Modeling and Design of Virtual inertia based

Rooftop PV
Deepak Pullaguram, Student Member, IEEE, Sukumar Mishra, Senior Member, IEEE, and
Nilanjan Senroy, Member, IEEE

Abstract—With ever increasing Photovoltaic (PV) penetration Furthermore, the above mentioned control scheme lack proper
in distribution level, they expected to respond for voltage and fre- synchronising capability, which causes high initial transients
quency deviations. It is also necessary to provide an uninterrupted during grid connection.
supply to the local loads during grid interruption. In this paper,
virtual inertia based single phase rooftop PV which responds to In this paper a modified virtual inertia control scheme
changes in the grid state, grid voltage and frequency is described, for single phase rooftop PV with the following features is
using instantaneous real and reactive power components. As the developed.
instantaneous real and reactive power components are used in
controller, it is embedded with power limiters to avoid over • Seamless transition from grid connected to the isolated
currents and to safe guard inverter switches. The Small signal mode.
model analysis is carried out for the system developed to observe • Supports both grid connected and isolated mode opera-
its stability under both isolated and grid connected modes. The tion.
performance of PV-battery system with proposed controller is
• Contributes as an ancillary service by supporting fre-
tested by considering various operating points and perturbations
in digital simulation environment. quency and voltage.
Index Terms—Linearised system model, power system, rooftop A linearised small signal model for the proposed system is
photovoltaic, virtual inertia
developed. Eigen analysis is carried out for both grid connected
and isolated modes, and important conclusions are drawn
I. I NTRODUCTION
regarding the stability of the system.
In many countries worldwide, most customers and utilities The rest of the paper is organized as follows, Section II
use battery based inverters to improve the reliability of their provides the modelling and control of rooftop PV system.
local power requirements. Considerable incentives provided by Section III describes the controller transition between islanded
renewable energy promotion policies in the form of feed-in and grid-connected modes along with its reference set points.
tariff, net metering, etc. [1], have encouraged many customers The small signal analysis of the system is carried out in Section
to install rooftop solar PV panels. These PV’s along with IV. Section V. Section VI concludes the paper.
batteries are operated to meet the relatively small local demand
and sell the excess power to the grid. Even though single II. ROOFTOP PV SCHEMATIC

phase renewable generation looks promising because of better The complete schematic of the rooftop PV is shown in
reliability and economic benefits for local customers, there Figure 1. The PV panel, which is augmented with a DC-
are issues from the network perspective mainly in protection, DC boost converter, is operated at maximum power point
stability and power quality. The impact of these issues will be (MPP) using the incremental conductance algorithm. The MPP
low if the renewable energy penetration is less. But with the tracking algorithm feeds the duty reference corresponding to
increase in PV penetration, the equivalent inertia of the system, the maximum power point voltage Vpv to the PWM genera-
especially in microgrids and islanded networks could be re- tor through which the boost converter is fired. An H-bridge
duced, which deteriorates its dynamic performance. The quality inverter, controlled by a modified virtual inertia controller, is
of power may also degrade due to the stochastic nature of sin- connected to the point of common coupling (PCC) through an
gle phase power injections. In order to have acceptable quality LCL filter circuit. The low pass LCL filter is designed such that
of power supply, it is expected that the network connected PV it limits the injection of excessive current harmonics, which are
systems respond to frequency and voltage deviations. In this generated due to the SPWM, into the network.
context, various droop based control architectures responding
to the frequency and voltage deviations have been reported
[2], [3]. But these control methodologies may not always Ib PWM
Bi-
Directional
Vdc ref
Vdc DPf
controller Ib f cal Synchronization
provide adequate inertial response. Further they are designed Vb
grid mode
& isolation detection

to operate either in grid connected or in isolated operation but + - H- bridge

vPCC
Lse Rf Lf Rc Lc
not for both. Control schemes that mimic the characteristics of I pv ii io PCC CB
Local Load

a synchronous generator using a virtual inertia concept have Cdc Vdc vi Cf vo

C pv V pv
been proposed for a three phase inverter system [4] and further PWM

extended to single phase systems for controlling power output

from energy storage system [5], [6]. However, these controllers PWM
Preflim
w wt vo io
may fail to limit the current flowing through the inverter during MPPT Based
vo
Qe
Current based Virtual inertia Po
SOGI based
Power
power limiter controller
network disturbances, and may damage the inverter switches. Controller Pe
lim
Qref
Qo
calculation

