2258 IEEE TRANSACTIONS ON SMART GRID, VOL. 8, NO.

5, SEPTEMBER 2017

A Dynamic Operational Scheme for Residential

PV Smart Inverters
Ahmad Reza Malekpour, Student Member, IEEE, and Anil Pahwa, Fellow, IEEE

Abstract—This paper presents an operational scheme for pho- strategies [5], [6]. In the cosϕ(P) strategy, inverter feed-in
tovoltaic (PV) inverter reactive power control to accommodate power is monitored; once the power exceeds a predetermined
higher levels and leverage efficient use of rooftop PV penetration limit, the PV inverter starts to absorb reactive power at a con-
in distribution systems. The scheme proposes three states of oper-
ation with specific operational goals for the PV inverter based on stant rate. On-load tap-changer (OLTC) control and cosϕ(P)
weather conditions and voltage at the interconnection point, and strategies for PV inverters were used in [7] to limit voltage
adapts the reactive power control strategy accordingly. In normal rise. Location-dependent cosϕ(P) characteristics were pro-
state with slowly changing solar irradiance, the control modu- posed in [5] and [8] to mitigate voltage rise and reduce power
lates reactive power to reduce power losses. In fluctuating state losses. In the Q(V) strategy, a droop characteristic based on
with rapidly varying solar irradiance due to intermittent pass-
ing clouds, the control dynamically changes the reactive power inverter bus voltage is used to calculate PV inverter reactive
in order to mitigate voltage fluctuations. In contingency state in power. Application of the Q(V) control strategy was described
which the PV terminal voltage violates the nominal operating in [5] and [9] for overvoltage prevention control of PV gen-
range, the control adjusts the PV inverter as reactive power sink erators in microgrids. The Q(V)/P(V) strategy was proposed
or source in order to push the voltage back within the range. This in order to mitigate voltage rise when reactive power control
paper also proposes a coordination strategy in order to switch
control between the states and manage interaction between the alone is inefficient. The method combines active power curtail-
fast PV inverter controllers, and the slow on-load tap-changer for ment with the Q(V) method. Referring to the literature, voltage
voltage regulation. A large-scale distribution network based on rise caused by reversed power during low demand and high
the IEEE 37-node test feeder was developed in order to investigate PV generation is known as primary network constraints that
performance of the proposed algorithm. limit PV penetration in distribution systems [10]. An online
Index Terms—Power distribution systems, power qual- overvoltage prevention control strategy based on active power
ity, reactive power control, renewable energy, smart grid, limit prediction was proposed in [11]. However, the concept
solar power. is suitable only for microgrids since identification of the sys-
tem Thevenin equivalent is the foundation of the algorithm.
I. I NTRODUCTION
Reference [12] discussed use of droop-based active power cur-
ARGE penetration of rooftop photovoltaic (PV) sys-
L tems present challenging power quality issues, such
as incremental power losses, voltage violation, and volt-
tailment for overvoltage prevention in radial LV feeders, while
the Q(V) and Q(V)/P(V) strategies and autonomous OLTC
operation were used in [6] to assess cost-benefit analysis of
age fluctuation for distribution system operators. Worldwide local control strategies.
interest in rooftop PV installation has created the need for Detrimental effects associated with high levels of PV
additional network regulations in order to achieve safe and integration are not limited to voltage rise. Power gener-
reliable operation of low voltage (LV) grid. Although pre- ation of rooftop PV can have large fluctuations due to
vious version of IEEE 1547 Std. [1] prohibited reactive rapid variations in solar irradiance on intermittent cloudy
power support by distributed generation (DG) in LV grids, days. For example, cloud cover can rapidly reduce solar
a recently published amendment [2] and standards issued in power generation, increasing flow of power from the grid
Germany [3] and Italy [4] specify DG reactive power control and potentially causing voltage drop problems. Recently,
strategies in order to maintain power quality levels or provide PV system studies during intermittent cloud movement have
ancillary services for the LV network. gained increasing attention. In [13], DSTATCOM was used
Several PV inverter reactive power control methods have to damp impacts associated with residential PV power fluc-
been proposed, including cosϕ(P), Q(V), and Q(V)/P(V) tuation on the OLTC operation. Reference [14] presented
inverter reactive power control for rooftop PV integration
Manuscript received May 4, 2015; revised August 21, 2015 and
November 19, 2015; accepted January 4, 2016. Date of publication that required detailed historical load demand and PV out-
February 24, 2016; date of current version August 21, 2017. This work put profile for each household. Moreover, inverter night-mode
was supported by the U.S. National Science Foundation under Award operation for the DSTATCOM was assumed; while the idea is
CNS-1136040. Paper no. TSG-00509-2015.
The authors are with the Department of Electrical Engineering, Kansas under research and development for solar farms [15] its suit-
State University, Manhattan, KS 66506 USA (e-mail: [email protected]; ability for house-level PV has not been demonstrated yet.
[email protected]). Studies in [16] and [17] proposed storage batteries to sup-
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. press the effect of large penetration of PV on power tran-
Digital Object Identifier 10.1109/TSG.2016.2521367 sients. However, cost-effective battery technologies able to
1949-3053 c 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
MALEKPOUR AND PAHWA: DYNAMIC OPERATIONAL SCHEME FOR RESIDENTIAL PV SMART INVERTERS 2259

