Adaptable Low-Cost Volt-VAR Control for Smart PV

Inverters
Shayan Ebrahimi, S.M. Safayet Ullah, Farzad Ferdowsi
Electrical and Computer Engineering Department, University of Louisiana at Lafayette, Lafayette, LA, United States
[email protected], [email protected], [email protected]
2022 North American Power Symposium (NAPS) | 978-1-6654-9921-7/22/$31.00 ©2022 IEEE | DOI: 10.1109/NAPS56150.2022.10012193

Abstract—Solar photovoltaic (PV) is becoming popular around Traditional devices such as on-load tap changers (OLTC),
the world for its decreasing cost and increasing efficiency. However,
voltage regulators, and capacitor banks are not quick enough
the high penetration of solar PV can bring several challenges to to provide the support to mitigate voltage fluctuations issue [4].
the power grid. Due to the intermittent characteristics of solar PV
generation, voltage violation, as well as fluctuation can become The grid support functions of Smart Inverters (SI) have been
major issues affecting the power quality of the grid. Volt-VAR game changers within the past few years providing solutions in
control in the smart inverter is used to provide reactive power the form of control schemes that are less costly compared with
support to mitigate voltage violation and fluctuation issues. In equipment-based solutions such as storage devices. In different
this paper, a corrective mechanism updates the inverter’s reactive research works, SIs are studied to control power parameters
power set-point to concurrently mitigate voltage violation and
fluctuation. The proposed “Delta-Q” approach is incorporated at the solar system’s point of interface (POI). The control
into the Volt-VAR control of the smart inverter. A modified parameters include but are not limited to voltage, active/reactive
IEEE 4 Bus system is utilized to validate the proposed control power, and frequency.
method. For real-time simulation, Typhoon Hardware In Loop Among SI’s control schemes, the Volt-VAR control for
(HIL) 402 series is utilized with real solar radiation and load voltage regulation at POI is quite common in the literature. In
data set. Furthermore, the control parts of the proposed Delta-Q
approach are implemented using dSPACE MicroLabBox device. [9], multi-objective optimization for different smart inverter
The results show the effectiveness of the proposed method in functions including Volt-VAR is proposed to select the optimal
comparison with the traditional Volt-VAR control in mitigating smart inverter curve. In [10], the authors presented an optimal
voltage violation and fluctuation. power flow-based approach to select the local Volt-VAR and
Index Terms—Solar Photovoltaic, Smart Inverter, Volt-Var Volt-Watt settings of SIs for voltage profile enhancement. In
Control, Voltage Fluctuation.
another recent study [11], the authors proposed Volt-VAR
I. I NTRODUCTION Optimization (VVO) integrated with Distribution Grid Optimal
Power Flow (DOPF) to overcome voltage fluctuation issues.
Recent aggressive environmental goals in the Unites States Although the method could lessen the voltage variations, it is
require a significant growth of about 40% in solar penetration. computationally extensive.
According to Solar Futures Study [1], more than 100,000 With many recent advancements in Artificial Intelligence,
MW of new solar capacity is expected to become operational AI-driven control functions for SIs have remained a top trend
by 2025. The increased demand for electrification would in SI control technologies within the past two years. To
require several thousands of gigawatts of solar capacity for be more specific, multi-agent Deep Reinforcement Learning
achieving decarbonization goals. Recent studies show that (DRL) is proposed in different recent research articles [12]–
the power networks might not be fully ready yet to host a [16] as an alternative to the model-based classic control. The
significant growth of solar penetration. In [2], power quality proposed control methods ensure the most-rewarded actions
issues including voltage rise and flicker are reported in actually (reactive power dispatch) by minimizing voltage violations and
existing power distribution networks in northwestern Louisiana active power losses. The effectiveness of DRL-driven Volt/VAR
for solar penetrations between 10% and 20%. control has been verified on different theoretical feeders
Common power grid issues associated with high penetration including IEEE 4, 33, 34, 69, and 123-bus systems. However,
of solar may include voltage violations, voltage fluctuations, the computational costs associated with DRL-based methods are
power losses, and instability reported in different literature [3]– not well addressed in the literature. That includes an additional
[6]. Traditionally, battery storage systems are utilized along capital cost for stronger GPUs as well as operational costs
with solar facilities to mitigate the above-mentioned power due to more power consumption. Although the application
quality issues [7]. This is mainly because batteries can respond of DRL-aided control in energy systems is on the rise [17],
really fast to solar-related dynamics. However, the battery’s DRL methods are not benchmarked to be compared with set-
support does not come without a cost. Batteries can be more point/rule-based strategies that are much simpler to implement.
costly if they are dispatched with a high Depth of Discharged That makes Reinforcement Learning act as Maslow’s hammer
(DoD) affecting the battery’s life adversely [8]. for setting up SI’s control functions including Volt/VAR control.
In this paper, the Delta-Q approach is proposed for Volt-