Pref Qref Fnom Vref

The authors are with the Department of Electrical Engineering, Indian
Institute of Technology, Delhi India e-mail: [email protected]. Fig. 1: Single phase rooftop PV
978-1-5386-1138-8/18/$31.00 c 2018 IEEE
A. Modelling of Virtual Inertia Controller B. Second order generalized integrator (SOGI)
The basic active power loop of the virtual inertia controller The average instantaneous power delivered to the system
is based on the swing equations of an imaginary synchronous
can be calculated by using dq frame voltage and current
machine [4], [6], as given by (1) and (2). components as given by (3) and (6). In order to obtain dq
dω Pref Po 1 transformation atleast two symmetrical signals are required,
J = − − D∆ω+ (ω0 − ω) (1) which may not be available for a single phase system. Hence,
dt ω0 ω0 2πRω0
a SOGI is used to obtain an additional orthogonal component
dθ along with the in-phase component [7]. Accordingly, the αβ
= ω0 − ω (2)
dt component of the single phase system are obtained as
where J, D are the imaginary machine inertia and damping v′ kωs
Hd (s) = = 2 (8)
coefficient, ω is the inverter operating frequency, R is the gain v s + kωs + ω 2
of the frequency feedback path similar to the frequency droop qv ′ kω 2
Hq (s) = = 2 (9)
constant used in a synchronous generators. Pref is the power v s + kωs + ω 2
reference input and Po is the real power output of the inverter, Here k effects the bandwidth of the SOGI system.
calculated using (3) The output time response of the SOGI for the system with
1 input vi = Vi sin(ω0 t) is given by
Po = (Vod Iod + Voq Ioq ) (3)
2 kωω0
v ′ (t) =Vi p sin(ω0 t − γ)
Here Vod , Voq , Iod , Ioq are the dq components of the voltage k ω ω0 + (ω 2 − ω02 )2
2 2 2
(10)
across the inverter filter capacitor and output current after the p −kωt
− Vi kα sin( (1 − k/2)2 ωt − ψ1 )e 2
capacitor, Cf , respectively.
The reactive power output of the inverter is controlled as per
the voltage deviation (at the PCC) to realize the voltage droop kω 2
qv ′ (t) =Vi p cos(ω0 t − γ)
characteristics as given in (4) and (5). k 2 ω 2 ω02 + (ω 2 − ω02 )2 (11)
p −kωt

Vi = (kpv (Vref − Vm ) − Qo )(kp + ki φi ) (4) − Vi kβ sin( (1 − k/2)2 ωt − ψ2 )e 2

dφi where
= (kpv (Vref − Vm ) − Qo ) (5)
dt q
2
4(4 − k 2 )(ω02 − ω 2 )2 ω 2 + 2kω(ω02 + ω 2 )
!
kω0
where kpv is voltage droop constant, kp , ki are proportional kα = p
and integral gains of the PI controller, and Vm is the voltage at k 2 ω 4 (4 − k 2 ) + (k 2 ω 2 − 2ω 2 + 2ω02 )2 1 − (k/2)2
the inverter output (given by (7)). The reactive power is thus
calculated using (6) 2kω0 ω
1 kβ = p
Qo = (Voq Iod − Vod Ioq ) (6) k 2 ω 4 (4 − k2 ) + (k 2 ω 2 − 2ω 2 + 2ω02 )2 1 − (k/2)2
2 " #
q
−1 ω02 − ω 2
Vm = 2 +V2
Vod oq (7) γ = sin p
k 2 ω 2 ω02 + (ω 2 − ω02 )2
"√ #
Using (1)-(7) the basic virtual inertia based control is devel- 4 − k 2 (ω 2 − ω 2 )
0
oped as shown in Figure.2. ψ1 = tan−1
k(ω02 − ω 2 ) + 2kω 2
" #
2 2 2 2
fcal −1 k ω − 2ω + 2ω0
CB status Isolation ψ2 = tan p
vo detection
ENB
2kω 2 1 − (k/2)2
gridmode ENB
-
e
-VPCC max

+
PCC
CB
Under frequency locked condition (typically during grid-
d 2 Df s
1
X
k ss
s
d Df s
+ == 0 connected mode), ω = ω0 . Substituting this in (10),(11) and
dt dt 2
Isolated mode PCC CB control applying Park’s transformation [8], we get
Synchronization
X kif