compensate for unexpected PV power fluctuations are still

under development [18].
In spite of detrimental impacts, increased rooftop PV pen-
etration could improve distribution system efficiency through
control strategies that favor better system performance [19].
For example, high PV penetration levels could help reduce line
losses [20] or improve load-margin and voltage profile [21]. In
general, new control approaches and additional inverter func- Fig. 1. Schematic operational curve of PV inverter: (a) Unity power factor
tionalities applicable in all weather conditions are required method and (b) cosϕ(P) method.
to mitigate undesirable effects while enhancing beneficial
impacts of PV installations on distribution systems.
This paper proposes a novel dynamic PV inverter reac-
tive power control scheme to address adverse impacts while
exploiting the benefits related to high penetration of rooftop
PV in various weather conditions. Three states of operation
with specific operational goals for the PV inverter are proposed
based on local measurements of voltage and PV generation
variations. The control scheme is implemented on PV invert-
ers based on the state of operation, including normal state Fig. 2. Schematic operational curve of PV inverter: (a) Q(V) method
in sunny/fully cloudy periods, fluctuating state in intermittent and (b) Q(V)/P(V) method.
cloudy periods, and contingency state in which PV terminal
B. Cosϕ(P) Control
voltage violates the nominal operating range. Objectives for
the proposed states include reducing power losses, smooth- The cosϕ(P) characteristic curve for the connection of gen-
ing out voltage fluctuations, and mitigating voltage violation eration units at low voltage is shown in Fig. 1(b). In order to
(overvoltage or undervoltage), respectively. A reactive power mitigate voltage rise, a PV unit must operate in reactive power
ramp rate limiter is introduced in order to prevent additional consumption mode when the feed-in active power exceeds
voltage fluctuations due to reactive power variation caused a specified threshold (Pth ). However, the approach does not
by transitioning between the states. Moreover, a coordina- actively regulate voltage at the PCC because unnecessary reac-
tion scheme is designed to match the low-power fast-response tive power absorption increases line losses and degrades power
PV inverters with higher amplitude short-period fluctuations quality due to voltage fluctuations caused by fast ramping of
originating from intermittent passing clouds while considering PV unit on intermittent cloudy days.
the slow response of the OLTC operation in order to match
C. Q(V) Control
the longer period voltage deviation in the system. Simulation
results demonstrate superior performance of the proposed con- The Q(V) control strategy determines PV inverter reactive
trol scheme for meeting the objectives as compared to other power based on voltage at the PCC, as shown in Fig. 2(a).
reactive power control strategies such as unity power factor, PV inverter absorbs/injects reactive power if its terminal volt-
cosϕ(P), and Q(V) approaches. age is higher or lower than the predefined upper/lower critical
U and V L , respectively. Voltage magnitudes at
voltages, Vcr
The rest of the paper is organized as follows. Section II cr
briefly reviews inverter control strategies, Section III dis- PV buses near the substation transformer typically are within
cusses inverter technical requirements, Section IV explains the range; therefore, these PV inverters do not contribute in
voltage sensitivity analysis, Section V presents the proposed voltage regulation. Only PV inverters near the end of the
control approach, Section VI presents simulation results, and feeder provide reactive power and voltage support, causing
Section VII provides the conclusions and future work. additional stress on these inverters. Because the controller is
designed for voltage rise situations, it does not react to voltage
fluctuations [5], [9].
II. I NVERTER R EACTIVE P OWER C ONTROL S TRATEGIES
Various inverter reactive power control strategies for LV grid D. Q(V)/P(V) Control
operation are discussed in this section. The Q(V)/P(V) characteristic curve is shown in Fig. 2(b).
The control strategy is similar to the Q(V) method with the
A. Unity Power Factor With Limit on Active Power addition of an active power curtailment (Pcurtail ) feature for
In the unity power factor strategy, as shown in Fig. 1(a), the the PV inverter in case of voltage rise at the PCC. As a result,
PV inverter operates with no injection of reactive power into the approach could not accommodate reactive power support
the grid in order to comply with the previous version of IEEE for power loss reduction or control voltage fluctuations.
1547 Std. (i.e., distributed resources do not regulate voltage at
the point of common coupling (PCC)). In some applications, III. I NVERTER R ATING AND T ECHNICAL R EQUIREMENTS
a fixed limit (70%) on PV maximum active power generation In order to accommodate large penetration of PV in
(Pmax ) is enforced to prevent voltage rise from the installed distribution system, some utilities are attempting to man-
PV systems without a utility remote control unit. date smart inverters for all new solar facilities within their
2260 IEEE TRANSACTIONS ON SMART GRID, VOL. 8, NO. 5, SEPTEMBER 2017

