Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

1

Comparison of Voltage Control Methods in


Distribution Systems Using Q-V Based PI and
Droop Controls of Solar Inverters
Severin Nowak1, Student Member, IEEE, Malcolm S Metcalfe1, Member, IEEE,
Wilson Eberle1, Member, IEEE and Liwei Wang1, Member, IEEE
1. The University of British Columbia, Okanagan, Canada, liwei.wang@ubc.ca

Abstract—The increasing penetration of photovoltaic power Alliance [6], which consists of the major industry participants
generation in distribution systems causes serious voltage of the sector. Recent PV inverters have the capability of
management issues. To mitigate the voltage variations due to solar generating real and reactive power independently as long as the
generation intermittency, this paper introduces a PI-based
reactive power control method of PV inverters. The proposed PI
maximal apparent power rating (ܵ௠௔௫ ) is not exceeded. While
controller adjusts the reactive power injection of the solar the real power generation is directly linked to the solar
inverters dynamically to drive the voltage at the Point of Common irradiation, there are various strategies to control the reactive
Coupling (PCC) to a target value. The simulation studies are power generation.
performed to evaluate the proposed PI-based reactive power The IEC and IEEE standards propose the fixed Power Factor
control using the IEEE 34 test feeder for a 24-hour period in (PF) mode where the PV inverters operate at a constant PF. In
OpenDSS and Matlab. The performance of the proposed PI this mode, the reactive power generation is directly linked to
controller is compared to the conventional PV inverter controls the real power generation [4]. Alternatively, the constant
with no reactive power generation and with the Q-V-based droop
reactive power mode can be used in which the reactive power
control at different penetration levels. The case studies
demonstrate that the proposed PI-based voltage control method
of the inverter is independent of the real power generation and
reduces effectively the voltage deviations at the PCC of the PV the voltage at the PCC, given that the inverter’s ܵ௠௔௫ is not
inverters. The voltage profiles under the proposed PI controller exceeded. The third reactive power control mode of a PV
have narrower variation bands compared with the conventional inverter is the so-called reactive-power-voltage (Q-V)-based
PV reactive power control methods particularly at high droop control mode in which the inverter supplies or absorbs Q
penetration levels of solar generation. with respect to the inverter AC voltage at the PCC (VPCC) [4].
There are several Q-V-based droop control modes: 1) the linear
Index Terms—Reactive power control, Voltage control, Solar Q-V droop control, where the Q generation is a linear function
power generation, Distributed power generation, Power
of the inverter voltage at the PCC; 2) the Q-V droop control
distribution.
with a dead-band at the reference voltage; and 3) the Q-V droop
control with a hysteresis model where the Q generation of the
I. INTRODUCTION
inverter is different for rising and falling voltages at the PCC.
E lectric utilities have serious concerns about the reliability of
power systems due to the increasing amounts of
photovoltaic (PV) generation directly connected to the
In addition to the control modes introduced by the standards,
the literature suggests several approaches for the reactive power
control with no communication channels. In [7], a reactive
Distribution System (DS) [1]. The Distributed Generation (DG) power control approach is introduced where a PV inverter
at customer sites not only impacts the utilities energy sales, but regulates its reactive power injection based on the P generation
also affects the power quality within the DS. The intermittent and the characteristic of the transmission line connected to the
and undispatchable real power (P) generation of PV plants leads PV inverter. The purpose is not to control the node voltage but
to significant voltage variations. Voltage variations are rather to guarantee that active power generation does not cause
proportional to the power generation of PV plants. In worst case voltage rise/drop locally. Voltage becomes almost independent
scenarios this can lead to violations of the ANSI voltage limits of P generation. The references [8, 9] present an approach to
(+/- 5%) and can also create reverse power flow in the DS [2, implement local control actions for DG units without the use of
3]. In addition, PV generation can interfere with conventional communication channels. The DG units are controlled to
voltage control mechanisms such as on-load tap changers, in- operate at an optimal point between active power curtailment
line voltage regulators, and Static Var Compensators (SVC), and reactive power generation to maintain the voltage within
causing undesired wear-out of the mechanical switches [2]. the limits.
Therefore, the use of PV inverters to regulate the DS voltages This paper introduces a new approach to regulate the voltage
locally through appropriate reactive power controls have been of a feeder to a target set point by controlling the Q output of
a very active research area. Both IEC (IEC61850-7-420) and the PV inverter. A PI-based inverter voltage control method is
IEEE (IEEE 1547.8.) have introduced standards for the reactive proposed to exploit the maximum available reactive power of
power control modes [4, 5] of PV inverters. Many PV inverters without curtailing real power generation. The
manufacturers of commercial PV inverters have remainder of this paper first specifies the Q-V droop control and
correspondingly adopted these control methods. For example, PV inverter capabilities in Section II, Section III introduces the
system interoperability has been pursued by the SunSpec proposed method, Section IV describes the case studies that

