You are on page 1of 14

Applied Energy 264 (2020) 114733

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Enhancing the voltage stability of distribution network during PV ramping T


conditions with variable speed drive loads
Salish Maharjan , Dhivya Sampath Kumar, Ashwin M. Khambadkone

Department of Electrical and Computer Engineering, National University of Singapore, Singapore 117583

HIGHLIGHTS

• Severe PV ramping at high penetration could cause voltage instability in presence of Induction motor loads.
• Replacement of Induction motor by Variable Speed Drive (VSD) loads enhances voltage stability of distribution network.
• Enhanced stability with VSD is studied theoretically with Q-V analysis and with time simulations.
• Minimum IM replacement by VSD is determined for realistic distribution network at various PV penetration levels.

ARTICLE INFO ABSTRACT

Keywords: The Photovoltaic (PV) sources have become the most popular renewable energy resources as it is modular and
Voltage stability are deployed quickly. However, the distributed PVs are highly volatile and can have high ramping characteristics
PV ramping where the ramping can go up to 90% of its capacity in about 20 s due to events such as the passing of fast-moving
Distribution network clouds. This kind of PV ramps with high magnitudes can lead to voltage instability in distribution networks that
Variable Speed Drives
are usually dominated by Induction motor (IM) loads. This paper explains the mechanism of such an instability
Q-V analysis
DigSILENT
analytically using Q-V analysis with distribution feeder and IM load characteristics. Highlighting the gradual
replacement of IM loads by Variable Speed Drive (VSD) loads due to energy efficiency policies, the paper de-
monstrates the enhanced voltage stability under VSD loads by both time simulation and analytical approach.
Furthermore, the analysis is extended to a realistic distribution network namely, United Kingdom General
Distribution System (UKGDS) and extensive case studies are conducted to analyze the voltage stability during PV
ramping events at various load compositions of IMs and VSDs. Moreover, the minimum VSD penetration level
required to avert the voltage instability at various PV penetration levels have also been determined.

1. Introduction equipped with slow voltage controllers such as On-Load Tap Changer
(OLTC), whose typical time delay of operation is of 20 to 60 s [5].
One of the promising options of all the renewable energies available Abrupt voltage changes brought in by PV ramps within the time delay
worldwide is photovoltaics (PV) sources, which is mainly due to the of OLTC can have severe impacts on the network voltage stability.
gradual decrease in their installation cost.The penetration of PVs is Additionally, due to higher R/X ratio of lines (ranges from 0.85 to 7.7
certain to grow in distribution grids in the coming decades as renewable [6]), the short-term PV ramping in the distribution network would
targets and energy efficiency have become an integral part of govern- bring faster transition on the network voltage. Inability of the dis-
ment policies around the world [2]. Though the proliferation of PVs tribution voltage regulator to respond timely to this voltage transition
support in the reduction of net power demand, they introduce abrupt could lead to issues such as violation of the voltage regulatory limits
power changes in the network due to their inherent ramping nature. [7,8], large voltage dips [9], poor power quality [10], and even dis-
The quantification of probable PV ramping shows that PV of several connection of PV sources [11], which could further worsen the voltage
MWp capacity could have a maximum ramp of ± 90 % of its rated ca- stability.
pacity in about 20 s and even higher at a larger time-span [3,4]. The The short-term voltage instability arising due to PV sources are well
short-term PV ramping (within 20–30 s) is a major concern for the reported in [12–18]. These reported cases of instabilities can be broadly
distribution network operation because, the traditional feeders are only classified into two categories namely, (a) instability due to


Corresponding author at Department of Electrical and Computer Engineering, National University of Singapore (NUS), Singapore 117583.
E-mail address: salish.maharjan@u.nus.edu (S. Maharjan).

https://doi.org/10.1016/j.apenergy.2020.114733
Received 29 November 2019; Received in revised form 20 February 2020; Accepted 22 February 2020
Available online 06 March 2020
0306-2619/ © 2020 Elsevier Ltd. All rights reserved.
S. Maharjan, et al. Applied Energy 264 (2020) 114733

Nomenclature iinv input DC current to inverter of VSD


n tap position of distribution transformer
A. Abbreviations PL (V ) P-V characteristics of load
PN (V ) P-V characteristics of network
D-FACTS Distributed Flexible AC Transmission System Pim Active power consumption of induction motor
DigSILENT A RMS/EMT capable power system simulation software PL, cri Critical load demand
[1] PL, nom Nominal load demand
IM Induction Motor Pmot Active power drawn by IM of VSD
MV Medium Voltage Pvsd, Q vsd Active, Reactive power demand of VSD
OLTC On-Load Tap Changer QL (V ) Q-V characteristics of load
PV Photovoltaic QN (V ) Q-V characteristics of network
RMS Root Mean Square Qim Reactive power consumption of induction motor
SCR Short circuit ratio Qtot total reactive power consumption of IM with shunt capa-
SVC Static Var Compensator citor
UKGDS United Kingdom General Distribution System R, X , Z Resistance, reactance and impedance of line
VSD Variable Speed Drives Rs , Rr Stator and rotor resistance of IM
Rtr , Xtr resistance and reactance of transformer
B. Variables s Slip of Induction Motor
u Commutation angle of diode bridge rectifier
V / tap incremental voltage per tap of transformer vdc Voltage at DC link of VSD
a Transformer turn ratio Voltage drop across diode in rectifier
Cdc Capacitance at DC link of VSD Xc commutation reactance of VSD
Eg Voltage at external grid Xm Magnetising reactance of IM
Erms RMS value of supply voltage to VSD Xs , Xr Stator and rotor reactance of IM
idc Current in DC link of VSD Xcap Reactance of capacitor shunted with IM for power factor
idgcom, iqgcom Grid side current during commutation period in dq correction
frame Zg impedance of external grid
idgcond , iqgcond Grid side current during conduction period in dq frame

disconnection of PV after a severe contingency [12–16] and (b) in- gradually evolving.
stability due to worst PV ramping [17,18]. The voltage instability for Customer loads were dominated by IM in the past, however, they
the first case is mainly due to the load demand crossing the loadability are being gradually replaced by Variable Speed Drives (VSD). European
limit when the PVs get disconnected just after a severe contingency in commission is actively promoting VSD as an alternative to IM for in-
the network. This phenomenon is explained analytically in [13], and it creasing the proportion of high efficient motors, in accordance with
is suggested to incorporate fault-ride-through schemes in PV inverter to their Motor Regulation (EC) 640/2009 act [23]. As a consequence, VSD
prevent such a kind of instability. An enhanced fault-ride-through is projected to replace 50% of the motor load in Europe by 2020. This
scheme with fast terminal voltage recovery for the restoration of PV change in the load trend has not been considered by the above-men-
power injection is achieved using dynamic voltage support in [14,15]. tioned literature works, where the researchers attempted to solve the
Alternatively, the voltage instability due to disconnection of PVs after a voltage instability problems caused by PV ramps assuming that the
severe contingency can be prevented by proper allocation of D-FACTS loads are mainly IM loads in the network. Moreover, many new
devices like static var compensator (SVC) as discussed in [12,19,20]. equipment/technique based solutions such as D-FACTS, energy storage,
However, the installation of the D-FACTS devices would have addi- regulators, control mechanisms, etc. that need to be additionally in-
tional cost as compared to embedding fault ride schemes in PV in- stalled in the feeder have been proposed in the literature works. VSDs
verters. Furthermore, the short-term voltage instability risk is fore- are known to have inbuilt energy storage at the DC link, which has the
casted day-ahead in [16] by predicting the weather and load across the capability of slowing down the dynamics of supply voltage in events
distribution network. Thereafter, the demand responding resources are such as PV ramps. Hence, this paper attempts to provide a fresh per-
optimally deployed during the period of higher risk to prevent voltage spective of the solutions proposed to solve the voltage instability pro-
instability. However, the reliability of this method depends on the blem considering only a high proportion of VSD loads in the network
performance of methodology used for forecasting voltage instability excluding any additional voltage regulating devices/techniques.
risk. The second case for instability is reported due to the worst short- This paper conducts voltage stability analysis considering worst PV
term ramping of PV sources, which are expected to happen during ramping events on a distribution feeder with both IM and VSD loads.
transient cloud passing [17,18].The stability due to PV ramping have Initially, the stability analysis is performed on a simplified distribution
been mainly conducted using time simulation methods as in [17,18]. feeder at various probable PV ramp rates using both analytical (Q-V
However, analytical studies corroborating the time domain simulations analysis) and RMS simulations (DigSILENT/Powerfactory). The Q-V
provide more specific inference of the studies, which has not been characteristics of the distribution feeder and loads (IM, VSD, and static
widely discussed in many of the literature works. The stability issues loads) are analytically modeled by considering the grid impedance, tap
due to the second case have been usually solved using battery storage changers, and R/X elements of lines for performing the Q-V analysis.
units, which can compensate the power fluctuations from PV sources, as This analysis is extended to a realistic distribution network namely,
discussed in [21,22]. Another possible solution is the local reactive United Kingdom General Distribution Network (UKGDS), with re-
power support from PV inverters as discussed in [18]. However, these sidential, commercial and industrial feeders. Various ratios of IM and
literature works have proposed solutions considering that the instability VSD are considered in the load composition of commercial and in-
is caused by PVs overlooking the load response from stiff loads like IM. dustrial customers, to determine the impacts of VSD penetration on
This paper focuses on the solution measure of voltage instability pro- voltage stability of the distribution network during severe PV ramping
blem from the customers’ load perspective, as the type of loads is conditions. Besides, the ratio of IM loads that need to be replaced by