978-1-6654-9921-7/22/$31.00 ©2022 IEEE

Authorized licensed use limited to: North South University. Downloaded on February 08,2024 at 03:56:20 UTC from IEEE Xplore. Restrictions apply.
VAR control of SIs to address the above-mentioned issues. The and high frequency can affect the power quality in terms of
main advantage of the Delta-Q approach is its simplicity which visible flicker [2].
makes it easy and affordable to implement in real systems.
III. P ROPOSED M ETHODOLOGY
Unlike machine learning-based control algorithms, the Delta-Q
method does not need a lot of communication or training. Thus, The aim of this study is to concurrently mitigate voltage
Delta-Q performs properly for both large-scale PV applications fluctuations and violation (voltage rise). The general approach
and small-scale ones, like residential roof-top PVs, where we of the proposed Delta-Q control strategy is to keep updating the
usually have fluctuation issues. Another advantage of the Delta- value of reference reactive power,Qref , calculated by traditional
Q approach is its adaptability to any sampling time due to its SI’s Volt-VAR control for each sample time Ts . In this study,
simplicity and fast performance. the sample time Ts is considered as 200 µs. The overview of
The limitations associated with traditional Volt-VAR control the control parts of the proposed Delta-Q is illustrated in figure
methods are described in Problem Statement, section II. The 2.
mathematical overview and analytical algorithms are outlined in
section III, Proposed Methodology. In section IV, the technical
analysis is discussed. Finally, the result and discussion of this
study are provided in section V followed by a brief conclusion.

II. P ROBLEM S TATEMENT

The concept of Volt-VAR control methods is to inject/absorb
a certain amount of reactive power, based on the reference
voltage to keep the voltage of POI within a permissible range,
e.g. 0.95pu to 1.05pu (ANSI standard) in each sample time Figure 2: Delta-Q Control Strategy
Ts [18]. The Q-V curve of the Volt-VAR control method is As shown in figure 2, the value of Qref is constantly updated
illustrated in figure 1. If the voltage remains in the dead band by the ∆Q term, thereby the Qref,updated is applied instead
region(zone 3 in figure 1), there will be no reactive power to mitigate the voltage fluctuations. The value of ∆Q is highly
support from the smart inverter. dependent on two main factors. The first aspect is to what
extent solar irradiance is intermittently led to abrupt variations
in the output voltage of PV. The other factor hangs on the
value of the threshold set for the proposed Delta-Q approach to
start taking action. If the difference between the instantaneous
voltage and the reference voltage is greater than the threshold,
the Qref is going to be updated by ∆Q.
A. Mathematical Overview of Delta-Q Approach
The mathematical equations lain behind the Delta-Q ap-
proach in the pu system are presented in this section. According
to figure 2, phase voltages and currents at node 4 of figure 3
are constantly measured and converted to the two components
system via the Clarke transformation formula.
 