Fig. 3. Operational areas for PV inverter at 0.8 PF, (a) standard, and
(b) oversized.
service territories [7]. Smart inverters are equipped with a
programmable logic controller (PLC) [22], [23] with extremely
fast response times (milliseconds) [24] that enables many func- Fig. 4. Voltage evolution in normal, fluctuating, and contingency states
tionalities including reactive power provision as a function without reactive power control.
of external setpoints [25]. Since the proposed control scheme
has a time resolution of 1-second, the response time of the
inverter has not been explicitly included in formulation of the where Pi , Qi , Vi , δi denote active power, reactive power, volt-
solution. If properly regulated, the PV inverter could act as age magnitude, and phase angle at bus i, respectively. Yij ∠θij
an active control component to eliminate or mitigate feeder is admittance of the line from bus i to bus j.
voltage variations while providing loss reduction and voltage Sensitivity matrix (S) is derived from partial derivation of
quality support by dynamically provisioning reactive power at P and Q with respect to V and θ as
the area of greatest need, the customer site. Fig 3. illustrates V [SVP ] [SVQ ] P
smart PV inverter operation in a 2-quadrant P-Q plane. The = (5)
θ [SθP ] [SθQ ] Q
feasible operating space is enclosed by dashed straight lines
that represent the total harmonic distortion (THD) limit [26] where SVP and SVQ are sensitivities of bus voltage magnitudes
and inverter rating curve. to active and reactive powers, respectively, and SθP and SθQ
As shown in Fig. 3(a), rooftop PVs with standard-sized are sensitivities of bus angles. In particular, voltage variation
inverters must reduce a portion of real power generation in in bus i due to 1 p.u. reactive power change at bus j can be
ij
order to accommodate a reasonable amount of reactive power interpreted as SVQ .
contribution near full active power. However, power reduction Voltage sensitivity analysis has been used for applications
is not desirable because customers assign priority to active such as voltage rise prevention via PV inverter reactive power
power generation. Moreover, if voltage rise or voltage drop consumption or active power curtailment [8], [12], [27], and
occurs in the system, tangible capacity from PV inverters may voltage variation mitigation at a target bus due to the oper-
be unavailable to provide higher levels of reactive power and ation of a wind turbine in a microgrid via reactive power
voltage support. Therefore, oversized inverters can be used support [28]. In this paper, voltage sensitivity analysis is used
(Fig. 3(b)) to increase maximum reactive power available to locally determine PV inverter reactive power due to its effi-
(Qmax
PV ) at rated PV power (P ) as cacy in quantifying power losses and bus voltage variations
 r
with respect to changes in PV and load profile.
PV = [(1 + γ ) . S] − Pr
Qmax 2 2 (1)
where γ is the inverter oversize factor. The THD limit also
imposes upper and lower bounds on reactive power generation V. P ROPOSED DYNAMIC C ONTROL S TRATEGY
specified by a given PV power factor (PF) as The proposed dynamic reactive power control strategy con-
  siders three states of operation and associated goals for each
−1
PV = PPV . tan cos PF
Qmax (2) PV unit in which the inverter is allowed to monitor its termi-
where tan is the tangent function. nal voltage. The first state (or normal state) is associated with
reactive power control in slow PV ramping periods (i.e., sunny
IV. VOLTAGE S ENSITIVITY A NALYSIS or overcast periods in which loss reduction is the main objec-
The purpose of voltage sensitivity analysis is to quantify tive). The second state (or fluctuating state) deals with fast
the relationship between nodal voltage magnitudes (V) and ramp-up and ramp-down PV power generation during inter-
angles (θ ) with respect to nodal active power (P) and reactive mittent cloudy periods with the goal of smoothing the voltage
power (Q) injections that are mathematically coupled by power profile throughout the feeder. The third state (or contingency
flow equations [12] as state) is experienced when PV terminal voltage deviates from
the normal range since low load with high generation or high