2
S. Maharjan, et al. Applied Energy 264 (2020) 114733

VSD loads (industrial and commercial customer) in the UKGDS network


is also determined to ensure voltage stability under severe PV ramping
events at various PV penetration levels. This approach can assist the
policymakers to set the required IM replacement targets in the future to
ensure network stability at a planned PV penetration levels.
The main contributions of this paper can be listed as:

• This paper theoretically determines the Q-V characteristics of a


simplified distribution feeder incorporating line resistance, re-
actance, OLTC, and grid impedance. Utilizing this, the voltage in-
stability due to severe PV ramping events is studied analytically in
the presence of both IM and VSD loads.

Fig. 1. Network and load Q-V characteristics at various loading conditions. (a)
The voltage stability analysis of the simplified distribution feeder is Normal loading (PL), (b) Critical loading (PLcri), and (c) loading above PLcri.
then analysed in time domain simulations using DigSILENT/
Powerfactory for corroboration.
• The stability studies are later extended to a UKGDS network to more volatile, which would bring a fast transition in network Q-V
characteristics. If the voltage controller is not able to adjust the network
analyze the network response during PV ramps with a higher pro-
portion of VSDs. Finally, the minimum required VSD penetration characteristics at the same pace of PV rampings, then there could be
levels to avert voltage instabilities are determined. voltage stability issues. This paper investigates these voltage stability
issues and provides measures to prevent them.
The rest of the paper is organized as follows. Section 2 revisits the
voltage stability analysis with Q-V analysis. Section 3 analytically de- 3. Q-V characteristics of distribution network and loads
termines the Q-V characteristics of a simplified distribution feeder and
loads (IM, VSD, and static loads). Section 4 performs the stability This section models Q-V characteristics of a simplified distribution
analysis during PV ramping events in the presence of IM and VSD loads feeder and various loads such as IM, VSD and static loads. These Q-V
using both Q-V analysis and time simulation. Section 5 investigates the characteristics are essentially used for stability analysis in subsequent
stability analysis in the UKGDS network with different load composi- sections.
tions at various PV penetration levels considering severe PV ramping Consider the distribution network represented by a simplified feeder
events. The quantification of the minimum required VSD penetration is shown in Fig. 2. It is connected to the external grid represented by a
described in Section 6. Finally, Section 7 concludes the paper. constant voltage source with internal impedance Zg . OLTC is the only
available voltage controller in this feeder like in many existing dis-
2. Voltage stability analysis tribution networks. The distributed PVs and loads are simplified by
aggregated PV and load in the load-bus.
The voltage stability of the power network is primarily determined
by the loadability limit. The loadability limit can be defined as the 3.1. Network Q-V characteristics
critical value of load demand beyond which the network voltage will
collapse. The loadability limit not only depends on the active power Let Z¯tr = Rtr + jXtr be the short circuit impedance of the tap-chan-
demand but also on the Q-V characteristic of network (QN (V ) ) and load ging transformer, referred from the low voltage (LV) side. As per the
(QL (V ) ) [24]. To assess the voltage stability at a particular loading Fig. 2, the input (Īg ) and output (Ī ) currents of the transformer are given
condition, firstly the real power delivered by network and demanded by by:
the load are equated (PN (V ) = PL (V ) ). Then, the Q-V characteristics of
the network and load are determined at that loading condition. Finally, E¯p/ a E¯
I¯ = ,
the network and load characteristics are plotted in a common Q-V Z¯tr (2)
diagram to determine the possible equilibrium points.
Fig. 1 shows the Q-V plot of a generic network and load at various E¯ E¯p
I¯g = + 2 .
loading conditions. At normal loading conditions, there could be mul- ¯
aZtr a Z¯tr (3)
tiple intersections of load and network characteristics (Fig. 1a), which
are referred to as equilibrium points. In Fig. 1a, there are two equili- Here, Ēp and Ē are the primary and secondary side voltages of the
brium points denoted by ‘s’ and ‘u’, which refer to stable and unstable transformer.1 a is the tap ratio which is a real number. Alternatively, Ī
points. The equilibrium points are stable if the condition (1) holds true, can be also calculated as:
otherwise it is unstable [24]. E¯ V¯
I¯ = .
dQN (V ) dQL (V ) Z¯ (4)
<
dV dV (1)
On equating, (2) and (4), we get
On further increasing the load demand (PL (V ) ), there would be only
Z¯ + Z¯tr Z¯tr
one solution, which is referred to as critical demand, denoted by PL, cri in E¯p = aE¯ aV .
Fig. 1b. Beyond this demand, there would be no equilibrium points, Z¯ Z¯ (5)
which will result in total voltage collapse (Fig. 1c). Furthermore, the Q- Similarly, Īg could be equated as in (5) to get (6).
V characteristic of the network becomes more complicated with asso-
ciated voltage controllers. As a consequence, the stable equilibrium E¯g E¯p E¯ E¯p
= + 2
point might lie below the technically acceptable voltage limit at certain Zg aZ¯tr a Z¯tr (6)
loading conditions. This condition is referred to as partial voltage col-
lapse where the system is still stable, but operates at reduced voltage. It
can also be referred as an emergency state [24], which is usually re- 1
In the rest of the paper, every phasor, its conjugate and magnitude are re-
solved by shedding the portion of loads in the network. spectively denoted with a bar above (e.g., X̄ ), below (e.g., X ) and without any
With the proliferation of PV sources, the net load demand becomes bars (e.g., X).

3
S. Maharjan, et al. Applied Energy 264 (2020) 114733

Therefore, the reactive power transfer capability of the network while


meeting certain load power (PS = PL) is given by:

QN = (aVEg ) 2 (p2 + q2) (PL + a2V 2p)2 a2V 2q. (15)

It is evident from (15), the Q-V characteristic of distribution feeder


Fig. 2. Simplified distribution feeder with tap changing transformer.
depends on short circuit impedance of external grid (Zg ), R/ X ratio of
line, active power demand, and tap ratio of the transformer.
Z¯g + a2Z¯tr aZ¯tr The following subsections studies the network Q-V characteristics
or , E¯ = E¯p Eg .
aZ¯g Zg (7) for a simplified distribution feeder shown in Fig. 2 with the parameters
listed in Table 1.
Using (5) to replace Ep from (7), we get
aEg Z¯ V¯ (Z¯g + a2Z¯tr ) 3.1.1. Dependence on short circuit ratio
E¯ = + 2 . The strength of the external grid is defined by Short Circuit Ratio
a2Z¯ + Z¯g + a2Z¯tr a Z¯ + Z¯g + a2Z¯tr (8)
(SCR), which could be defined as:
The complex power delivered by the network could be derived as:
Fault MVA
SCR =
S¯ = V¯ I = V¯
E V
. Transformer Rating (16)
Z (9)
The SCR between 3 and 5 is considered to be moderate grid, whereas
On substituting Ē from (8) in (9), we get above 5 is considered as strong grid [25]. For a given SCR, the internal
¯ g impedance of the external grid is calculated as:
aVE a2V 2
S¯ = .
a2Z + Z g + a2Z tr a2Z + Z g + a2Z tr (10) Eg2 Eg2
Zg = = .
Fault MVA SCR × Transformer Rating (17)
Using V¯ = V (cos jsin ), S¯ could be decomposed into active and
reactive parts respectively, as: With all the parameter fixed, the Q-V characteristic of the distribution
feeder is estimated using (15), at no load condition, for various value of
PN = aVEg (pcos + qsin ) a2V 2p , (11)
SCR at the connection point to the external grid, which is shown in
Fig. 3(a). According to this, the reactive power transfer capability of the
QN = aVEg (qcos psin ) a2V 2q. (12)
network is reduced for weaker grid.
Here, p + jq = , which can be expanded as:
1
a2Z + Z g + a2Z tr