‹,(((
2

were performed in order to evaluate the proposed method. proposed voltage control method eliminates the voltage error
Section V presents the simulation results and the discussion and between the reference voltage and the actual voltage using an
finally the paper is concluded in Section VI. integral controller in addition to the proportional controller,
forming a PI control scheme. The integral controller of the
II. Q-V-BASED DROOP CONTROL METHOD proposed approach compensates for the accumulated error
In the Q-V-based droop control mode, the PV inverter between ܸ௉஼஼ and ܸ௥௘௙ over time and drives the steady-state
absorbs or supplies reactive power in proportion to the error in (3) to zero.
difference of the inverter voltage at PCC (ܸ௉஼஼ ) and the ܸ௘௥௥௢௥ ሺ‫ݐ‬ሻ ൌ ܸ௥௘௙ ሺ‫ݐ‬ሻ െ ܸ௉஼஼ ሺ‫ݐ‬ሻ (3)
reference voltage (ܸ௥௘௙ ሻ, which is generally assumed to be 1.0 The reactive power output of the PV inverter before saturation
pu. The reactive power generated by the inverter (ܳ௢௨௧ ) is a (ܳ ᇱ ) is a function of the input voltage error ܸ௘௥௥௢௥ in (4).
linear function of the ܸ௉஼஼ in pu as given in (1): ௧
ܳ ᇱ ሺ‫ݐ‬ሻ ൌ ݇௣ ቀܸ௘௥௥௢௥ ሺ‫ݐ‬ሻ ൅ ݇௜ ‫׬‬଴ ܸ௘௥௥௢௥ ሺ߬ሻ ή ݀߬ቁ (4)
ܸ௉஼஼ ൏ ͲǤͻͷ ‫ܳ ׷‬௢௨௧ ൌ ܳ௠௔௫
ܳ where ݇௣ is the proportional gain and ݇௜ is the integral gain.
൞ͲǤͻͷ ൑ ܸ௉஼஼ ൑ ͳǤͲͷ ‫ܳ ׷‬௢௨௧ ൌ ܳ௠௔௫ െ ௠௔௫ ሺܸ௉஼஼ െ ͲǤͻͷሻ Clearly, the performance of the proposed PI controller is
ͲǤͲͷ
ܸ௉஼஼ ൐ ͳǤͲͷ ‫ܳ ׷‬௢௨௧ ൌ െܳ௠௔௫ constrained by the available reactive power of the PV inverter.
(1) Whenever ܳ ᇱ reaches ܳ௔௩௔௜௟௔௕௟௘ as in (2), the inverter voltage
where ܳ௠௔௫ is the maximum available Q of the inverter. ܸ௉஼஼ is not able to track ܸ௥௘௙ . Hence ܸ௘௥௥௢௥ does not reduce
The Q-V droop control is essentially a proportional control towards zero which can result in an integral windup. Therefore,
scheme having steady-state error (SSE) between ܸ௉஼஼ and ܸ௥௘௙ . an anti-windup technique is included in the proposed PI control.
The real power generation of the PV inverter only uses its full The integrator term is clamped when (5) is activated which is
capacity (i.e. ܵ௠௔௫ ) for a few hours around noon on sunny found to perform well under various conditions [11].
days, depending on the PV layout, the weather and the season. ሺܸ௘௥௥௢௥ ሺ‫ݐ‬ሻ ή ܳ ᇱ ሺ‫ݐ‬ሻ ൐ Ͳሻ‫ܦܰܣ‬൫ܳ ᇱ ሺ‫ݐ‬ሻ ് ܳ௢௨௧ ሺ‫ݐ‬ሻ൯ (5)
During cloudy days and nights, there is significant reactive The implementation of the proposed Q-V-based PI voltage
power capacity available from the inverters [10]. The Q control scheme of the PV inverter is illustrated in Figs. 2 and 3.
availability (ܳ௔௩௔௜௟௔௕௟௘ ) is related to the apparent power The PI controller is tuned using the Matlab PID controller
(ܵ௠௔௫ ) and the actual real power (Pact) of the PV inverter as tuning toolbox. For the proposed PI controller, a proportional
gain of ݇௣ = 0.85 and an integral gain of ݇௜ = 200 proved to
ܳ௔௩௔௜௟௔௕௟௘ ሺ‫ݐ‬ሻ ൌ ටܵ௠௔௫ ଶ െ ܲ௔௖௧ ሺ‫ݐ‬ሻଶ (2) offer a fast reaction while providing stability of the closed-loop
system.
The available reactive power of the PV inverter is also Line Line
limited by the maximum PF of the inverter. Most PV inverters impedance impedance
have PF limits around 0.7−0.85 (inductive or capacitive). In this
paper we assume a maximum PF of 0.7 (inductive or capacitive)
Qout
Substation