N
   
Pi = |Vi | · |Vj ·|Yij  · cos θij − δi + δj (3) load with low generation periods are possible operational sce-
i=1 narios. The objective is to eliminate or reduce overvoltage
 or undervoltage in the grid. Fig. 4. shows voltage evolution
N
   
Qi = |Vi | · |Vj ·|Yij  · sin θij − δi + δj (4) through time in normal, fluctuating, and contingency states
i=1 without reactive power control.
MALEKPOUR AND PAHWA: DYNAMIC OPERATIONAL SCHEME FOR RESIDENTIAL PV SMART INVERTERS 2261

A. Normal State QiPV calculated using (8) and replaces it with a Q(V) droop
The normal state includes two control logics based on local characteristic [9]
voltage. When voltages are within the predefined upper and i,(k)
QPV = QPV
i,(k)
lower critical voltages (VcrL , V U ), the objective is to provide ⎧ i,(k)
QPV + Qimax  i,(k) 
cr
⎪
⎪
reactive power for voltage support and power loss reduction ⎪
⎨ VPV − VcrU i,(k)
if VPV ≥ VcrU
in the network. If voltage reaches the upper (Vcr U ) or lower 1.05 − VcrU
− i  
⎪ i,(k)
⎪ QPV − Qmin V L − V i,(k) if V L ≤ V i,(k)
L
(Vcr ) threshold, the objectives are to provide reactive power ⎪
support for loss reduction, prevent local overvoltage or under- ⎩ L − 0.95 cr PV cr PV
Vcr
voltage, and maintain voltages within the normal range. The (9)
control strategies are local with no broadcast command needed
to switch between actual operation controls. In this paper, ini- where Qimax and Qimin are maximum and minimum PV inverter
i (k) < V U
reactive power capacities at node i. Whenever VPV
tially lower and upper critical voltages are set to 0.958 p.u. cr
and 1.042 p.u. around the normal operating voltage range from or VPV (k) > Vcr , the inverter reactive power control (9) is
i L