a2R + Rg + a2Rtr + j (a2X + Xg + a2Xtr ) 3.1.2. Dependence on R/X ratio of line


p + jq = . For the fixed length of line and conductor configuration, X remains
(a2R + Rg + a2Rtr )2 + (a2X + Xg + a2Xtr ) (13)
2
constant. In contrast, R depends on the size of conductors. Therefore,
It can be noticed, when R 0, Rtr 0, Xtr 0, Rg 0, Xg 0 , and the R/X of the line will be different for various conductor sizes. Under
a = 1, the above relation simplify to the case for transmission line, as this consideration, the network Q-V characteristics are plotted for the
shown in [24]. Furthermore, on squaring and adding (11) and (12), we various values of R/X ratio at no-load condition and is shown in
get: Fig. 3(b). It can be observed that the higher value of R in the line lowers
the Q transfer capability.
(PN + a2V 2p) 2 + (QN + a2V 2q) 2 = (aVEg )2 (P 2 + q2 ). (14)
3.1.3. Dependence on active power demand
Table 1 The active power demand reduces the Q transfer capability of the
Simplified distribution feeder parameters. network. The impact of active power demand on Q-V capability of the
Grid 33 kV SCR = 5 pu network is shown in Fig. 3(c).

Line 11 kV 6 km
3.1.4. Dependence on tap position of transformer
R = 0.322 ohm/km X = 0.167 ohm/km
The tap position (n) change the turns ratio of the transformer as:
Transformer Rtr = 0.124 pu Xtr = 0.173 pu 10 MVA
a = 1 + n ( V / tap) (18)
V/tap = ±1% tap = +3 to −15 at HV side
The incremental voltage per tap ( V / tap ) is a fixed quantity and usually
1 to 2% [5]. The change in tap position of the transformer also modifies

Fig. 3. Q-V capabilities of the distribution network when changing the parameters independently, such as (a) SCR, (b) R/X. (c) PL , and (d) tap position (n).

4
S. Maharjan, et al. Applied Energy 264 (2020) 114733

the Q-V capability of network as shown in Fig. 3(d). In this plot, one can Table 2
observe the increased Q transfer capability when the tap position is Induction Motor parameters for various customer types [27].
lowered. This plot is drawn for step down transformer with tap provi- IM Parameters (p.u.)
sion at the HV side.
type Rs Xs Xm Rr Xr H T0
3.2. Q-V characteristic of induction motor load
R 0.077 0.107 2.22 0.079 0.098 0.74 0.46
C 0.031 0.1 3.2 0.018 0.18 0.7 0.6
The induction motor (IM) is usually connected with a shunt capa- I 0.013 0.067 3.8 0.009 0.17 1.5 0.8
citor for power factor correction. For this reason, the Q-V or P-V
characteristics of IM and shunt capacitors are to be aggregated. The
active and reactive power consumption of IM at various input voltages at the operating region. However, when the supply voltage is reduced to
is modeled in [26] and the following relations are obtained. the stalling point, the IM stops which leads to a drop in active power
consumption. On further reducing the voltage, the active power con-
V 2 [(Rs + Re) + j (Xs + Xe )]
Pim + jQim = sumption also reduces (here active power consumption is mainly due to
(Rs + R e) 2 + (Xs + Xe )2 (19) losses). On the other hand, Q-V characteristics of IMs at the operating
where, region show the gradual increase of reactive power consumption as the
supply voltage is lowered to some extent. However, the reactive power
jXm (Rr / s + jXr )
Re + jXe = scales when the supply voltage is lowered near to the stalling region. On
Rr /s + j (Xm + Xr ) further lowering the supply voltage, the reactive power of IM is reduced
The slip (s) is obtained by solving the torque equation shown below. gradually.

V 2Xm2
Te = 3.3. Q-V characteristic of variable speed drive load
(R1 + Rr / s )2 + (X1 + Xr )2 (20)
where, V/f control of induction motors is a widely used variable speed drive
technology. The commonly used configuration for V/f control is with a
jXm (Rs + jXs )
R1 + jX1 = front end rectifier unit, intermediate DC link capacitor and finally the
Rs + j (Xm + Xs ) inverter for supplying controlled voltage and frequency to drive the
The combined reactive power (Qtot ) consumption of IM and shunt ca- induction motor, which is shown in Fig. 5. VSD can be designed to ride-
pacitor is given by: through voltage sags as mentioned in [28]. Usually, energy storage is
connected to the DC link to support the drive units during voltage sags.
V2 The energy storage units are allocated to support the load for 3 s for
Qtot = Qim
X cap (21) 80% sag [29], or 4 s for 85% sag [30]. With this information, the energy
storage unit (considered to be the capacitor) would be 180µF /kW for a
Fig. 4 shows the P-V and Q-V characteristics of IM at both operating
VSD connected to a 11 kV supply. The capacitor sizing lies in the range
and stalling region. Although the IM is not operated at the stalling re-
gion, it can reach this region during voltage emergency situations, and
remain here for several seconds until it is tripped by overload protec-
tion. The stalling of IM for a few seconds can worsen the voltage
emergency situations by further reducing the network voltage, which
could consequently lead to voltage instability before it is tripped by
overload protection.
The P-V and Q-V characteristics of the residential, commercial and
industrial induction motors (11 kV, 5 kW) with shunt capacitor having
the parameters of Table 2 are shown in Fig. 4. The shunt capacitor is
calculated to operate the IM at 0.98 pf at rated load. These character-
istics are obtained by solving the model of IM with shunt capacitor
((19)–(21)) at various supply voltage (V) at a nominal torque (T0 ) Fig. 5. V/f control of induction motor.
(shown in Table 2). All the IMs have a fairly constant P-V characteristic

Fig. 4. P-V and Q-V characteristics of induction motors with shunt capacitor.

5
S. Maharjan, et al. Applied Energy 264 (2020) 114733

as stated in [31] for commercial VSD units. Furthermore, the commu- where u is the commutation angle, which is calculated as:
tation reactance Xc is usually 5% of the base impedance of the induc-
2X c idc
tion motor [32]. The main idea behind this discussion of the sizing of u = cos 1 1 .
capacitor and reactance is that their sizing impacts the load char- 6 Erms (30)
acteristic of the VSD. Finally, the active and reactive power consumed by the VSD is calcu-
We have used the average model of the VSD for the estimation of its lated as:
P-V and Q-V characteristics. If vqg = 2 Erms and vdg = 0 is the phase
voltage at the supply node in the qd reference frame, then the DC link 3
Pvsd + jQvsd = [(vdg idg + vqg iqg ) + j (vqg idg vdg iqg )].
voltage can be obtained as: 2 (31)

3 6 3 didc The P-V and Q-V characteristics of the VSD load were estimated for
vdc = Erms X c idc 2l c 2Vdiode, various input voltages (Erms ) using (31), where the input currents (idg
dt (22)
and iqg ) were calculated using (25) and (26). The resulting character-
where Xc = g lc , lc is the inductance of the commutation reactance, and istics of the VSDs, whose parameters are from Table 2, are portrayed in
Vdiode is the voltage drop across the diode, which is neglected. The DC Fig. 6. This figure shows that the active power consumption of the VSD
current to the inverter is given by: load is constant at various input voltages during the steady-state.
vdc However, the reactive power gradually increases at lower voltages, but
iinv = idc Cdc .
dt (23) not abruptly as in the case of IMs.
It is to be noted that the active power consumption of the VSD is not
In the steady state,
constant during the transient state of network voltage (while the supply
Pmot voltage is falling or rising). During such transients, the energy storage
iinv = idc = ,
vdc (24) unit compensates the active power from the grid, which slows the dy-
where Pmot is the active power drawn by the induction motor. Moreover, namics of the supply voltage. This dynamic nature of the VSD is ben-
the input current of the rectifier is calculated [29,33] as: eficial to the grid during severe PV ramps and is discussed in Section
4.2.
idg = idgcom + idgcond and (25)
3.4. Static loads
iqg = iqgcom + idgcond. (26)
Here, (idgcom, iqgcom ) and (idgcond , iqgcond ) are the currents during the The aggregated load seen from a MV node is determined either by
commutation and conduction periods, which are calculated as: measurement [34] or by component-based approach [35]. These ap-
proaches have been experimentally validated in [36] and still used in
2 3 5 5 3 2 Erms
idgcom = idc cos u + cos + sin (u) power system industries [37]. However, the load model of electrical
6 6 Xc
appliances which were studied in the 80s and 90s are not applicable at
3 2 Erms 3 2 Erms present since modern appliances comprise of more electronic devices
sin (2u) u,
4 Xc 2 Xc [38,39]. Accordingly, this paper utilizes the ZIP load models studied for
2 3 7 5 New York City in 2011 [40] for residential, commercial and industrial
idgcond = idc cos + cos u + ,
6 6 (27) consumers which are listed in Table 4. Fig. 7 represents the char-
acteristics of static loads belonging to various customers in the dis-
2 3 5 5 3 2 Erms tribution network.
iqgcom = idc sin u sin + (cos (u) 1) +
6 6 Xc
3 2 Erms
(1 cos (2u)), and 4. Stability analysis during PV ramping events
4 Xc (28)
The impact of PV ramping on voltage stability of the distribution
2 3 7 5 network is studied on a simplified distribution network, shown in Fig. 2.
iqgcond = idc sin sin u +
6 6 (29) The specific parameters of the network is listed in Table 1. The dead-

Fig. 6. P-V and Q-V characteristic of VSD.