which leads to the PV inverter’s P-Q diagram shown in Fig. 1. PI AC


Control
Fig. 3 DC
VPCC

Vref
PV
Fig. 2 Single line diagram of radial feeder with PV inverter and PI controller
Vref + Verror Q’ Qout
+
VPCC - kp
+
Fig. 1 a) P/Q capability PV inverter; b) Q-V droop control function 0
1
The PV inverter operates within the shaded area as shown in ki s
0 1
Fig. 1a. Figure 1b illustrates the two commonly used Q-V droop x
controls: linear and deadband droop control modes. The
deadband droop control maintains a unity PF around the
reference voltage of 1pu. The linear Q-V droop control mode is >0
further explained in the next section and will serve as a
reference for the case studies. AND

III. PROPOSED Q-V-BASED PI CONTROL METHOD Fig. 3 Block diagram of Q-V-Based PI Control
The proposed Q-V control method is based on a similar idea The proposed Q-V-based PI voltage control method can be
as the Q-V-based droop control by regulating the VPCC using the conveniently realized in the commercial PV inverters since the
reactive power generation of the PV inverter. However, the Q control option has been embedded into the PV inverters. The

‹,(((
3

PV inverter with integrated PI voltage control will bring the (1) real power only without reactive power generation (unity
advantage of fast reaction to voltage variations by automatic power factor); (2) the linear Q-V-based droop control; and (3)
reactive power generation control. It is also technically possible the proposed Q-V-based PI control. In the second case study,
to implement the proposed PI voltage control method in a the same test scenarios are simulated while the ܵ௠௔௫ of the
microcontroller outside the inverter which controls the inverter inverters is over dimensioned by 50% to provide more reactive
through Modbus commands. However, the latency of the power.
communication protocol between the external microcontroller 1
Commercial load Residential load
and the inverter may lead to communication and control delays.
Therefore, it is desirable to implement the PI controller inside 0.8

normalized power
the PV inverter for better control performance and reliability.
The reference voltage (ܸ௥௘௙ ) can either be set once at 0.6
commissioning or controlled remotely based on feeder
measurement. 0.4