0.95 p.u. to 1.05 p.u. as an example for illustration of the pro- switched back to the original control (8).
posed methodology. Two additional ranges of critical voltages
were tested to see their effect on the results, which showed B. Fluctuating State
small changes in the results but no significant impact on the Transient cloud movement is the key cause for short-term
control scheme. intermittency in PV output. PV power variability is reflected in
Because calculation of exact power losses via local measure- rapid bus voltage or grid power fluctuations. The PV inverter
ments is impossible, deviation of PV bus voltage magnitude can react to rapid voltage and power changes in a few mil-
from nominal value is considered as an approximate measure liseconds. Fig. 4. shows transition from state 1 to state 2 as
for power losses. Once the difference between bus voltages voltage starts to fluctuate due to sudden PV generation drop
and unity is reduced, the voltage profile will be flatter and the around 100s.
losses will be decreased. Consider a feeder supplying a house Considering the local voltage regulation at bus i, voltage
through a line with impedance R + jX, the voltage drop at the variation can be approximated as
end of the line with respect to the beginning of the feeder can
be approximated as Vi = SVP
ii
.Pi + SVQ
ii
.Qi . (10)
    For ideal voltage regulation (Vi = 0), the PV inverter can
R. PiL − PiPV + X. QiL − QiPV make voltage variation caused by real power fluctuations zero
V = (6)
V by adjusting its reactive power output as
where PiL and QiL are the active and reactive power of load ii
SVP
and PiPV and QiPV are the active and reactive power of PV at Qi = − ii
· Pi . (11)
SVQ
bus i, respectively, and V is the nominal voltage [29], [30].
In order to achieve zero voltage drop (i.e., ideal voltage The variable reactive power control strategy to mitigate
regulation), the reactive power generation of PV unit i at the voltage fluctuations can be derived as
generic time step k should be ii
SVP
  Qi,(k) i,(k−1)
PV = QPV − ii
· PiPV (12)
i,(k) i,(k) i,(k) i,(k) SVQ
QPV = QL − (R/X). PL − PPV . (7)
where PiPV is PV power output variation at bus i. Variable
However, defining a unique R/X ratio for the reactive reactive power modulation is initiated when PV variation
power control logic in (7) is not straightforward due (PiPV /dt) is greater than the PV ramp rate threshold (ε).
to various network parameters, conductor types (cable or
overhead lines), and feeder length. In contrast, voltage sen- C. Contingency State
sitivity to active/reactive power variations at each bus can
Whenever the voltage goes outside of the normal operat-
be calculated for each network [12]. Hence, (7) can be
ing range, the system transitions to voltage deviation state.
reformulated as
In this state, the objective is to inject/absorb reactive power
ii
SVP   so that voltages are pushed back to the normal range. The
i,(k) i,(k) i,(k) i,(k)
QPV = QL − ii
. PL − PPV (8) control includes two control logics to prevent overvoltage or
SVQ undervoltage situations.
1) Overvoltage Control: A PV unit with terminal voltage
ii , Sii are voltage sensitivity indices at bus i due to
where SVP VQ of 1.05 p.u. would have fully utilized its reactive power adjust-
1 p.u. active/reactive power change at bus i, respectively. ment capability and cannot mitigate the overvoltage without
i,(k)
While QPV is calculated using (8) at each time step, proceeding to active power curtailment. Further, accord-
i,(k)
PV terminal voltage VPV is monitored to determine if it is ing to electrical requirements for inverter-based distribution
U
higher than Vcr or lower than VcrL . If so, the PV inverter con- systems [12], [31] the upper operating voltage bound (VUB )
trol overrides the regular reactive power provisioning using is set to 1.058 p.u. Hence, to prevent additional voltage rise,
2262 IEEE TRANSACTIONS ON SMART GRID, VOL. 8, NO. 5, SEPTEMBER 2017