6
S. Maharjan, et al. Applied Energy 264 (2020) 114733

Fig. 7. Normalized P-V and Q-V characteristics of residential, commercial, and industrial type static loads. Here, the power factor is assumed to be 0.98.

band of the OLTC controller is ±2% of the reference voltage and has an
intentional and mechanical time-delay of 10 s. The OLTC is designed to
regulate the low voltage side of the transformer. The voltage regulatory
limit of the feeder is adopted from EN50160 [41] (0.9–1.1 p.u.) and the
same is followed throughout the paper.
A PV of 3 MWp capacity and a load of 5 MW rating is connected at
the load bus. The PV is assumed to be operated at unity power factor as
many of the power networks do not allow to inject reactive power from
distributed generators. The voltage stability of this feeder is studied
using both simulation and analytical based approach. The simulation is
conducted in DigSILENT Power Factor [1] using its RMS simulation
tool, whereas Q-V analysis is performed for analytical study of stability.
It is reported that the static loads usually are not the critical elements
for voltage stability compared to dynamic loads such as IM [42].
Therefore, the voltage stability has been studied with only IM and its
alternative VSD during PV ramping events. The study has mainly con-
centrated on PV ramp-down cases rather than ramp-up cases. It is be-
cause the PV ramp-up events did not have stability issues other than the
voltage regulation problem.
The stability analysis of simplified feeder is conducted using both
time simulation and Q-V analysis considering the presence of the fol-
lowing loads namely,

• Induction motor load and Fig. 8. A simplified distribution network simulated with PV ramping down

• Variable Speed Drive load events. All plots are obtained with 10 MVA and 11 kV as the base.

4.1. Considering induction motor load ramping leads to large voltage dips in short-time (refer to 90%/20 and
90%/30 s ramping case in Fig. 8), causing the IM to draw more reactive
An induction motor of 6.2 MVA capacity is considered to be oper- power while consuming constant active power. Such a response of IM
ating at 5 MW load during a sunny day, where the PV is generating its load will further exaggerate the voltage dip, finally leading to partial
rated power. An IM is having a shunt capacitor for local var generation voltage collapse before the OLTC can respond. The voltage stayed at 0.5
which resulted in a load power factor to 0.98. The induction motor p.u. level after the partial voltage collapse, which stalled the IM loads.
parameter used for the study is belonging to commercial type and is During the stalling condition, the current drawn by IM is nearly equal to
listed in Table 2. the starting current. As a result, the IM will be tripped only after few
seconds of stalling by the over-current protection relay. Once the vol-
4.1.1. Simulation approach tage collapses, the OLTC response would not be able to boost the vol-
As per the reported values of PV ramps [3], the PV is subjected to tage back until customers’ loads are intentionally shedded. The reason
90% change in output power within 20, 30 and 40 s in the simplified behind the partial voltage collapse will be analysed and discussed in
distribution feeder. The simulated results are shown in Fig. 8. When the detail from the analytical stability studies in the next subsection.
PV ramps down severely, the abrupt decline of active power from PV
have to be supplied by the grid, which leads to increased voltage drop 4.1.2. Analytical approach
along the line. If the voltage drop is not quickly controlled, the reactive The partial voltage collapse due to PV ramping and load response
power consumption of the IM will increase (refer to Fig. 4), which leads from IM could be well understood using the concept of Q-V analysis.
to even higher voltage drop along the line. The OLTC might have suf- The Q-V curve for IM remains fixed for a given loading condition,
ficient time to respond for voltage correction in case of slow PV whereas the Q-V curve of the network depends on active power demand
ramping (refer to 90%/40 s ramping case in Fig. 8). However, faster and the tap position of OLTC. The network Q-V curve shrinks with time

7
S. Maharjan, et al. Applied Energy 264 (2020) 114733

4.2.1. Simulation approach


A ramping event of 90% within 20 s in a PV generator was simu-
lated in the simplified distribution network with VSD loads. The results
from RMS simulation for this case is shown in Fig. 10. Unlike with IM
loads, the ramping down of PV did not lead to partial voltage collapse,
even under 90%/20 s ramp down scenario. With VSD loads, the net-
work voltage dynamics were slower than in the case of IM. As a result,
the tap changers had sufficient time to respond for boosting the node
voltage, which helped to prevent from partial collapse.
The static characteristics of VSD show that it has a constant active
power consumption at various supply voltage, as seen in Fig. 6. How-
ever, it is not constant during the supply voltage dynamics. During
voltage declining period, a part of the VSD load is contributed from the
energy storage at the DC link. This phenomenon lowers the active/re-
active power consumption, which is shown by the dotted ellipse in
Fig. 10. As the PV ramping stops, the energy storage elements are
charged, which increases both the active and reactive power, causing a
small voltage dip that is corrected by the OLTC. In summary, the in-
teraction of the VSD with the supply voltage causes the voltage to drop
slowly during PV ramping conditions. This makes the OLTC action more
effective, although it has a delayed response, to begin with. This is one
of the important dynamic load characteristics of VSD, which is bene-
Fig. 9. Tracing time-domain simulations in the Q-V diagram at various time ficial to the distribution network in terms of voltage stability stand-
instants to portray voltage stability issues due to short-term PV ramping and IM point.
load response.

4.2.2. Analytical approach


as PV power ramps down and the rate of Q-V curve shrinking depends The advantage of the VSD load to the distribution network is more
on the PV power ramping rate. For the case of 90%/20 s ramp down of pronounced when analysing the simulated results in the Q-V diagram.
PV, the Q-V curves for both the network and the IM load are traced out The major advantage is that the Q-V characteristic of VSD is its flat
for various time-intervals of the simulation, as shown in Fig. 9. During 0 nature over the wide range of operating voltage than that of IM, which
to 10 s interval, the system is at a steady-state, and there are three causes slow falling of node voltage, while PV power ramps down. The
possible equilibrium points defined as s, u , and s in Fig. 9. According Q-V plots of VSD loads and network at various time interval of the si-
to (1), the point u is unstable, whereas s and s are stable. However, s mulation are traced out in Fig. 11. During the interval (10–30 s) of PV
does not lie in the operating region of the IM. As a result, the network ramp down, the network Q-V curve did not shrink drastically as it was
remains at the Stable Equilibrium Point (SEP) s, which can be verified in case of IM load. This is due to reduced active power drawn from the
from the time simulation shown in Fig. 8. As the PV ramps down by supply side as the part of the load power will be supported by energy
90% of its capacity during a 10 to 30 s interval, the Q-V curve of the storage in the DC link. Soon after the ramping event, the active power
network shrinks at a faster rate than the OLTC can respond for ex- gradually increases to charge the storage and supply the load com-
panding the Q-V capability. As a result, there is no SEP at the end of pletely from the grid. This phenomenon leads to further shrinking of the
30 s, causing the voltage to collapse. Subsequently, the IM stalls due to Q-V curve during 30–50 s intervals. Due to sufficiently slow voltage
low network voltage, causing the Q-V curve of the network to expand as
the IM drops its active power consumption. This happens within a short
period of time from 30 to 32.6 s. At the end of 32.6 s, s is the only
equilibrium point. As time progresses, the OLTC continuously changes
its tap setting until it reaches the maximum limit, but still cannot re-
store the network voltage as seen in Fig. 9. It can be observed that once
the voltage collapses, the OLTC or any other voltage restorer requires a
large effort to bring it back into normal limits. Had any voltage reg-
ulator acted timely before the 30 s instant, the effort required to correct
the voltage would have been far less.
The partial voltage collapse due to 90%/30 s ramp down of PV was
also studied, and the result of analysis were similar to 90%/20 s case.
However, for 90%/40 s, the partial voltage collapse is averted due to
the timely operation of OLTC, which expanded the Q-V characteristics
of the network before it lost the stable equilibrium point (s).