!p

s Vid (t) = Vi cos(ω0 t − ωt) − Vi sin 1 − (k/2)2 ω0 t sin ωt
+ f ref
- Df
1/ R X s  (−kω0 t)/2
+
fcal
- +
+ !p
2

e
f nom
− cos 1 − (k/2) ω0 t − ψ1 cos(ωt) p
DPf
1
Droop mode
1 − (k/2)2
2p
+ Current
+ 1 1
CB status (12)
Pref based
÷ mod sin
wt 
+ power
-
Js+B s !p

limiter 0
¸
2
vi
Viq (t) = Vi sin(ω0 t − ωt) − Vi sin 1 − (k/2)2 ω0 t cos ωt
Pe
X PWM
0  (−kω0 t)/2
Pvirtual CB status
Active Power loop !p
e
Qe
− cos 1 − (k/2)2 ω0 t − ψ2 sin(ωt) p
-
Virtual inertia
control
1 − (k/2)2
Current based Vi
Vref
+
k pv power limiter +
PI (13)
-

Vm Reactive Power loop When grid considered as the reference, ω0 t − ωt = θ. Thus for
a small disturbance, cos θ ≈ 1 and sin θ ≈ θ. Moreover, the
Fig. 2: Modified virtual inertia Control schematic exponential term in the above expression decays to zero with
kω0 /2 time constant. Hence Vid and Viq may be approximated low inertia constant in order to improve its response time.
(in frequency domain) as [9] The ROCOF would be of high magnitude and may cross the
threshold for any network disturbance other than grid failure,
Vid = Vi ; Viq = V θ̂; (14)
thus leading to mal-operation. The ROCOF sign may be used
kω0 /2 instead of a threshold value to avoid such a scenario. Whenever
θ̂ = θ
s + kω0 /2 grid fails the electrical load Po on the inverter will be either
increased or decreased, with Pref being constant. Accordingly
The equation for θ̂ may be written in differential form as
the controller will sense either a monotonic rise or fall of the
dθ̂ frequency. The sign of the ROCOF may be monitored for 20
= (θ − θ̂)kω0 /2; (15)
dt to 50 electrical cycles before triggering the controller. During
C. AC Filter circuit this time if the ROCOF sign is either +ve or −ve for all the
cycles then it is treated as LoM and the CB tripped.
The inverter filter circuit is considered in dq domain assum-
ing that the inverter produces the commanded voltage vi [10].
ROCOF ≤ 0 or ROCOF ≥ 0 (22)
Equations corresponding to the LCL-filter circuit is given as
dIidq Rf 1 1 The IEEE 1547 standards [11] for the LOM detection with
=− Iidq ± ωIiqd + Vidq − Vodq (16) in 2s is fulfilled by the proposed technique. Further, over-
dt Lf Lf Lf
dVodq Gf 1 1 voltage and under-voltage relays may be used as backup, in
=− Vod ± ωVoqd + Iidq − Iodq (17) case the system has large number of impedance loads where
dt Cf Cf Cf
the voltages are effected and the CB tripped based on the
dIodq Rc 1 1
= − Iodq ± ωIoqd + Vodq − Vgdq (18) relay operation. Once the LoM is detected, the PCC CB is
dt Lc Lc Lc opened using the logic given in ’PCC CB control’ block in
III. S EAMLESS TRANSITION OF CONTROLLER Figure 2. Subsequently, the controller will operate in isolated
A. Synchronization mode supplying its local load and the frequency of the system
is maintained by the integral controller with gain kif in the
As described earlier in Section I, smooth operation en-
frequency loop which gets activated when e−Vpcc > 0, as
tails minimum transients during the transition between grid
shown in isolated mode block in Figure 2. With the addition
connected and islanded modes. In order to achieve this, a
of secondary frequency control, the active power dynamic
virtual ’power’ signal is used as a feedback to the controller
expression of (1) is modified as
and made it zero by setting Pref = 0. The virtual signal
is calculated from (3) using the virtual current obtained by
following expression. dω Pref Po 1 kif
J = − − D∆ω+ (ω0 − ω) + θ (23)
dt ω0 ω0 2πRω0 2πω0
vi − vpcc
ivirtual = (19) where kif is the secondary frequency control gain used to
Zvirtual
where Zvirtual is a mathematical model of the line impedance maintain the local frequency isochronously. Further, when the
and vpcc is instantaneous voltage at the PCC. grid is not available, the equation (18) is modified as