no active power generation is allowed beyond 1.058 p.u. But tracks both the voltage at the PCC and PV power variations.
for terminal voltage between 1.05 p.u. and 1.058 p.u., the PV If voltage violation occurs for the past 10 seconds and PV
inverter curtails active power as shown in Fig. 2(b). More power variation (PiPV /dt) > ε remains, then PV operational
specifically, the PV reactive power setting is frozen to Qimin control transitions from state 2 to state 3. If the current state
while the dynamics of PV active power are chosen as is 3 and PV power fluctuation is detected, the control for state
 i,(k)
 2 is activated. Meanwhile, voltage at the PCC and PV power
i,(k) i,(k) VPV − 1.05 variations are monitored. If voltage violation is not present
PPV = PPV . 1 − . (13)
1.058 − 1.05 in the past 10 seconds and (PiPV /dt) > ε, the control in
state 2 continues. If (PiPV /dt) < ε, operational control tran-
If overvoltage persists for a predefined period, OLTC steps
sitions from state 2 to state 1. The transition to state 2 is fast
down the tap position in order to achieve voltage regulation
in order to cancel out or smooth voltage fluctuations caused
within operating voltage range (0.95-1.05 p.u.).
by power intermittency of the PV inverters. For transitions
from state 2 to states 1 and 3, however, a reactive power ramp
D. Undervoltage Control rate limiter (εQ ) is needed to prevent additional voltage fluc-
Referring to (8), the PV inverter provisioned its maximum tuations caused by reactive power swings from state to state
reactive power if its terminal voltage has reached the lower transitioning.
bound. Typical actions that can be taken to push the voltage ⎧  
back to the normal range are 1) upstream capacitor switching ⎨ Qi,(k) , if Qi,(k) − Qi,(k−1)  < εQ
QiPV (k) =  i,(k−1) 
PV PV PV
(if any), and 2) OLTC/regulator step-up voltage regulation in ⎩ Q + ε σ +
i,(k)
PV Q Q PV (1 − σ ), otherwise
order to increase voltage.
(16)
E. OLTC Operation Control Because out-of-bound voltage could trigger the OLTC oper-
The primary purpose of an OLTC transformer is to keep ation, a reactive power injection rate (σ ) is added to provide
the voltage on the low voltage side of the power transformer a trade-off between smooth reactive power variation and tap
within the standard regulatory range. OLTC control strate- changing prevention.
gies include [32] setting fixed voltage at the transformer’s low TD − Tt
voltage side, using a line-drop compensator, measuring local σ = (17)
TD
power flow, and utilizing remote voltage measurements. In this
paper, the latter strategy is applied based on minimum and The σ value close to 1 represents a situation in which OLTC
maximum voltage measured at the PCC. is less likely to operate, so reactive power injection could
The control method initiates a time delay counter (Tt ) that smoothly change to the desired calculated value (QiPV (k)). As
counts up when the measured remote voltage is out of standard σ approaches 0, the OLTC is more likely to operate; there-
regulatory bound and counts down to zero when the measured fore, the inverter injects the desired calculated reactive power
remote voltage is in bound. (QiPV (k)) depending on the state of operation.
 i,(k)
Fig. 5 shows the flowchart for the proposed coordinated
max(0, Tt − 1), if Vmin ≤ VPV ≤ Vmax reactive power and voltage control scheme for the PV inverter
Tt = (14)
Tt + 1, otherwise and OLTC operation.
The primary purpose of a timer is to provide a time delay
(TD ) in order to prevent unnecessary OLTC operations due VI. S IMULATION R ESULTS
to temporary voltage fluctuations. Once Tt becomes greater A. Network Description
than TD , a control pulse is sent to the OLTC mechanism in
A distribution network based on IEEE 37-node test feeder
order to move the tap up or down by one position.
 is developed to investigate performance of the proposed PV
Tap + 1, if Vmax < Vi (k), Tt > TD inverter operational strategy. Fig. 6 shows the modified IEEE
Tap = (15)
Tap − 1, if Vi (k) < Vmin , Tt > TD 37 node test feeder in which the extended network is modeled
by lines branching out of primary feeder 32 (i.e., from node
The time delay is set to 50 seconds in this paper [32].
39 to 42). The assumption was made that each distribution
transformer distributes power to a neighborhood consisting of
F. Coordination Between the States four homes. Characteristics of the transformers and service
In order to achieve the desired objectives in each state drop cables are described in detail in [19]. All nodes in the
and mitigate the adverse impact of high penetration of PV original IEEE 37-node test feeder are extended using the same
on power quality of the system while reducing operation of analogy.
OLTC, a coordinated control is required. In the proposed coor- The modified system consists of 560 nodes and 144, 144,
dination scheme, PV active power variations and terminal bus 160 homes in phase A, B, C, respectively. PV-enabled homes
voltage are used to determine the PV inverter operating state. If are selected randomly in each phase with 50% PV penetration
the current state is 1 and PV power intermittency begins, oper- per phase. In particular, PV-enabled homes in the extended
ational control for state 2 is triggered. While reactive power feeder are located at nodes 44, 46, 52, 55, 56, 57, 59, 60,
support continues during PV power fluctuations, the inverter 61, and 62. OLTC tap position ranges from 0.95 to 1.05 with
MALEKPOUR AND PAHWA: DYNAMIC OPERATIONAL SCHEME FOR RESIDENTIAL PV SMART INVERTERS 2263

Fig. 5. Flowchart of the proposed reactive power control strategy.

Fig. 7. Typical home data (left) and global horizontal irradiance (right).