4.2. Considering variable speed drive loads

In this case-study, the IM is replaced by a VSD load of the same


rating in the simplified distribution network. The considered VSD
configuration is described in Section 3.3. The previous case study with
IM showed that the voltage collapse is more likely under the high ramp-
rate scenario of PV sources. Therefore, the voltage stability analysis for
the rest of the paper will be based on 90%/20 s cases of PV ramp. Fig. 10. Distribution network with VSD load simulated for various PV ramping
down events.

8
S. Maharjan, et al. Applied Energy 264 (2020) 114733

Fig. 11. Portraying the stable operating point at various instances of 90%/20 s
ramp down of PV in the distribution network.

dynamics, the OLTC aids to expand the network Q-V characteristics,


which eventually brought the voltage under the regulating limit. Fig. 11
depicts the movement of a stable operating point at various instances of
90%/20 s PV ramping down event. For other ramping down cases, the
stability analysis curves are not shown as they exhibit behaviour similar
to 90%/20 s ramp down case.
Inference: Section 4 analyzed voltage stability of a simplified dis-
tribution feeder considering two types of dynamic loads, namely IM and Fig. 12. The United Kingdom General Distribution Grid with feeder classifica-
VSD loads, during severe PV ramping events. The stability analysis was tion.
conducted first with RMS simulation in DigSILENT PowerFactory and
then verified with Q-V analysis. The first analysis conducted with IM modeled using PQ controlled inverter in DigSILENT. The various sce-
load show that the PV ramps of 90%/20 s can lead to partial voltage nario of PV penetration is created using the following definition of PV
collapse. The reason behind the partial voltage collapse are due to Q-V penetration as in [45].
characteristics of IM and slow acting nature of OLTC. However, the
second analysis show that severe PV ramps do not cause voltage in- Peak PV power
stability when IM is replaced by VSD. As VSD drives have energy sto- PV penetration =
Peak load Apparent Power (32)
rage at its DC link, it partially supports the active power to motor load
during the voltage falling period. In addition, the Q-V characteristics of The load model comprises of the dynamic load (composition of IM
VSD do not have stalling region where it has abrupt reactive power and VSD) and static loads (representative model of aggregated heating,
demand. Both these characteristics of VSD reduces the rate of voltage lighting and electronic loads). The important aspect of load modeling is
decline during severe PV ramping condition, which allows the slow the composition of the static and dynamic load, which is specific to
acting controller like OLTC to respond and restore the network voltage each customer type [37]. Certain research work such as [18,46] assume
before it tends to collapse. the composition of static and dynamic load to be fixed at 25% and 75%.
However, this paper uses the load composition as studied in [40] for a
5. Investigation of PV ramping impacts on voltage stability of modern city for various customer types, which are listed in Table 4. The
UKGDS network ZIP parameters for the aggregate static load and IM parameters for
various customer types considered are listed in Tables 4 and 2 respec-
The UKGDS is an urban distribution network, operating at 11 kV, tively.
which is connected to a 33 kV grid using two distribution transformers, The worst PV ramping impacts on UKGDS network is studied using
each having a rating of 26.4 MVA, as shown in Fig. 12. At the point of RMS simulation in DigSILENT under normal and grid contingent con-
connection to the 33 kV grid, the fault MVA is considered to be five ditions. As voltage stability is affected by load models, the investigation
times the rated capacity of the substation (i.e., 264 MVA) [43]. The is carried out at two different compositions of loads. The first one is the
distribution transformer is equipped with OLTC and its parameters are composition of IM and static load. However, IMs are being replaced by
listed in Table 1. Its control configuration is the same as described for VSDs and are expected to grow in the future. To represent this scenario,
the simplified distribution network Section 4. The feeders of the the second load composition which comprises of IM, VSD and static
UKGDS, as seen in Fig. 12, are classified into residential, commercial loads is considered to investigate the impact of PV ramping on voltage
and industrial feeders. The classification was performed as per the an- stability in a realistic distribution network.
nual electricity consumption of the corresponding customers in London, The voltage stability under severe PV ramping condition is in-
whose distribution are 35%, 25% and 40% respectively [44]. Mean- vestigated for the following cases:
while, the peak load at each bus is shown in Table 3. The PV sources are

9
S. Maharjan, et al. Applied Energy 264 (2020) 114733

Table 3
Peak loads in the UKGDS network.
Feeder Nodes Modified Nodes Modified
load (MW) load (MW)

F1 1101–1102 1.117 + j0.227 1103 0.851 + j0.173


F2 1104–1105 1.117 + j0.227 1106 0.851 + j0.173
F3 1107–1108 0.887 + j0.180 1109 0.620 + j0.126
F4 1110–1112 0.670 + j0.136 1113–1114 0.898 + j0.182
F5 1115–1121 0.701 + j0.142 1122–1125 0.496 + j0.100
F6 1126–1132 0.410 + j0.083 1133–1136 0.205 + j0.041
F7 1137–1145 0.418 + j0.085 1146–1150 0.210 + j0.043
F8 1151–1166 0.330 + j0.067 1167–1175 0.215 + j0.044

Table 4
Load composition in various types of feeders and their characteristics [40].
Feeder Load composition Characteristics Active Dynamic
Type Static Dynamic (Z,I,P) Reactive (Z,I,P)

R 34 66 (1.56, −2.49, 1.93) IM


(10.1, −16.72, 7.65)
C 25 75 (0.4, −0.41, 1.01) IM
(4.43, −7.98, 4.56)
I 8 92 (1.21, −1.61, 1.41) IM
(4.35, −7.08, 3.72)
Fig. 13. (a) PV ramping down at 90%/20 s. Voltage at different bus locations at
∗R: Residential, C: commercial, I: industrial feeder. (b) 25% and (c) 35% PV penetration during PV ramping events.

• Composite load comprising of IM and static loads under: supplying the substation [47]. Due to faults or yearly maintenance, one
– Normal grid conditions and of the sub-transmission lines may be out of service. This is referred to as
– Grid contingency conditions a grid contingency period in this paper. Under such a condition, the grid
• Composite load comprising of IM, VSD and static loads under: impedance nearly doubles. The impact of PV ramping for this scenario
– Grid contingency conditions is now evaluated. For a fair comparison, the same PV ramping magni-
tude and loading conditions as in the case of normal grid conditions are
5.1. Considering composite-load comprising of IM and static loads utilized.
Firstly, the PV ramping was simulated at a 25% penetration level.
The load composition ratio of static and IM loads for the residential, The simulation results, as shown in Fig. 14(a), demonstrate that the
commercial and industrial customers have been adopted from [40]. It voltage at certain buses such as 1114, 1136, 1150 and 1175 is below
shows that 92% of the industrial loads and 66% of residential loads are 0.85 p.u. for more than 2 s. Voltage dips of such magnitude and
motor loads. The IM parameters and ZIP parameters for static loads are duration cause the distributed PV to disconnect (as per the PV dis-
shown in Table 2 and 4 respectively. With this load composition, the PV connection rule in [48]), which could further worsen the grid voltage.
ramping impact on voltage stability of UKGDS network is studied under Secondly, the PV penetration was raised to 35%, and the PV ramping
normal and grid contingent condition. impact was again investigated. The results are shown in Fig. 14(b). In
this case, the PV ramping caused a partial voltage collapse in the dis-
5.1.1. Impacts of PV ramping during normal grid conditions tribution network, thereby lowering the voltage at all the buses below
The external grid is characterized by a grid impedance ( Xg ) which 0.5 p.u. Although the OLTC provided a delayed response, it was not
remains fairly constant during normal conditions. It is estimated as: effective in bringing the voltage back to the normal limits.
(Base Voltage )2
Xg = .
Fault MVA (33)
5.2. Considering composite-load comprising of IM, VSD and static loads
During normal grid conditions, the fault MVA or short circuit ratio at
the substation remains at the rated level. As as result, it has higher Q-V VSD is gradually replacing IM due to its superiority in performance
capability at a given power demand. The worst PV ramping of 90% in and energy saving during the operation [23]. The current VSD pene-
20 s was simulated in UKGDS network using RMS simulation, where the tration over the motor loads in residential customers is 40% [49]
load composition comprised of IM and static loads. The simulation is whereas, it is only 16% in commercial and industrial customers [50].
conducted for various PV penetration levels defined by (32). The node However, according to [23], the VSD load is anticipated to take over
voltages of distribution network during a PV ramping event of 90%/ 50% of the motor load mainly in commercial and industrial customers
20 s at 25% and 35% PV penetration are shown if Fig. 13(b) and (c) in European cities. Hence, it is imperative to assess the impact of in-
respectively. Under normal grid conditions, this (worst) ramping did creasing PV penetration and the change in customer load composition
not create any significant impact to the UKGDS network in either the on the distribution voltage stability.
25% or 35% PV penetration levels. Although small voltage drops are Considering the anticipated VSD penetration in 2020 from [23],
observed for some duration, at the buses located at the far end of the 50% of the motors pertaining to the commercial and industrial loads
feeders, they were within the EN50160 regulation limits [41]. Subse- shown in Table 4 are modeled as VSDs, and the rest 50% as IM. The
quently, those buses are regulated by the OLTC. residential VSD penetration is however kept at the current value of 40%
[49]. This is because the motor load is more dominant in the com-
5.1.2. Impacts of PV ramping during grid contingency mercial and industrial sector than among the residential customers.
At the 33 kV level, there will be at least two sub-transmission lines Additionally, regulations such as the European Commission (EC) 640/