 Va
2 cos 0 cos 2π 4π
  
Figure 1: Volt-VAR control in smart inverters Vα 3 cos 3  Vb 
= (1)
Vβ 3 sin 0 sin 2π3 sin 4π
3 Vc
Traditional Volt-VAR control methods keep  
 Ia
injecting/absorbing a specific amount of reactive power 2 cos 0 cos 2π 4π
  
Iα 3 cos 3  Ib 
for the whole one sample time Ts [19]. For instance, when = (2)
Iβ 3 sin 0 sin 2π3 sin 4π
3
the reference voltage falls in zone 4, the Volt-VAR control Ic
unit of SI sends a command of absorbing a constant value By having Vα and Vβ from 1, the instantaneous voltage
of Qref for the whole duration of one sample time Ts . The reference, Vref,cal , during one sample time Ts can be calculated.
problem, however, is prompted when the solar irradiance is This voltage is continuously compared with the reference
abruptly changing. This can have a negative impact on the voltage calculated by SI’s traditional Volt-VAR control unit of
output voltage profile of PV’s SI and can lead to voltage PV, Vref,pv , to produce the error of voltage as below:
fluctuations since in statistic Volt-VAR control methods the
value of Qref remains constant for the whole sample time Ts . ∆V = Vref,pv − Vref,cal (3)
The mentioned problem can reveal its negative impact more Where ∆V is the calculated error between two reference
on small-scale roof-top PVs. Fluctuations with large magnitude voltages and it is constantly compared with a specific threshold

2
Authorized licensed use limited to: North South University. Downloaded on February 08,2024 at 03:56:20 UTC from IEEE Xplore. Restrictions apply.
determined based on the amount of sensitivity required for from Typhoon HIL 402 through Typhoon-dSAPCE interface
applications. Once ∆V is greater than the threshold, the board connection to dSPACE to take required actions and send
proposed control strategy starts taking action to minimize the the final result of Qref,updated back to the Typhoon. Since the
difference between Vref,cal and Vref,pv by updating the amount whole process of updating Qref takes less than 200 µs, the
of reference reactive power as follows: proposed Delta-Q is fast enough to update Qref a couple of
times during one sample time Ts .
Qref,updated = Qref,old + ∆Q (4)
IV. T ECHNICAL A NALYSIS
Where Qref,updated and Qref,old are the updated and
traditional Volt-VAR reference reactive power, respectively. To evaluate the performance of the proposed dynamic Volt-
Moreover, ∆Q is the additional amount of reactive power re- Var control method, a modified IEEE 4 bus system is applied,
quired for mitigating voltage violations which can be calculated figure 3.
through 5.

∆Q = |∆V ||∆Iβ | (5)

Where ∆Iβ is the difference between the reference Iβ
of the traditional SI’s Volt-VAR control unit of PV and Iβ
calculated in equation 2. It is worth mentioning that adding or
subtracting ∆Q to Qref,old is highly dependent on both the
sign of ∆V and SI’s operation zone which is further discussed Figure 3: Modified IEEE 4-bus system with solar PV plant
in the next section. Thus, the absolute amounts of ∆V and
∆Iβ are applied to calculate the value of ∆Q in 5. To analyze the major impact of higher PV penetrations in
this IEEE 4 bus system, a single solar plant is connected to bus
B. Analytical Algorithm of Delta-Q Approach 4 which is the most remote place in this radial IEEE system.
The sign of ∆V and SI’s operation zone have a vital role The nominal active power of the solar PV plant is 180 kW.
to play in updating Qref,updated in 4. Here in table I, the The peak load apparent power is 181.96 kVA. So, a high PV
proposed Delta-Q approach is presented in an algorithmic penetration level is utilized in this research which is around
way aiming to mitigate voltage violations/fluctuations. 98%. Higher solar irradiance is required to produce maximum
solar power from solar PV plant [21], [22]. For this purpose,
1-minute interval of higher solar radiation data is utilized from
Table I: Algorithm of Delta-Q Approach [23]. Real load data set is utilized on the load bus of this
test system. The smart inverter is operating with the reactive
Step 1 : Check the Value of ∆V : power priority mode. Therefore, it will always prioritize reactive
If the difference of Vref,pv and Vref,cal is bigger than power support to the power grid by curtailing the active power
threshold generation of the solar PV plant.
Step 2 : Check SI Operation Zone:
Detecting the zone according figure 1
Step 3 : Check the Sign of ∆V :
If it is positive → ∆Q in equation 5 with positive sign
If it is negative → ∆Q in equation 5 with negative sign
Otherwise → ∆Q in equation 5 should be equal to zero
Step 4 : Updating Reference Q:
Apply the result of step 3 to update the equation 4