Home load data were extracted from the eGauge

website [33], which provides load data with up to 1-minute
resolution. Typical home data is shown in Fig. 7. Load reactive
power is defined in proportion to the real load connected at the
Fig. 6. Modified IEEE 37node test feeder.
same bus with a power factor of 0.9 lagging. The 1-second res-
olution PV generation is obtained from NREL data measured
from a station near Hawaii’s Honolulu International Airport
tap steps of 0.01. In order to select ε parameter, the PV out- on the island of Oahu [34]. Global Horizontal Irradiance is
put power ramp-rate is measured during slow variations. We shown in Fig. 7, which includes both the clear sky and tran-
observed that more than 99% of the ramp-rates were less than sient cloud movement periods. The assumption was made that
15 W/s. Therefore, ε is set slightly higher than that to 20 W/s. the nodes are geographically close in the network such that
εQ parameter is set to 50% of the var available based on heuris- outputs of PV units follow the same generation pattern. In
tic. If larger values are selected, additional voltage fluctuations addition, inverters are assumed to be oversized by 20% and
could happen due to large reactive power swings originating provide reactive power up to 0.8 power factor.
from state to state transitioning. Conversely, selecting smaller
values could degrade performance of the proposed control
strategy in damping voltage fluctuations because only a small B. Results
capacity of inverter for reactive power injection/absorption will A three-phase unbalanced power flow was developed in
be used. MATLAB to implement the proposed scheme. Results from
2264 IEEE TRANSACTIONS ON SMART GRID, VOL. 8, NO. 5, SEPTEMBER 2017

Fig. 8. Voltage sensitivity coefficients.

Fig. 9. PV active power and reactive power generation at bus 439.

the developed program for the test system with no controls

were verified with the results obtained from OpenDSS [35].
Because the control time step should be granular enough to
capture dynamic interactions among PV inverters and OLTC
in various operational states, test results were obtained by
sequentially running load flow and implementing the proposed
scheme over a period of one day. The proposed control strategy
was compared to the following PV control strategies: 1) unity Fig. 10. Effect of the proposed approach on tap operation and voltage
power factor, 2) cosϕ(P) control, and 3) Q(V) control. Voltage at bus 439. (a) Voltage profile with and without the proposed approach,
(b) comparison of voltage with the cosϕ(P) method, and (c) comparison of
sensitivity coefficients for all the nodes with solar PV are voltage with the Q(V) method.
presented in descending order in Fig. 8. The values of Svp
ranged from 0.1366 to 0.0759 and that of Svq ranged from
0.0606 to 0.0238. Since the voltage did not exceed the upper The proposed reactive power control strategy has also been
limit of 1.042 p.u. in the simulations, the overvoltage control compared to the cosϕ(P) and the Q(V) control methods, as
mentioned in (13) was not triggered. shown in Figs. 10(b) and 10(c). Notice that the cosϕ(P)
Figs. 9-11 show results including tap operation due to method consumes reactive power in high feed-in power peri-
varying irradiance in a zoomed-in duration of 500 seconds. ods and provides no reactive power provisioning for the rest of
Fig. 9 shows the PV active power and reactive power genera- the simulation time. In fact, the cosϕ(P) strategy only consid-
tion for 500 seconds and state transitioning for the PV inverter ers the weather condition (solar irradiance) to enforce reactive
at bus 439. Fig. 10(a) shows voltage at bus 439 and the tap power support; it simply assumes that grid voltage increases in
changer operation with and without the proposed PV inverter high solar irradiance periods regardless of load variation. As
reactive power control strategy. a result, the control method fails to mitigate voltage fluctua-
The proposed strategy effectively mitigates voltage fluctu- tions caused by moving clouds and provides poor performance
ations when the PV inverter operates in state 2 and provides during high PV generation periods by causing voltage drop,
voltage support when it operates in state 1. Further, operation thereby triggering the tap operation more than that of other
in state 3 requires only one tap operation within the zoomed-in methods (see Fig. 10(b) and Fig. 11).
study period to push voltage into the range, whereas the unity Although basically designed to prevent voltage violation, the
power factor control needs two tap operations. Q(V) strategy could not mitigate voltage violation, as shown in
MALEKPOUR AND PAHWA: DYNAMIC OPERATIONAL SCHEME FOR RESIDENTIAL PV SMART INVERTERS 2265

TABLE I
DAILY P OWER L OSSES , N UMBER OF OLTC O PERATIONS AND
MVD VIA U NITY P OWER FACTOR , COSϕ(P), Q(V),
AND DYNAMIC A PPROACHES