IV. CASE STUDIES 0.2


Solar generation
In order to evaluate the proposed method, a simulation of the
IEEE 34-node test feeder [12] was performed using OpenDSS 0
0 6 12 18 24
and GridPV Toolbox in Matlab, a commonly-used distribution
grid simulation toolset [13]. The single line diagram of the Time (hours)
Fig. 5 Normalized load profiles and solar generation over 24 hours
IEEE 34-node test feeder is shown in Fig. 4 which is an actual
feeder located in Arizona with a nominal voltage of 24.9 kV The voltage profiles of the PV inverters at the PCC depend
and a voltage at the substation of 1.05 pu. The IEEE 34-node on their locations within the DS. The feeder voltages reduce
test feeder includes two in-line voltage regulators, an in-line from the substation to the load terminals along the transmission
transformer for a short 4.16 kV section, two shunt capacitors, lines. The target set points of the feeder voltages are obtained
and unbalanced three-phase loads. The existing test feeder in by running a power flow solution at 60% of the rated loads
OpenDSS (IEEE34-Mod2 accommodates for the distributed without PV generation. The resulting voltage set-point for each
loads) is modified as follows to facilitate the case studies: 1) a PV inverter decreases with respect to the distance between the
total of 30 single-phase PV inverters are installed at every PV inverter and the substation. The voltage set points are within
distributed load and 2) the voltage regulators are set to the rated the ANSI limits of 1 ± 0.05 pu as shown in Fig. 6.
ratio of 1:1 since the PV inverters are used to control the voltage
instead of the voltage regulators.
848

822 846

820 844
864
818 842

802 806 808 812 814 850 824 826 834 860 836
858 840
816
832 862
800 888 890
810
838
852
Fig. 6 Vref for each PV inverter with respect to the distance from substation
828 830 854 856

Fig. 4 Single line diagram of IEEE 34 test feeder V. SIMULATION RESULTS AND DISCUSSION
Commercial and residential load profiles based on real The PV bus voltages, real and reactive power generations of 3
statistical data from Flagstaff, Arizona are used for each spot different PV inverters for the three reactive power control
load and distributed loads, respectively [14]. A highly modes at a penetration level of 100% are shown in Figs 7−9.
intermittent solar generation profile based on measured AC (blue: close to the substation at bus 810, red: in the middle of
power data with one second resolution is applied to all the PV the feeder at bus 820 and yellow: at the end of the feeder at bus
systems [15]. The solar generation and load profiles are 840). It is observed in Fig. 7 that the voltage variations at PV
normalized based on the maximum capacity of the feeder and buses depend on the loads and the solar real power generations.
is illustrated in Fig. 5. The OpenDSS simulations over a 24- The reactive power is set to zero and is not used to regulate the
hour period with a simulation time step of one second were PV bus voltages. Therefore, the voltages vary largely between
performed at different penetration levels and reactive power the voltage limits of 1 ± 0.05 pu. Fig. 8 shows that the PV
control modes of the PV inverters. The penetration level is inverters in Q-V droop control mode generate Q with respect to
defined as the ratio of the total solar generation capacity to the the voltage at the PV bus. The voltages at PV buses (Fig. 8 a)
total load capacity of the feeder. are in a narrower band than without Q generation. In Fig. 8 b)
In the first case study, the PV systems placed within the DS, it can be seen that the full Q capabilities of the inverters are not
were varied in their sizes proportional to the individual loads used as the maximum Q output is not reached.
from 0 to 100% penetration at three different Q control modes:

‹,(((
4

VPVbus (pu) 1.05 close to sub Fig. 9 a) illustrates that the proposed PI controller provides a
middle of feeder
smoother voltage profile, which is not as much affected by load
1 and solar variations. Fig. 9 b) shows, that for the same inverter
size, there is considerably more Q output of the PV inverters
(a)
0.95 end of feeder compared to the Q-V droop control. The inverters now operate
at their maximum rating during a longer period. At t=0 the total
20
Qinv (kvar)

load on the feeder is low, therefore the PV inverters absorb


0 significant amounts of Q to lower the PV bus voltages to the
reference level ܸ௥௘௙ . It can be observed in Fig. 9 that the
-20 (b) maximum solar generation around noon does affect the PV bus
40 voltages because the inverters saturate.
Pinv (kW)

For the two case studies of the linear droop and the PI control,
20 the maximum standard deviation (SD) of the feeder voltage and
the overall energy loss in the distribution system are compared
(c) over the total simulation time of 24 hours. The SD of the feeder
0
0 6 12 18 24 voltage is calculated as
Time (hours) ଵ ଶ
Fig 7. Real power only without Q generation at 100% penetration, a) voltages ܵ‫ ܦ‬ൌ ට σே
௜ୀଵหܸ௉஼஼௜ െ ܸ௔௩௚ ห (6)
ேିଵ
at PV buses; b) Q generations; and c) P generations of inverters
close to sub
where ܸ௔௩௚ is the average of the respective VPCC; N is the
1.05
VPVbus (pu)

number of samples in the 24-hour simulation period.