It is worth mentioning that according to IEEE 1547-2018

standard [20], the reactive power injecting/absorbing capacity
is set to 0.44 pu. It means that the maximum reactive power
support injection in zone 1 of the figure 1 is limited to 0.44
pu. Thus, ∆Q in 5 should be equal to zero due to maximum
Figure 4: Real-time analysis set-up of the case study
injection by the traditional Volt/VAR control unit. The same
limitation is applied to absorption in zone 5, thereby ∆Q is For the real-time test, Typhoon HIL 402 and dSPACE
equal to zero in this case. MicroLabBox are used. The setup is shown in figure 4.
In addition, the above-mentioned Delta-Q control algorithm After measuring the required inputs for control units including
units including traditional Volt/VAR control units which can be Vref,cal , Iβ , and Vref,pv , they are transferred through the inter-
seen in figure 2 are mounted in dSAPCE MicroLabBox to run face board from Typhoon HIL 402 to dSPACE MicroLabBox.
as a real-time application. The measured signals are transferred Then, the control unit starts processing the data to calculate

3
Authorized licensed use limited to: North South University. Downloaded on February 08,2024 at 03:56:20 UTC from IEEE Xplore. Restrictions apply.
the amount of ∆Q required for mitigating voltage fluctuations Delta-Q approach takes action to mitigate voltage fluctuations
during each sample time Ts . Eventually, the updated value of via updating, the current waveform also becomes smooth.
Qref,updated is sent back to Typhoon HIL 402 due to applying In figure 7, the amounts of delivering active power generation
on the SI. by PV for both the traditional Volt-VAR control method and the
Delta-Q approach are illustrated. Although the traditional Volt-
V. R ESULTS AND D ISCUSSION VAR control method provides more active power in comparison
The operation and behavior of the proposed Delta-Q control with the Delta-Q approach, it is not able to mitigate voltage
algorithm is simulated and discussed in this section. The main fluctuations issues. This means that to lessen the number of
purpose of this study is to mitigate the voltage fluctuations fluctuations in the voltage curve during abrupt solar irradiance
led by abrupt solar irradiance variations. To better evaluate the variation, more active power should be curtailed to temporarily
performance of the proposed Delta-Q approach, the traditional absorb more reactive power and have an acceptable voltage
Volt-VAR control method with reactive power priority is first profile.
utilized to compare with the operation of the Delta-Q approach.
As it can be seen in figure 5, the traditional Volt-VAR control
method keeps the voltage below 1.05 pu. Its voltage curve,
however, contains abrupt voltage fluctuations which can bring
power quality issues in the distribution network for large-scale
PV applications. In this voltage variation situation, the Delta-
Q approach takes effective actions to reduce the fluctuations
to further smooth the voltage profile. Not only the Delta-Q
method brings the peak voltage down, but also it can reduce
the abrupt voltage variations to a great extent which can be
seen in figure 5.

Figure 7: Active Powers of Traditional Volt/VAR Control method

and Delta-Q Control Approach
The amounts of absorbing reactive power by SI for both the
traditional Volt-VAR control method and the Delta-Q approach
are shown in figure 8. Due to a large number of voltage
fluctuations, Delta-Q updates Qref in a way that absorbs more
to the minimum reactive power limit of 44% in comparison to
traditional Volt-VAR control. Although this means curtailing
more active power, it makes a substantial contribution to
smoothing the voltage profile and addressing power quality
issues to a great extent.
Figure 5: Voltages of Traditional Volt/VAR Control method and
Delta-Q Control Approach