10
S. Maharjan, et al. Applied Energy 264 (2020) 114733

Inference: Section 5 conducted voltage stability analysis in UKGDS


network using RMS simulation during the 90%/20 s PV ramping events.
To highlight the impact of load responses on voltage stability, the
analysis was performed for (a) composite-load of IM and static load
(first type), and (b) composite-load of IM, VSD, and static load (second
type). In a realistic distribution network like UKGDS with the first type
of composite load, severe PV ramps caused voltage collapse only at grid
contingency condition above 35% PV penetration level. However, the
voltage collapse is averted in UKGDS network with second type of
composite load, which was formed by replacing 50% IM. Corroborated
by the analytical study in Section 4, larger VSD proportion in customer
load reduced the rate of voltage fall during PV ramping events in
UKGDS network, which allowed the slow acting controller like OLTC to
restore the network voltage.

6. Minimum VSD penetration policy for distribution network

Fig. 14. Voltage at different bus locations at (a) 25% (b) 35% PV penetration It is revealed in Section 5.2 that the VSD loads aids to avert partial
during PV ramping down (90%/20 s) events during grid contingency condition. voltage collapse during the worst PV ramping events. VSD achieves it
by partially reducing the active power consumption from the grid
during voltage emergency conditions and locally supports the load
power from its storage in DC link. This process happens naturally due to
VSD dynamics. Replacement of stiff loads like IM by VSD would be
beneficial to the grid and alleviate the impact of PV ramps. Although
there are solutions for voltage regulation by deploying local voltage
controllers in PV inverter, it should be realized that reactive compen-
sation would be less effective than active power compensation in a
distribution network with a high R/X ratio. Although IM would not be
replaced quickly by VSD in the existing distribution network, policies
favoring the VSD would certainly help. Therefore, the paper aims to
quantify the minimum VSD penetration required to alleviate partial
voltage stability problems in UKGDS network at several PV penetration
levels. It is to be noted that the VSD penetration signifies the percentage
of VSDs as motor loads.
Many cities have declared their PV penetration targets in the next
5–10 years, and the governments are facilitating it with renewable
energy policies [2]. To accomodate future PV penetration, the dis-
tribution network operator could plan the VSD penetration target if it
can quantify the minimum VSD percentage required at their targeted
PV penetration level. Therefore the minimum VSD percentage required
Fig. 15. Voltage at different bus locations at (a) 25%, and (b) 35% PV pene- in the UKGDS network has been determined for various PV penetration
tration during ramping (90%/20 s) events during grid contingency conditions targets. The corresponding results are depicted in Fig. 16.
considering 50% VSD load in the commercial and industrial customers.

2009 mainly apply to the industrial and commercial motors only [23].
Therefore the resulting composition of VSD in dynamic loads is 40%,
50%, and 50% in residential, commercial and industrial customers,
respectively. The ratio of static load is kept intact as shown in Table 4.
As the voltage stability of the distribution network is weak during
grid contingent conditions, the impact of PV ramping at various PV
penetration level are studied for the same case. The composite load
with VSD was assigned to each load of UKGDS network, and the si-
mulated result for 25% and 35% PV penetration level are shown in
Fig. 15 (a) and (b). In both the scenarios of 25% and 35% PV pene-
tration, voltage reduced slowly than the rate at which PVs were
ramped. In both penetration levels, the worst ramp down of PV created
the maximum voltage dip which is not below 0.85 p.u. This voltage dip
would neither disconnect the PV generator (as per the PV disconnection
rule in [48]) nor trip the VSD load, as seen in Fig. 15. After a mo-
mentary voltage violation, the OLTC action regulated the voltages back
to the statutory limit. These simulations reveal that the presence of VSD
loads aids the distribution network in averting a partial voltage collapse
Fig. 16. Minimum VSD penetration requirement at various PV penetration le-
during weak grid conditions. Additionally, it also prevents the tripping
vels to avert voltage collapse in the UKGDS network due to PV ramping under
of the network loads.
grid contingencies.

11
S. Maharjan, et al. Applied Energy 264 (2020) 114733

Fig. 16 shows that, for 25% PV penetration target, 10% VSD in the partial voltage collapse in UKGDS network due to PV ramps was
industrial and commercial customer would be sufficient for voltage averted. On identifying its grid supportive nature, the paper finally
security against severe PV ramps. Similarly, the minimum VSD pene- quantified the minimum VSD penetration required for various PV pe-
tration is 70% at 45% PV penetration level. For any PV penetration netration targets in UKGDS network. This quantification would be
target, the VSD penetration target should lie in the yellow zone of the useful for system operators to formulate suitable VSD penetration po-
Fig. 16 for ensuring voltage stability against PV ramps. licies to ensure voltage stability in their network as they set future PV
penetration targets.
7. Conclusion The local voltage control from PVs could have lessened the PV
ramping impacts on voltage stability, however, this is still not practiced
This paper analyzed the partial voltage collapse in a distribution in many countries. Nevertheless, the VSDs in the distribution network
network caused by severe PV ramps. It was demonstrated in a simplified would bolster the voltage stability along with local voltage control from
distribution network as well as in realistic distribution network like PVs, where local voltage control is practised. This combined contribu-
UKGDS. The RMS simulation from DigSILENT was deployed as the main tion of VSD and local voltage control on short-term voltage stability of
tool of study. In addition, simulation results were further corroborated the distribution network will be studied as our future work.
using analytical stability analysis like Q-V analysis. For Q-V analysis, Q-
V characteristics of the simplified distribution feeder were mathema-
Declaration of Competing Interest
tically modeled considering OLTC and R/X element of the distribution
line. Additionally, the Q-V characteristics of IM and VSD loads were
The authors declare that they have no known competing financial
also modeled. These models were used to identify the cause of partial
interests or personal relationships that could have appeared to influ-
voltage collapse observed in the simulation. It is determined from the
ence the work reported in this paper.
analysis that the primary cause for partial voltage collapse is due to
slow-acting voltage controllers in the network and stiff loads like IM.
Acknowledging the anticipated VSD penetration data in 2020, the CRediT authorship contribution statement
paper further investigated the voltage stability considering the pro-
liferation of VSD loads. From the analysis, it is observed that VSDs’ Salish Maharjan: Writing - original draft, Data curation, Software,
static and dynamic load characteristics are beneficial to the grid during Methodology, Conceptualization. Dhivya Sampath Kumar: Writing -
PV ramping events as they support the grid during voltage falling review & editing, Investigation, Visualization. Ashwin M.
period by reducing the power intake from the grid partially and com- Khambadkone: Writing - review & editing, Conceptualization,
pensate from its local storage. When the IMs are replaced by VSD, the Supervision.

Appendix A. Line and bus data of UKGDS network

See Table 5.