Zvirtual = R + jXL (20) dIod Rc 1 1

= − Iod + ωIoq + Vod − (Rl Iod − ωLl Ioq )
dt Lc Lc Lc
where the values of R and XL may be assumed close to dIoq Rc 1 1
the actual resistance and reactance of the line connecting the = − Ioq − ωIod + Voq − (Rl Ioq + ωLl Iod )
dt Lc Lc Lc
inverter output terminal to the PCC. (24)
An additional synchronizing loop is appended to the existing
where Rl , Ll are the local load resistance and inductance.
frequency loop as shown in Figure. 2 to synchronise the
inverter to grid. Optimal selection of the integral gain ks s C. Control loop reference set points
can speed up the synchronization process. The synchroniza-
tion process and the grid connected operation is dictated by The controller has both frequency and voltage loops which
the gridmode switch, the PCC circuit breaker (CB) status are controlled individually using the active and reactive power
CBstatus and the PCC point r.m.s voltage measurement i.e, set points respectively. Usually at the LV distribution level,
(1 − e−Vpcc ) > 0. The PCC CB is switched ON when the it is expected that the renewable generation should operate
inverter is synchronized with the grid, which is confirmed by close to unity power factor, so that the maximum utilization
(21). of the inverter can be achieved in the form of active power.
d∆fs d2 ∆fs At the same time in such systems at low voltage levels, the
+ == 0 (21) line currents are high and there will be large voltage drops in
dt dt2
the lines because of high line impedance which may further
B. Islanding detection and operation reduce the voltage at different load points. Hence, occasionally
Whenever there is a grid failure, it may be required to the PV system may be operated at lower power factor [11] as
identify loss of mains (LoM) and isolate the rooftop PV for per the grid operating codes for voltage support. Considering
safety. As per IEEE standard 1547, inverters must be able to these factors, in grid connected mode the total KVA rating of
detect the LoM and isolate themselves by using appropriate the inverter is shared such that 98% loading will be active
islanding schemes [11]. The existing schemes like rate of power and 20% will be reactive power. Here, a 20% of band
change of frequency (ROCOF) [12] may not be suitable for is selected for reactive power, without compromising much in
the controller proposed here due to the selection of a very inverter active power carrying capacity.
 Further, the current flowing through the inverter should be The parameters and the initial operating conditions of the
limited, as over currents may damage the inverter. Thus, the system are tabulated in Table A2 of the appendix. The eigen-
set-points to the controller are chosen using the current based values along with the state participation under these conditions
power limiters satisfying above mentioned constraints. The for both grid connected and isolated mode are tabulated in
instantaneous active and reactive current components [13] are Table. I
dynamically calculated using
TABLE I: Eigenvalues and participation factor of each state
√ √
2po 2Qinv variable
iP = ; iQ = (25)
Vm Vm Grid connected mode Isolated mode
Eigen Eigen Participation factor Eigen Eigen Participation factor
where Qinv = 1/2(Voq Iid − Vod Iiq ) name value name value
λ1 −164.0 θ̂(92.03%), ω(4.18%), λ1 0 θi (100.0%)
As the current flowing through the inverter il = io − ic , is to θ(3.8%)
λ2 −700.0 Vi (100.0%) λ2 −7.79 ω(92.29%), θ(07.71%),
be restricted, the active and reactive current limits are obtained λ3,4 −0.497± ω(44.53%), θ(45.73%), λ3 −157.0 θ̂(100.0%)
as 34.3i θ̂(9.74%)
√ λ5,6 −23.0± iid (25.85%),
578.0i
iiq (25.8%),
iod (24.15%), ioq (24.15%),
λ4 −700.0 Vi (100.0%)