Figure 8: Reactive Powers of Traditional Volt/VAR Control method

and Delta-Q Control Approach
Figure 6: Currents of Traditional Volt/VAR Control method and According to table I, the concept of Delta-Q is to update
Delta-Q Control Approach
the value of reactive power reference to mitigate the abrupt
Figure 6 depicts the current of both the traditional Volt- voltage fluctuations caused by the intermittent nature of solar
VAR control method and the Delta-Q approach in pu. As it irradiance. Therefore, based on the number of voltage variations
can be seen, once the voltage starts fluctuating, the current is and the value of the threshold set associated with the level
affected accordingly and has variations. Therefore, once the of sensitivity, the Delta-Q approach starts playing an effective

4
Authorized licensed use limited to: North South University. Downloaded on February 08,2024 at 03:56:20 UTC from IEEE Xplore. Restrictions apply.
role and changing the value of Qref accordingly. Even though [8] A. J. Abianeh and F. Ferdowsi, “Sliding mode control enabled hybrid
Delta-Q in a situation with lots of voltage variations curtails energy storage system for islanded dc microgrids with pulsing loads,”
Sustainable Cities and Society, vol. 73, p. 103117, 2021.
more active power in comparison with traditional Volt/VAR [9] T. O. Olowu, J. Sarochar, and A. I. Sarwat, “Pareto optimal smart inverter
control methods, it can considerably reduce the negative effects curve selection for high photovoltaic penetration,” in 2021 IEEE Green
of voltage fluctuations. It is worth mentioning that the Delta-Q Technologies Conference (GreenTech). IEEE, 2021, pp. 1–6.
[10] T. O. Olowu, A. Inaolaji, A. Sarwat, and S. Paudyal, “Optimal volt-var
approach is designed to optimally improve Q control to mitigate and volt-watt droop settings of smart inverters,” in 2021 IEEE Green
fluctuations. Furthermore, the total harmonic distortion (THD) Technologies Conference (GreenTech). IEEE, 2021, pp. 89–96.
of the Delta-Q approach is 3.5% which is in the acceptable [11] A. Savasci, A. Inaolaji, and S. Paudyal, “Distribution grid optimal power
flow with adaptive volt-var droop of smart inverters,” in 2021 IEEE
range according to IEEE standards. Industry Applications Society Annual Meeting (IAS), 2021, pp. 1–8.
[12] Y. Gao, W. Wang, and N. Yu, “Consensus multi-agent reinforcement
VI. C ONCLUSION learning for volt-var control in power distribution networks,” IEEE
Transactions on Smart Grid, vol. 12, no. 4, pp. 3594–3604, 2021.
In this study, the impact of variable solar data with lots of [13] D. Cao, J. Zhao, W. Hu, F. Ding, Q. Huang, and Z. Chen, “Attention
ups and downs on the solar PV’s voltage profile at the POI enabled multi-agent drl for decentralized volt-var control of active
is investigated. In large-scale PV applications, variable solar distribution system using pv inverters and svcs,” IEEE transactions
on sustainable energy, vol. 12, no. 3, pp. 1582–1592, 2021.
data can cause abrupt voltage fluctuations led to power quality [14] H. Liu, C. Zhang, Q. Chai, K. Meng, Q. Guo, and Z. Y. Dong, “Robust
issues, and decrease the longevity of loads in the distribution regional coordination of inverter-based volt/var control via multi-agent
network. The issue is more severe in small-scale PVs used for deep reinforcement learning,” IEEE Transactions on Smart Grid, vol. 12,
no. 6, pp. 5420–5433, 2021.
residential applications which cause serious damage to end-user [15] H. Liu and W. Wu, “Online multi-agent reinforcement learning for
equipment. To tackle mentioned issues, the Delta-Q approach is decentralized inverter-based volt-var control,” IEEE Transactions on
introduced in this paper. Not only does it substantially mitigate Smart Grid, vol. 12, no. 4, pp. 2980–2990, 2021.
[16] H. Liu and W. Wu, “Two-stage deep reinforcement learning for inverter-
voltage fluctuations in the voltage of SI to a great extent, but based volt-var control in active distribution networks,” IEEE Transactions