middle of feeder The calculation results show that the PV bus voltage at the
1 bus 844 has the highest SD. Therefore, it is the worst case
(a) end of feeder scenario within the entire feeder and is used for the comparison
0.95
of the three voltage control approaches. The max voltage SD is
20 illustrated in Fig. 10 for different penetration levels from 0 to
Qinv (kvar)

0 100%. It is shown in Fig. 10 that the proposed Q-V-based PI


controller decreases the SD significantly at penetration levels
-20 (b)
(b) larger than 50%. It is noted that the PV inverter at bus 844 is
40 relatively small compared to the other inverters and therefore
has limited Q capabilities leading to higher SD of this particular
Pinv (kW)

bus voltage. 50% over-dimensioned PV inverters are used for


20
both linear droop and PI control modes in the second case study.
(c) Over-dimensioning of PV inverters leads to better voltage
0 control and smaller SD as shown in Fig. 10. At higher
0 6 12 18 24 penetration levels, the increased Q capability of PV inverters
Time (hours) allows for voltage values closer to the target set points.
Fig. 8 Q-V droop control at 100% penetration, a) voltages at PV buses; b) Q
generations; and c) P generations of inverters 0.03
close to sub
1.05 no Q 100%
VPVbus (pu)

/ 150%
middle of feeder
max SD (pu)

1 0.02
Q/V dro
op 100%
PI
(a) con
0.95 end of feeder trol
100 Q/V dro
PI % op 150%
20 (b) co
Qinv (kvar)

ntr
ol
15
0 0.01 0%

-20
40 0 0.2 0.4 0.6 0.8 1
Penetration
Pinv (kW)

Fig. 10 Largest SD within the feeder for different Q control modes at varying
20 PV penetration.
(c)
0 The total energy loss in the feeder using different control
0 6 12 18 24 approaches is illustrated in Fig. 11. Generally, the total losses
Time (hours) over the 24-hour simulation time decrease at higher penetration
Fig. 9 The proposed Q-V based PI control at 100% penetration, a) voltages at
levels. The more the P and Q are supplied locally, the less
PV buses; b) Q generations; and c) P generations of inverters energy is delivered from the substation through the feeder
which results in lower losses. The proposed PI controller has
slightly higher losses than the Q-V droop control. Particularly

‹,(((
5

when oversizing the PV inverters, there is increased Q flow to driving the voltage at the PCC to target set points. The
buses with no PV systems. Hence, the voltage regulation performance of the proposed method was studied on the IEEE
capability of the proposed method is at the expense of 34 test feeder at different penetration levels and PV inverter
moderately increased losses in the example feeder. However, sizes. Time-domain simulations over a 24-hour period based on
by installing PV systems at every bus, Q consumption could be real statistical data in OpenDSS and Matlab show that the
compensated locally to reduce the Q flow in the feeder. The proposed voltage control can regulate the inverter voltages at
proposed method could ideally be combined with CVR, which PCC to the target set points. The impact of PV generation on
could further reduce energy consumption and losses. the voltage in the DS can be reduced significantly by the
2400 proposed approach although it leads to slightly higher losses
compared to a standard Q-V droop control method due to
Total energy loss (kWh)

increased Q flow in the feeder.