limit 2 ∗ 0.98 ∗ KV Arating λ7,8 −214.0± vod (25.65%), voq (23.95%), λ5,6 −6.6e3± iid (16.35%), iiq (16.42%),
iP = 1.5e4i iid (12.18%), iiq (11.68%), 495.0i iod (27.27%), ioq (27.12%),
Vm iod (13.37%), ioq (13.17%) vod (06.42%), voq (06.93%)
√ (26) λ9,10 −790.0± vod (24.38%), voq (25.97%), λ7,8 −2.9e3± vod (25.47%), voq (24.11%),
limit 2 ∗ 0.20 ∗ KV Arating 1.5e4i iid (11.99%), iiq (12.49%), 1.3e4i iid (14.47%), iiq (13.54%),
iQ = iod (12.49%), ioq (12.69%) iod (11.17%), ioq (11.25%)
Vm λ9,10 −3.4e3± vod (24.51%), voq (25.87%),
1.4e4i iid (13.36%), iiq (14.13%),
The reference set-point to the controller, which restricts the iod (10.98%), ioq (11.15%)

power flowing through the inverter and thereby limiting the

currents, [14] are given by
There is a slow mode of oscillation with natural frequency
( limit
ωn = 35.5 rad/s and damping ratio ζ = 0.014 in the grid
limit Pref ∗ eκ(|iP |−|iP |) if |iP | ≥ |ilimit
P |. connected system as observed from the eigenvalue λ3,4 . From
Pref =
1 if |iP | < |ilimit
P |. participation factor it is noted that the influence of the state
( limit
(27) variables ω, θ on this mode of oscillations is significant. Hence,
Qref ∗ eκ(|iQ |−|iQ |) if |iQ | ≥ |ilimit |. the active power loop parameters J and R (Eqn. (1)) tuning
Qlimit
ref = Q
1 if |iQ | < |ilimit
Q |. is critical for the stability of this mode. The other modes
obtained from the eigenvalues λ5,6 , λ7,8 , λ9,10 have relatively
where κ decides the exponential decay slope.
high damping factor.
In the isolated mode, the operating power factor (PF) is
because of the load PF. Hence there is no restriction for the In the isolated system, the low frequency modes are critically
P and Q sharing although the current should be limited. Thus, damped with the addition of feedback integral controller ki f .
for the isolated mode the the set-points are given by Due to lack of a reference angle, there is an eigen value λ1 = 0.

V. S IMULATION R ESULTS
( limit
limit Pref ∗ eκ(|iS |−|iS |) if |iS | ≥ |irating
S |. AND
Pref =
1 if |iS | < |irating
S |. The performance of the proposed controller, was tested by
( limit
(28) simulating the test system shown in Figure 1 using Simulink/-
limit Qref ∗ eκ(|iS |−|iS |) if |iS | ≥ |irating
S |. MATLAB. A rooftop PV with MPPT tracking was simulated
Qref =
1 if |iS | < |irating
S |. in conjunction with a 160V , 10Ah Ni-Cd battery which was
connected to a single phase 230V /50Hz grid with a short
IV. S MALL SIGNAL A NALYSIS
circuit capacity of 1.48M V A. The PV source had a rating
The complete state space model of the system is obtained by of 3.1kW at a maximum operating DC voltage of 400V and
combining the dynamic equations of the controller, SOGI-dq current of 7.75A. The switches of the dc/dc boost converter
transformation, and inverter filter. The system dynamic model were chosen according to this current carrying capacity. The
consists of ten differential equations given by (1),(2),(5), (15), switching frequency of the inverter was selected as 5 kHz.
(16)-(18) and a single algebraic equation given by (4). This
A. Non-Ancillary mode
can be expressed in the form of
This subsection presents simulation results of the PV system
dx
= f(x, z, u); 0 = g(x, z, u) (29) during transition between grid-connected and islanded modes.
dt Duration Event
where t = 1.0s Grid is ON (synchronization begins)
x = [ω, θ, φ, θ̂, Iidq , Vodq , Iodq ] is state variable vector. t = 3.8s Grid connected
z = [Vi ] is algebraic variable vector t = 8.0s Pref change from 0W to 2000W
u = [Pref , Vref ] is input vector. t = 13.0s Grid is tripped
The state space model has to be initialised by solving DAE During this operation the grid voltage and frequency are
model with all derivatives set to zero [15]. The linearised state maintained at the nominal values. The simulation commences
matrix is obtained by neglecting the higher order terms of in the isolated mode operation with inverter supplying a local
Taylor series expansion for Eq.(29), around an initial operating load of 1000 W . At time t = 1s the grid is turned ON and
point (x0 , u0 ). synchronization begins at this instant as shown in Figure 3.
"  −1 #
∂f ∂f ∂g ∂g The oscillations observed at t = 1s in the Figure 3a are the
A= − (30) pre-synchronizing oscillations caused by controller, which does
∂x ∂z ∂z ∂x
x0 ,u0 not impact the PV inverter power output supplied to the local
 52 50.05 P inv Duration Event
4000
49.95
P ref t = 0.0s Grid is ON (synchronization begins)