also it can reduce the pick of voltage in comparison to the on Smart Grid, vol. 12, no. 3, pp. 2037–2047, 2020.
traditional Volt-VAR control method. Another significance of [17] A. Perera and P. Kamalaruban, “Applications of reinforcement learning
in energy systems,” Renewable and Sustainable Energy Reviews, vol.
Delta-Q is its simple concept which considerably reduces cost in 137, p. 110618, 2021.
comparison to DRLs. Moreover, it is capable of performing and [18] S. S. Ullah, A. J. Abianeh, E. B. Osunwoke, and F. Ferdowsi, “Compara-
reacting fast with no need for training like DRLs. Furthermore, tive analysis of volt-var control parameter settings of smart pv inverters: A
case study,” in 2021 North American Power Symposium (NAPS). IEEE,
due to its time sample adaptable capability, there is no need 2021, pp. 01–06.
to make extra sample time which brings redundant complexity [19] S. Ebrahimi, S. S. Ullah, F. Ferdowsi, and M. Barati, “Machine learning-
to the control process like other dynamic Volt-VAR control aided anomaly detection for secure smartification of solar inverters,” in
2022 IEEE Kansas Power and Energy Conference (KPEC), 2022, pp.
methods. 1–6.
[20] “IEEE standard for interconnection and interoperability of distributed
ACKNOWLEDGMENT energy resources with associated electric power systems interfaces,” IEEE
Std 1547-2018 (Revision of IEEE Std 1547-2003), pp. 1–138, 2018.
This work was partially supported by the Louisiana Board [21] M. F. Nayan and S. S. Ullah, “Modelling of solar cell characteristics
of Regents under grant number: LEQSF(2021-24)-RD-B-06. considering the effect of electrical and environmental parameters,” in
2015 3rd International Conference on Green Energy and Technology
(ICGET). IEEE, 2015, pp. 1–6.
R EFERENCES [22] M. F. Nayan, S. S. Ullah, and S. Saif, “Comparative analysis of pv module
[1] R. Margolis, “Solar futures study databook,” National Renewable Energy efficiency for different types of silicon materials considering the effects
Laboratory-Data (NREL-DATA), Golden, CO, United States, Tech. Rep., of environmental parameters,” in 2016 3rd International Conference
2021. on Electrical Engineering and Information Communication Technology
[2] G. B. Upton Jr, F. Ferdowsi, A. Kargarian, and S. Mehraeen, “The future (ICEEICT). IEEE, 2016, pp. 1–6.
of solar in louisiana,” 2019. [23] “Online,” https://www.nrcan.gc.ca/energy/renewable-electricity/
[3] M. Guleryuz Halacli and A. Demiroren, “Techno-economic management solar-photovoltaic/18409, accessed: July 11th, 2022.
of daily volt/var planning considering reactive power contribution of high
penetrated-pv solar farms utilized as statcom: A case study in istanbul,
turkey,” International Transactions on Electrical Energy Systems, vol. 31,
no. 3, p. e12772, 2021.
[4] V. B. Pamshetti and S. P. Singh, “Optimal coordination of pv smart
inverter and traditional volt-var control devices for energy cost savings
and voltage regulation,” International Transactions on Electrical Energy
Systems, vol. 29, no. 7, p. e12042, 2019.
[5] A. Moghassemi, S. Ebrahimi, and F. Ferdowsi, “A novel control scheme
for transzsi-dvr to enhance power quality in solar integrated networks,”
in 2021 North American Power Symposium (NAPS), 2021, pp. 1–6.
[6] J. Olamaei, S. Ebrahimi, and A. Moghassemi, “Compensation of
voltage sag caused by partial shading in grid-connected pv system
through the three-level svm inverter,” Sustainable Energy Technologies
and Assessments, vol. 18, pp. 107–118, 2016. [Online]. Available:
https://www.sciencedirect.com/science/article/pii/S2213138816300881
[7] I. Ranaweera, S. Sanchez, and O.-M. Midtgård, “Residential photovoltaic
and battery energy system with grid support functionalities,” in 2015
IEEE 6th International Symposium on Power Electronics for Distributed
Generation Systems (PEDG). IEEE, 2015, pp. 1–7.

5
Authorized licensed use limited to: North South University. Downloaded on February 08,2024 at 03:56:20 UTC from IEEE Xplore. Restrictions apply.