Table 5
Network data for UKGDS [47].
From bus To bus Line Impedance Transformer Impedance parallel load bus Load

R pu X pu R pu X pu P (MW) Q (MVAR)

301 1100 – – 0.04707 0.65409 2 301 – –


1100 1101 0.2038 0.1056 – – 1 1101 1.117 0.227
1101 1102 0.2038 0.1056 – – 1 1102 1.117 0.227
1102 1103 0.0624 0.017 – – 1 1103 0.851 0.173
1100 1104 0.2038 0.1056 – – 1 1104 1.117 0.227
1104 1105 0.2038 0.1056 – – 1 1105 1.117 0.227
1105 1106 0.0624 0.017 – – 1 1106 0.851 0.173
1100 1107 0.2038 0.1056 – – 1 1107 0.887 0.18
1107 1108 0.2038 0.1056 – – 1 1108 0.887 0.18
1108 1109 0.0624 0.017 – – 1 1109 0.62 0.126
1100 1110 0.266 0.1378 – – 1 1110 0.67 0.136
1110 1111 0.266 0.1378 – – 1 1111 0.67 0.136
1111 1112 0.266 0.1378 – – 1 1112 0.67 0.136
1111 1113 0.0663 0.018 – – 1 1113 0.898 0.182
1112 1114 0.0663 0.018 – – 1 1114 0.898 0.182
1100 1115 0.0745 0.0574 – – 1 1115 0.701 0.142
1115 1116 0.0745 0.0574 – – 1 1116 0.701 0.142
1116 1117 0.0745 0.0574 – – 1 1117 0.701 0.142
1117 1118 0.0745 0.0574 – – 1 1118 0.701 0.142
1118 1119 0.0745 0.0574 – – 1 1119 0.701 0.142
1119 1120 0.0745 0.0574 – – 1 1120 0.701 0.142
1120 1121 0.0745 0.0574 – – 1 1121 0.701 0.142
1116 1122 0.0542 0.0147 – – 1 1122 0.496 0.101
1118 1123 0.0542 0.0147 – – 1 1123 0.496 0.101
1119 1124 0.0542 0.0147 – – 1 1124 0.496 0.101
1121 1125 0.0542 0.0147 – – 1 1125 0.496 0.101
1100 1126 0.0745 0.0574 – – 1 1126 0.41 0.083
1126 1127 0.0745 0.0574 – – 1 1127 0.41 0.083
1127 1128 0.0745 0.0574 – – 1 1128 0.41 0.083
(continued on next page)

12
S. Maharjan, et al. Applied Energy 264 (2020) 114733

Table 5 (continued)

From bus To bus Line Impedance Transformer Impedance parallel load bus Load

R pu X pu R pu X pu P (MW) Q (MVAR)

1128 1129 0.0745 0.0574 – – 1 1129 0.41 0.083


1129 1130 0.0745 0.0574 – – 1 1130 0.41 0.083
1130 1131 0.0745 0.0574 – – 1 1131 0.41 0.083
1131 1132 0.0745 0.0574 – – 1 1132 0.41 0.083
1127 1133 0.0542 0.0147 – – 1 1133 0.205 0.042
1129 1134 0.0542 0.0147 – – 1 1134 0.205 0.042
1130 1135 0.0542 0.0147 – – 1 1135 0.205 0.042
1132 1136 0.0542 0.0147 – – 1 1136 0.205 0.042
1100 1137 0.0917 0.0706 – – 1 1137 0.418 0.085
1137 1138 0.0917 0.0706 – – 1 1138 0.418 0.085
1138 1139 0.0917 0.0706 – – 1 1139 0.418 0.085
1139 1140 0.0917 0.0706 – – 1 1140 0.418 0.085
1140 1141 0.0917 0.0706 – – 1 1141 0.418 0.085
1141 1142 0.0917 0.0706 – – 1 1142 0.418 0.085
1142 1143 0.0917 0.0706 – – 1 1143 0.418 0.085
1143 1144 0.0917 0.0706 – – 1 1144 0.418 0.085
1144 1145 0.0917 0.0706 – – 1 1145 0.418 0.085
1138 1146 0.0571 0.0155 – – 1 1146 0.21 0.043
1140 1147 0.0571 0.0155 – – 1 1147 0.21 0.043
1141 1148 0.0571 0.0155 – – 1 1148 0.21 0.043
1143 1149 0.0571 0.0155 – – 1 1149 0.21 0.043
1145 1150 0.0571 0.0155 – – 1 1150 0.21 0.043
1100 1151 0.0665 0.0512 – – 1 1151 0.33 0.067
1151 1152 0.0665 0.0512 – – 1 1152 0.33 0.067
1152 1153 0.0665 0.0512 – – 1 1153 0.33 0.067
1153 1154 0.0665 0.0512 – – 1 1154 0.33 0.067
1154 1155 0.0665 0.0512 – – 1 1155 0.33 0.067
1155 1156 0.0665 0.0512 – – 1 1156 0.33 0.067
1156 1157 0.0665 0.0512 – – 1 1157 0.33 0.067
1157 1158 0.0665 0.0512 – – 1 1158 0.33 0.067
1158 1159 0.0665 0.0512 – – 1 1159 0.33 0.067
1159 1160 0.0665 0.0512 – – 1 1160 0.33 0.067
1160 1161 0.0665 0.0512 – – 1 1161 0.33 0.067
1161 1162 0.0665 0.0512 – – 1 1162 0.33 0.067
1162 1163 0.0665 0.0512 – – 1 1163 0.33 0.067
1163 1164 0.0665 0.0512 – – 1 1164 0.33 0.067
1164 1165 0.0665 0.0512 – – 1 1165 0.33 0.067
1165 1166 0.0665 0.0512 – – 1 1166 0.33 0.067
1152 1167 0.0729 0.0198 – – 1 1167 0.215 0.044
1154 1168 0.0729 0.0198 – – 1 1168 0.215 0.044
1155 1169 0.0729 0.0198 – – 1 1169 0.215 0.044
1157 1170 0.0729 0.0198 – – 1 1170 0.215 0.044
1159 1171 0.0729 0.0198 – – 1 1171 0.215 0.044
1161 1172 0.0729 0.0198 – – 1 1172 0.215 0.044
1162 1173 0.0729 0.0198 – – 1 1173 0.215 0.044
1164 1174 0.0729 0.0198 – – 1 1174 0.215 0.044
1166 1175 0.0729 0.0198 – – 1 1175 0.215 0.044

system base voltage = 11 kV, base power = 100 MW

References UKSim-AMSS 9th IEEE European modelling symposium on computer modelling and
simulation 2016. p. 260–5. https://doi.org/10.1109/EMS.2015.46.
[9] Wang L, Yan R, Saha TK. Voltage regulation challenges with unbalanced PV in-
[1] DigSILENT. PowerFactory. http://www.digsilent.de/index.php/products- tegration in low voltage distribution systems and the corresponding solution. Appl
powerfactory.html. Energy 2019;256(June):113927. https://doi.org/10.1016/j.apenergy.2019.
[2] Renewable2016 Global Status Report. Tech. rep. Renewable Energy Policy Network; 113927.
2016. URL http://www.ren21.net/resources/publications/. [10] Nieto A, Vita V, Maris TI. Power quality improvement in power grids with the
[3] Marcos J, Marroyo L, Lorenzo E, Alvira D, Izco E. Power output fluctuations in large integration of energy storage systems. Int J Eng Res Technol (IJERT)
scale pv plants: one year observations with one second resolution and a derived 2016;5(07):438–43.
analytic model. Prog Photovoltaics Res Appl 2011;19(2):218–27. https://doi.org/ [11] Craciun BI, Kerekes T, Séra D, Teodorescu R. Overview of recent drid codes for PV
10.1002/pip.1016. power integration. Proceedings of the international conference on optimisation of
[4] Zhao Q, Wu K, Khambadkone AM. Optimal sizing of energy storage for PV power electrical and electronic equipment, OPTIM 2012. p. 959–65. https://doi.org/10.
ramp rate regulation. In: ECCE 2016 - IEEE energy conversion congress and ex- 1109/OPTIM.2012.6231767.
position, proceedings. doi:10.1109/ECCE.2016.7855318. [12] Glavic M, Novosel D, Heredia E, Kosterev D, Salazar A, Habibi-Ashrafi F, et al. See it
[5] Cutsem T, Vournas C. Voltage stability of electric power systems. Boston, MA: fast to keep calm: real-time voltage control under stressed conditions. IEEE Power
Springer US; 1998. https://doi.org/10.1007/978-0-387-75536-6. Energ Mag 2012;10(4):43–55. https://doi.org/10.1109/MPE.2012.2196332.
[6] Engler A. Applicability of droops in low voltage grids. Int J Distrib Energy Resour [13] Kawabe K, Tanaka K. Impact of dynamic behavior of photovoltaic power generation
Smart Grids 2005(1):1–5. systems on short-term voltage stability. IEEE Trans Power Syst
[7] Chamana M, Chowdhury BH. Optimal voltage regulation of distribution networks 2015;30(6):3416–24. https://doi.org/10.1109/TPWRS.2015.2390649.
with cascaded voltage regulators in the presence of high PV penetration. IEEE Trans [14] Kawabe K, Ota Y, Yokoyama A, Tanaka K. Novel dynamic voltage support capability
Sustain Energy 2018;9(3):1427–36. https://doi.org/10.1109/TSTE.2017.2788869. of photovoltaic systems for improvement of short-term voltage stability in power
[8] Vita V, Alimardan T, Ekonomou L. The impact of distributed generation in the systems. IEEE Trans Power Syst 2017;32(3):1796–804. https://doi.org/10.1109/
distribution networks’ voltage profile and energy losses. Proceedings - EMS 2015: TPWRS.2016.2592970.