Active Power (W)

Frequency (Hz)

50 3000 t = 3.1s Grid connected

2000 1040
t = 6.0s Ancillary mode is ON
48 960 t = 8.0s Grid frequency changed from 50.5 to 49.5Hz
f cal 1000 t = 10.0s Pref change from 0W to 2000W
f PCC
46 0
t = 15.0s Grid is tripped
0 2 4 6 8 10 12 14 16 18 0 2 4 6 8 10 12 14 16 18 The ancillary mode is selected at t = 6s to allow the inverter
Time (s) Time (s)
(a) Frequency (b) Active power output
54 5000
P inv
800 4000
P ref

Active Power (W)

Qe
230 52

Frequency (Hz)
Reactive Power (VAr)
Qref
600 3000
225
Voltage (V)

50
400 2000
220
200 48
f cal 1000
215 f PCC
Vo 0 0
46
V PCC 2 4 6 8 10 12 14 16 18 0 2 4 6 8 10 12 14 16 18
210 -200 Time (s) Time (s)
0 2 4 6 8 10 12 14 16 18 0 2 4 6 8 10 12 14 16 18
Time (s) Time (s) (a) Frequency (b) Active power output
(c) Output voltage (d) Reactive power output
250 600
Fig. 3: Non- Ancillary operation Qe

Reactive Power (VAr)

240 400 Qref

load, as depicted in Figure 3b. Further the frequency of these

Voltage (V)
200
230
oscillations is less as compared to the oscillations occurring 0
220
at t = 8s - 10s because of the additional integrator with -200

gain kss , in synchronizing unit. The synchronization process 210 VO

-400
V PCC
completes around t = 3.8s and connection to the grid is 200 -600
0 2 4 6 8 10 12 14 16 18 0 2 4 6 8 10 12 14 16 18
accomplished by switching PCC CB in Figure 1. Although the Time (s) Time (s)
time to synchronize time is relatively high, it still conforms to (c) Output voltage (d) Reactive power output
the IEEE 1547 standards [11]. Very small frequency and power
Fig. 4: Ancillary mode operation
oscillations are observed within the band of 0.01Hz and 40W
respectively, as shown in Figure 3a, 3b, which is in acceptable to respond to the frequency deviation. The inverter starts to
range as per the standards given in [11]. supply power as per the droop in the frequency controller
The power command Pref is altered to 2000W at t = 8s chosen. With 0.1% droop, the active power supplied through
which leads to power oscillations as shown in Figure 3b. These the inverter reduces by 500W for a 0.5Hz increase in
oscillations correspond to the low frequency mode, λ34 , in frequency. Accordingly, the shortfall in the active power
Table ??. Although the oscillations are significant they damp delivered to the loads beyond the PCC point is made up by
out within 1s because of the controller damping provided by drawing more active power from the grid, thus leading to
virtual inertia and droop constants. When the LoM event takes increased line currents and a reduced voltage at the PCC. The
place at t = 13s, it is observed that the frequency steadily voltage loop supplies additional reactive power in response to
decreases, as the load connected beyond the PCC is greater PCC voltage drop as shown in Figure 4c, 4d.
than the Pref . The slope of frequency is continuously tracked At t = 8s the frequency of the grid is reduced by 1Hz which
using ROCOF for 25 cycles. As there is no change in the leads to increased active power supply to the grid by 1000W
ROCOF sign, it indicates the grid failure event, the PCC CB is as shown in Figure 4b. When the inverter power set point is
tripped and the inverter is operated in the isolated mode. Once changed to 2000W at t = 10s, the output settles to around
the PV system is isolated, it settles down smoothly to a steady 2500W due to the continued frequency dip. Subsequently,
state without any low frequency oscillations indicating that the changes in the PCC voltage cause the inverter reactive power
system is critically damped, as analysed in Section IV through output to change again.
eigenvalue analysis. In this case study, the reactive power is The grid voltage is again increased to 240V at t = 12s
varied as per the terminal voltage Vref deviation based on the which causes the inverter to consume reactive power from the
droop constant kpv as illustrated in Figure 3c, 3d. grid as per the droop value of 70 V AR/V , as shown in Figure
B. Ancillary mode operation 4c, 4d. The grid mains is turned OFF at t = 15s; this time with
In this subsection, the ability of the controller to respond the connected load beyond PCC less than the inverter power
to deviations in the PCC voltage and frequency is presented reference. This makes the frequency rise monotonically for the
through simulations. next 25 cycles triggering the islanding detection, which finally
The grid voltage and frequency are perturbed and the operates to switch OFF the PCC CB.
response of the controller analysed. The grid is ON at t = 0s
with grid frequency 50.5Hz, with a local load of 1000W
connected to the inverter terminals. The synchronization C. Fault and induction machine load operation
process starts at t = 0s, with the steady state grid frequency
being other than the nominal value. The synchronizing unit In this subsection, the interaction of the PV system and an
tracks the grid frequency within 3.1s as depicted in Figure 4. induction machine load is studies under fault conditions.
 54 3000
current does not exceed its limits. The overall system stability
is verified by eigen value analysis, and the results are related

