. Miret,
“Output impedance design of parallel-connected UPS inverters with
evaluated experimentally. In this test, both DG units are wireless load sharing control,” IEEE Trans on Industrial Electronics,
controlled with identical virtual inductance of 3mH initially. vol. 52, no. 4, pp. 1126-1135, Aug. 2005.
At t=0.7sec, the virtual impedance method is disabled. Fig.21 [11] J. M. Guerrero, L. G. Vicuna, J. Matas, M. Castilla, and J. Miret, “A
wireless controller to enhance dynamic performance of parallel
shows the real and reactive power flow of DG1 (P-a/Q-a) and inverters in distributed generation systems,” IEEE Trans. Power
DG2 (P-b/Q-b). It can be seen that the system stability is Electron., vol. 19, no. 4, pp. 1205-1213, Sep, 2004.
maintained with virtual inductance control. Once the virtual [12] E. A. A. Coelho, P. C. Cortizo, and P. F. D. Garcia, ”Small signal
impedance control is disabled, power oscillations occur and stability for single phase inverter connected to stiff ac system, ” in
Conf. Rec. IEEE-IAS Annu. Meeting, Vol.4, Oct. 1999, pp. 2180-2187.
the system becomes unstable. [13] E. A. A. Coelho, P. C. Cortizo, and P. F. D. Garcia,“Small-signal
stability for parallel-connected inverters in stand-alone AC supply
VII. CONCLUSION systems, ” IEEE Trans on Industry Applications, vol. 38, no. 2, pp.
533-542, Mar/Apr. 2002.
In this paper, flexible complex number based small-signal [14] D. N. Zmood and D. G. Holmes, “Stationary frame current regulation
models are first developed to evaluate the stability and of PWM inverters with zero steady-state error,” IEEE Trans. Power
dynamics of a LV microgrid. Based on the developed model, Electron., vol. 18, pp. 814-822, May, 2003.
[15] Corradini. L, Mattavelli. P, Corradin. M, and Polo, F, “Analysis of
together with the DG power capacity and P-Q decoupling parallel operation of uninterruptible power supplies though long wiring
requirements, the desired DG output impedance range can be cables,” IEEE Trans. Power Electron., vol. 25, no. 4, pp. 2806-2816,
designed. In a LV microgrid with resistive line feeders, the Apr. 2010.
virtual impedance method can effectively adjust the total [16] J. M. Guerrero, L. Hang, and J. Uceda, “Control of distributed
uninterruptible power supply systems,” IEEE Trans on Industrial
impedance to be within the designed range, and the microgrid Electronics, vol. 55, no. 8, pp. 2845-2859, Aug. 2008.
stability and dynamics can be improved accordingly. In

2864

You might also like