13
S. Maharjan, et al. Applied Energy 264 (2020) 114733

[15] Islam M, Nadarajah M, Hossain J. Short-term voltage stability enhancement in re- 2000;36(3):904–10. https://doi.org/10.1109/28.845069.
sidential grid with high penetration of rooftop PV units. IEEE Trans Sustain Energy [32] E. ELECTRICAL, Harmonic Distortion from Variable Frequency Drives; 2012.
2018;10(4):2211–22. https://doi.org/10.1109/TSTE.2018.2883453. p. 1–8.
[16] Dong Y, Xie X, Shi W, Zhou B, Jiang Q. Demand-response-based distributed pre- [33] Krause PC, Wasynczuk O, Sudhoff SD. Analysis of electric machinery and drive
ventive control to improve short-term voltage stability. IEEE Trans Smart Grid system. Wiley; 2002.
2018;9(5):4785–95. https://doi.org/10.1109/TSG.2017.2670618. [34] Maitra A, Gaikwad A, Zhang P, Ingram M, Mercado DL, Woitt WD. Using system
[17] Yaghoobi J, Mithulananthan N, Saha TK. Dynamic voltage stability of distribution disturbance measurement data to develop improved load models. In: 2006 IEEE PES
system with a high penetration of rooftop PV units. IEEE power and energy society power systems conference and exposition, PSCE 2006 - Proceedings; 2006. p.
general meeting 2015-September 2015. p. 1–5. https://doi.org/10.1109/PESGM. 1978–85. doi: 10.1109/PSCE.2006.296230.
2015.7286608. [35] Models SL, Flow P, Simulation DP. Standard load models for power flow and dy-
[18] Yan R, Saha TK. Investigation of voltage stability for residential customers due to namic performance simulation. IEEE Trans Power Syst 1995;10(3):1302–13.
high photovoltaic penetrations. IEEE Trans Power Syst 2012;27(2):651–62. https:// https://doi.org/10.1109/59.466523.
doi.org/10.1109/TPWRS.2011.2180741. [36] Price W, Wirgau K, Murdoch A, Mitsche J, Vaahedi E, El-Kady M. Load modeling for
[19] Paramasivam M, Salloum A, Ajjarapu V, Vittal V, Bhatt NB, Liu S. Dynamic opti- power flow and transient stability computer studies. IEEE Trans Power Syst
mization based reactive power planning to mitigate slow voltage recovery and short 1988;3(1):180–7. https://doi.org/10.1109/59.43196.
term voltage instability. IEEE Trans Power Syst 2013;28(4):3865. https://doi.org/ [37] Milanović JV, Yamashita K, Martínez Villanueva S, Djokić S, Korunović LM.
10.1109/TPWRS.2013.2271260. International industry practice on power system load modeling. IEEE Trans Power
[20] Xu Y, Dong ZY, Meng K, Yao WF, Zhang R, Wong KP. Multi-objective dynamic VAR Syst 2013;28(3):3038–46. https://doi.org/10.1109/TPWRS.2012.2231969.
planning against short-term voltage instability using a decomposition-based evo- [38] Quilumba FL, Lee WJ, Jativa-Ibarra J. Load models for flat-panel TVs. IEEE Trans
lutionary algorithm. IEEE Trans Power Syst 2014;29(6):2813–22. https://doi.org/ Ind Appl 2014;50(6):4171–8. https://doi.org/10.1109/TIA.2014.2313660.
10.1109/TPWRS.2014.2310733. [39] Lu N, Xie Y, Huang Z, Puyleart F, Yang S. Load component database of household
[21] Marcos J, Storkël O, Marroyo L, Garcia M, Lorenzo E. Storage requirements for PV appliances and small office equipment. In: IEEE power and energy society 2008
power ramp-rate control. Sol Energy 2014;99:28–35. https://doi.org/10.1016/j. general meeting: conversion and delivery of electrical energy in the 21st century,
solener.2013.10.037. PES; 2008. p. 1–5. doi: 10.1109/PES.2008.4596224.
[22] Sukumar S, Mokhlis H, Mekhilef S, Karimi M, Raza S. Ramp-rate control approach [40] Bokhari A, Alkan A, Dogan R, Diaz-Aguilo M, De Leon F, Czarkowski D, et al.
based on dynamic smoothing parameter to mitigate solar PV output fluctuations. Int Experimental determination of the ZIP coefficients for modern residential, com-
J Electr Power Energy Syst 2018;96(October 2017):296–305. https://doi.org/10. mercial, and industrial loads. IEEE Trans Power Delivery 2014;29(3):1372–81.
1016/j.ijepes.2017.10.015. https://doi.org/10.1109/TPWRD.2013.2285096.
[23] Energy Efficiency with Electric Drive Systems, Tech. rep., European Committee of [41] CENELEC, En 50160, European Standard; 2005. p. 1–20.
Manufacturers of Electrical Machines and Power Electronics; 2015. URL http:// [42] Yan R, Marais B, Saha TK. Impacts of residential photovoltaic power fluctuation on
cemep.eu/data/CEMEP%5C_Energy%5C_Efficiency%5C_with%5C_Electric%5C_ on-load tap changer operation and a solution using DSTATCOM. Electr Power Syst
Drive%5C_Systems.pdf? Res 2014;111:185–93. https://doi.org/10.1016/j.epsr.2014.02.020.
[24] Machowski J, Bialek JW, Bumby JR. Power system dynamics. John Wiley and Sons [43] Calculation of System Fault Levels, Tech. Rep. 2, SP Energy Networks; 2008. URL
Ltd.; 2008. https://www.spenergynetworks.co.uk/userfiles/file/ESDD-02-006.pdf.
[25] Kundur P. Power system stability and control. McGraw-Hill; 1994. [44] Sub-National Electricity and Gas Consumption Statistics. Tech. Rep. December,
[26] Sharifzadeh M, Raoufi H. Effect of induction motors mechanical load on its model Department for Business, Energy & Industrial Strategy; 2016. URL https://www.
for purpose of static voltage stability analysis. In: 2011 19th Iranian conference on gov.uk/government/uploads/system/uploads/attachment_data/file/579203/Sub-
electrical engineering. ICEE; 2011. national_electricity_and_gas_consumption_summary_report_2016.pdf.
[27] Patel A, Wedeward K, Smith M. Parameter estimation for inventory of load models [45] Hoke A, Butler R, Hambrick J, Kroposki B. Maximum photovoltaic penetration le-
in electric power systems. In: World congress on engineering and computer science, vels on typical distribution feeders. IEEE Trans Sustain Energy 2012(July):1–14.
vol. 1; 2014. p. 233–8. [46] Aristidou P, Valverde G, Van Cutsem T. Contribution of distribution network con-
[28] Epperly RA, Hoadley FL. Considerations when applying ASD’s in continuous pro- trol to voltage stability: a case study. IEEE Trans Smart Grid 2017;8(1):106–16.
cesses. IEEE Trans Ind Appl 1997;33(2):389–96. https://doi.org/10.1109/28. https://doi.org/10.1109/TSG.2015.2474815.
568000. [47] Centre for Sustainable Electricity and Distributed Generation. https://github.com/
[29] Liang X, Xu W. Aggregation method for motor drive systems. Electr Power Syst Res sedg/ukgds.
2014;117:27–35. https://doi.org/10.1016/j.epsr.2014.07.022. [48] Bründlinger R. Grid Codes in Europe for Low and Medium Voltage: an overview. In:
[30] Wymann T, Jorg P. Power loss ride-through in a variable speed drive system. In: Conference: 6th international conference on integration of renewable and dis-
11th petroleum and chemical industry conference Europe electrical and in- tributed energy resources (IRED), At Kyoto. doi:10.13140/2.1.2045.2167. http://
strumentation applications, PCIC; 2014. doi: 10.1109/PCICEurope.2014.6900057. www.nedo.go.jp/english/ired2014/program/pdf/s2/s2_4_roland_bruendlinger.pdf.
[31] Bollen MH, Zhang LD. Analysis of voltage tolerance of AC adjustable-speed drives [49] Fourth Quarter FY 2015 Quarterly Update; 2015.
for three-phase balanced and unbalanced sags. IEEE Trans Ind Appl [50] Meza M. Industrial LV Motors & Drives: A Global Market Update; 2014.

14

You might also like