Active Power (W)

2000
Frequency (Hz)
52
to nonlinear time-domain simulations at different operating
50 1000
conditions.
0 A PPENDIX A
48
f cal
f PCC
TABLE A2: Specifications and initial conditions
-1000
46 0 2 4 6 8 10 12 14 16 18
0 2 4 6 8 10 12 14 16 18
Time (s) Controller Specifications Grid connected mode Isolated
Time (s) mode
(a) Frequency (b) Active power output Parameter value Parameter value value
Inertia J 0.01kgm2 ω 314.1593 314.1593
Droop constant R 0.1% θ 0.0546rad 0.0309rad
Frequency integral gain 100 φ 0.000 0.000
3000 kif
220 0.3 Vi 325.6430V 324.1670V
Reactive Power (VAr)
synchronizing integrator
gain kss
Voltage (V)

2000 voltage loop PI kp , ki 0.0001, 0.0007 Vidq [325.1577, [324.0123,

200 17.7713]V 10.0152]V
voltage droop constant 70 Iidq [13.5839, [14.2141,
180 kpv 0.8628]A 0.0645]A
1000 synchronizing integrator 0.3 Vodq [325.0891, [323.3467,
gain kss 8.1140]V −0.0231]V
160 voltage loop PI kp , ki 0.0001, 0.0007 Iodq [12.2938, [12.9207,
0.4219]A -0.3418]A
0 2 4 6 8 10 12 14 16 18 0 Nominal frequency ω0 314.159rad/s θ̂ 0.0546rad 0.0309rad
0 2 4 6 8 10 12 14 16 18 Virtual impedance 0.03 + 1.099j Pref , Vref 2kW , 325.27 25Ω, 0
Time (s) Time (s) Zvirtual