2200
REFERENCES
PI control 150% [1] S. Blumsack, "Utilities, solar energy and the fight for your roof," The
PI control 100% Conversation, 2016. [Online]. Available:
2000 no Q 100% https://theconversation.com/utilities-solar-energy-and-the-fight-for-your-
no Q 150% roof-54019. Accessed: Nov. 1, 2016.
Q/V droop 100% [2] R. A. Walling, R. Saint, R. C. Dugan, J. Burke and L. A. Kojovic, "Summary
Q/V droop 150% of Distributed Resources Impact on Power Delivery Systems," IEEE
Transactions on Power Delivery, vol. 23, pp. 1636-1644, 2008.
1800 [3] M. A. Mahmud, M. J. Hossain and H. R. Pota, "Voltage Variation on
0 0.2 0.4 0.6 0.8 1 Distribution Networks With Distributed Generation: Worst Case
Penetration Scenario," IEEE Systems Journal, vol. 8, pp. 1096-1103, 2014.
Fig. 11 Total energy loss in the feeder for different Q control modes at [4] EPRI, “Common Functions for Smart Inverters,” Version 3. EPRI, Palo
varying PV penetration Alto, CA: 2013. 3002002233.
[5] S. Gonzalez, F. Hoffmann, M. Mills-Price, M. Ralph and A. Ellis,
Fig. 12 illustrates the voltages at bus 844 for the PI-based "Implementation of advanced inverter interoperability and functionality,"
Photovoltaic Specialists Conference (PVSC), 2012 38th IEEE, 2012, pp.
reactive power control using normal-size inverters (100%) and 001362-001367.
50% oversized inverters. It can be observed in Fig. 12 that the [6] SunSpec Alliance, "About the SunSpec Alliance," SunSpec Alliance, 2016.
voltage variations for the PI control approach with oversized [Online]. Available: http://sunspec.org/sunspec-about/. Accessed: Nov. 1,
inverters are noticeably smaller than for normal-size inverters. 2016.
[7] P. M. S. Carvalho, P. F. Correia and L. A. F. M. Ferreira, "Distributed
The benefit of oversizing the PV inverter is that there is less Reactive Power Generation Control for Voltage Rise Mitigation in
impact of inverter reactive power saturation when the real Distribution Networks," IEEE Transactions on Power Systems, vol. 23, pp.
power generation is close to ܵ௠௔௫ around noon. This helps to 766-772, 2008.
smoothen the feeder voltage profile. A similar effect could also [8] V. Calderaro, G. Conio, V. Galdi, G. Massa and A. Piccolo, "Optimal
decentralized voltage control for distribution systems with inverter-based
be achieved by introducing curtailment of P generation without distributed generators," IEEE Transactions on Power Systems, vol. 29, pp.
oversizing of the inverters. Particularly at higher penetration 230-241, 2014.
levels, the voltages at the PCC become almost independent of [9] T. Sansawatt, L. F. Ochoa and G. P. Harrison, "Integrating distributed
the load and the solar generation. This effect could significantly generation using decentralised voltage regulation," in IEEE PES General
Meeting, 2010, pp. 1-6.
decrease the number of mechanical switching operations of [10] A. Maknouninejad, N. Kutkut, I. Batarseh and Zhihua Qu, "Analysis and
voltage regulators and reduce their wear-out. control of PV inverters operating in VAR mode at night," Innovative Smart
Grid Technologies (ISGT), 2011 IEEE PES, 2011, pp. 1-5.
[11] A. Visioli, Practical PID Control. Springer, 2006, pp. 35-60.
1.05 [12] W. H. Kersting, "Radial distribution test feeders," Power Engineering
PI control Society Winter Meeting, 2001. IEEE, 2001, pp. 908-912 vol.2.
PI control inv size 150% [13] M. J. Reno and K. Coogan, "Grid Integrated Distributed PV (GridPV)
VPVBus844 (pu)

inv size 100% Version 2," Sandia National Laboratories SAND2013-20141, 2014.
[14] Office of Energy Efficiency & Renewable Energy (EERE), "Commercial
and residential hourly load profiles for all TMY3 locations in the United
1 States - OpenEI Datasets," [Online]. Available:
http://en.openei.org/datasets/dataset/commercial-and-residential-hourly-
load-profiles-for-all-tmy3-locations-in-the-united-states. Accessed: Nov.
3, 2016.
no PV [15] EPRI, "Distributed PV monitoring and feeder analysis," [Online],
0.95 Available: http://dpv.epri.com/measurement_data.html. Accessed: Nov. 3,
2016.
0 6 12 18 24
Time (hours)
Fig. 12 Voltage at bus 844 for a 24h simulation period at Q-V based PI control
mode with normal sized and 50% over sized inverter

VI. CONCLUSION
A Q-V based PI controller for Q control of PV inverters was
introduced. The proposed voltage control method is capable of

‹,(((

You might also like