(c) Output voltage (d) Reactive power output

Fig. 5: Grid Fault and induction machine operations R EFERENCES

[1] K. Solangi, M. Islam, R. Saidur, N. Rahim, and H. Fayaz, “A review on
Duration Event global solar energy policy,” Renewable and Sustainable Energy Reviews,
t = 3.0s 3000W impedance load tripped vol. 15, no. 4, pp. 2149 – 2163, 2011.
t = 6.0s Grid is ON (synchronization begins) [2] R. Bhatt and B. Chowdhury, “Grid frequency and voltage support
using pv systems with energy storage,” in 2011 North American Power
t = 9.1s Grid connected Symposium, Aug 2011, pp. 1–6.
t = 11.0sto11.1s LG-fault on line [3] W. Yao, M. Chen, M. Gao, and Z. Qian, “Development of communi-
A 0.5HP single phase induction motor (IM) is operated cationless hot-swap paralleling for single-phase ups inverters based on
adaptive droop method,” in 2009 Twenty-Fourth Annual IEEE Applied
along with a 3000W impedance load. Due to the induction Power Electronics Conference and Exposition, Feb 2009, pp. 1283–1287.
machine there is a voltage drop of 11V at the PCC as shown [4] Q. C. Zhong and G. Weiss, “Synchronverters: Inverters that mimic
in Figure 5c which further reduces the active power drawn synchronous generators,” IEEE Transactions on Industrial Electronics,
vol. 58, no. 4, pp. 1259–1267, April 2011.
by the impedance load. A large perturbation is created, by [5] J. A. Suul, S. D’Arco, and G. Guidi, “Virtual synchronous machine-based
switching off the 3000W impedance load at t = 3s, Figure 5a control of a single-phase bi-directional battery charger for providing
depicts the rise in frequency, due to a sudden decrease in total vehicle-to-grid services,” IEEE Transactions on Industry Applications,
vol. 52, no. 4, pp. 3234–3244, July 2016.
connected load from 2960W to 372W induction motor load. [6] S. Mishra, D. Pullaguram, S. A. Buragappu, and D. Ramasubramanian,
The slow decay of the frequency is due to the high damping “Single-phase synchronverter for a grid-connected roof top photovoltaic
provided by the large gain of the integrator, kif , during the system,” IET Renewable Power Generation, vol. 10, no. 8, pp. 1187–
1194, 2016.
islanded mode. When grid is again available at t = 6s, the [7] “A new single-phase pll structure based on second order generalized
synchronization process is started and the frequency oscilla- integrator,” in 2006 37th IEEE Power Electronics Specialists Conference,
tions during this period has a minimal effect on the power June 2006, pp. 1–6.
[8] F. Xiao, L. Dong, L. Li, and X. Liao, “A frequency-fixed sogi-based
absorbed by the IM. The inverter gets connected to the grid at pll for single-phase grid-connected converters,” IEEE Transactions on
t = 7.9s with Pref set to 0W . A 5 cycle LG fault is simulated Power Electronics, vol. 32, no. 3, pp. 1713–1719, March 2017.
in the network beyond the PCC at t = 11s. During this period [9] S. Golestan, M. Monfared, F. D. Freijedo, and J. M. Guerrero, “Dynamics
assessment of advanced single-phase pll structures,” IEEE Transactions
voltage dips to to 158V causing the inverter reactive power on Industrial Electronics, vol. 60, no. 6, pp. 2167–2177, June 2013.
output to rise due to the action of the voltage control loop. [10] H. J. Avelar, W. A. Parreira, J. B. Vieira, L. C. G. de Freitas, and E. A. A.
The power reference limiter in this loop restricts the rise to Coelho, “A state equation model of a single-phase grid-connected inverter
using a droop control scheme with extra phase shift control action,” IEEE
2820 V AR instead of the intended 4700V AR. Transactions on Industrial Electronics, vol. 59, no. 3, pp. 1527–1537,
March 2012.
VI. C ONCLUSION [11] “Ieee standard for interconnecting distributed resources with electric
A multi-functional controller for single phase rooftop PV power systems - amendment 1,” IEEE Std 1547a-2014 (Amendment to
IEEE Std 1547-2003), pp. 1–16, May 2014.
system is proposed in this paper. The inverter can be operated [12] W. Freitas, W. Xu, C. M. Affonso, and Z. Huang, “Comparative anal-
in various modes as per the changes in the grid condition. ysis between rocof and vector surge relays for distributed generation
The controller allows the PV to operate in both grid connected applications,” IEEE Transactions on Power Delivery, vol. 20, no. 2, pp.
1315–1324, April 2005.
and isolated mode without causing any power interruptions to [13] A. Nabae and T. Tanaka, “A new definition of instantaneous active-
the local loads. An islanding detection technique is developed reactive current and power based on instantaneous space vectors on
which isolates the inverter from grid during LoM and supply polar coordinates in three-phase circuits,” IEEE Transactions on Power
Delivery, vol. 11, no. 3, pp. 1238–1243, Jul 1996.
uninterrupted power to the local loads. Inverter is made to [14] A. Tabesh and R. Iravani, “Multivariable dynamic model and robust
respond to the grid frequency and voltage variations as an control of a voltage-source converter for power system applications,”
ancillary service in grid connected mode. Isochronous control IEEE Transactions on Power Delivery, vol. 24, no. 1, pp. 462–471, Jan
2009.
of the system frequency during islanded condition is posible [15] F. Mei and B. Pal, “Modal analysis of grid-connected doubly fed
due to the activation of the integrator, kif , present in the induction generators,” IEEE Transactions on Energy Conversion, vol. 22,
controller driven by the PCC voltage. Further the active power no. 3, pp. 728–736, Sept 2007.
set points to the controller are developed such that the inverters