Fig. 12. VFI at buses 11, 439 with unity power factor, cosφ(P), Q(V) and
proposed methods.

where M is the number of time steps of one second. In the

paper, M = 86400 was used.
As shown in Table I, without control (unity power factor
method as base case), PV output variations cause significant
voltage deviation in bus voltages. In particular, MVD experi-
enced at bus 439 as voltage drop is 0.0603 p.u. and the number
Fig. 11. Tap operation with cosϕ(P), Q(V) and proposed methods. of OLTC operations are 3.
Application of the cosϕ(P) method results in 3 tap opera-
Fig. 10(c) because in this strategy, only PV inverters with ter- tions throughout the day, while MVD is 0.0575 p.u. at bus
minals voltages outside the predefined range (i.e., 0.958-1.042) 439. Moreover, power losses in the ABC phases increase
provide reactive power provision. As a result, only some PV 2.24%, 1.57%, and 2.18%, respectively. When the Q(V) con-
inverters provide reactive power support based on their termi- trol strategy is used, the bus voltage at node 439 exhibits
nal voltage. In addition, the Q(V) strategy could not address MVD of 0.0551 p.u., and power losses in the ABC phase
voltage fluctuation issue. are slightly reduced by 0.47%, 0.28%, and 0.5% with respect
In contrast to the cosϕ(P) and Q(V) control methods, all to the base case. Three tap operations are recorded for this
PV inverters in the proposed approach contribute via reactive strategy. The proposed dynamic approach is the case in which
power support in order to achieve the goal associated with their performance metrics show maximum improvement. In the pro-
state of operation as determined by solar irradiance variation posed approach, MVD experienced at bus 439 improves from
and terminal voltage. The proposed strategy is a perfect match 0.0603 p.u. to 0.0545 p.u.; power losses reduce by 2.5%,
to react when the PV output is subject to rapid variation due to 4.06%, and 3.93% in respective phases, and the numbers of
scattered clouds passing over the system. Proposed approach OLTC operations decrease to 2. Compared to other strategies,
also offers voltage support during cloud shadows with low PV power losses reduce because PV generation is matched with
generation and prevents voltage violation via dynamic reac- load consumption a majority of the time.
tive power injection. Although the number of OLTC operation Fig. 12 displays the VFI at selected buses; i.e., PV-enabled
within the zoomed-in study period is 2 for the unity power homes located at bus 439 (connected to phase A) and bus 11
factor and Q(V) control strategies and 3 for the cosϕ(P) con- (three phase bus). When the unity power factor method is uti-
trol strategy, the tap operation is reduced to 1 for the proposed lized, VFI at bus 439 is 0.6375 p.u., whereas VFI at bus 11
approach. in ABC phases are 0.4263 p.u., 0.2027 p.u., and 0.2915 p.u.,
respectively. With the cosϕ(P) control, VFI at bus 439 and
phase A of bus 11 is decreased by 18.86% and 13.6%, while
C. Performance Analysis
VFI is increased by 52.45% and 39.6% at bus 11 of phase B
For quantitative comparison, whole-day simulations using and C, respectively. VFI associated with the Q(V) control in
the investigated PV inverter reactive power control methods bus 439 (phase A) and 11 (ABC phase) are reduced by 4.62%,
are summarized in Table I and Fig. 12. Performance metrics 4.03%, 1.85%, and 0.03%, respectively. When the proposed
include maximum voltage deviation (MVD) from the nominal strategy is implemented, VFI values at bus 439 (phase A) and
voltage limits, total power losses, and the number of OLTC tap 11 (ABC phase) are effectively decreased by 45.93%, 50.18%,
operations. In addition, the summation of voltage magnitude 11.48%, and 14.82%, respectively.
change between the current and previous time slice at each The results shown above are based on setting the upper and
bus for the duration of the simulation is defined as the bus lower critical voltages to 1.042 p.u. and 0.958 p.u., respec-
voltage fluctuation index (VFI). tively. We consider it the base case with total power losses of

M 1453.067 kW and MVD of 0.0545 p.u. In order to examine the
VFI = |Vi (k) − Vi (k − 1)| (18) effect of these limits on the results with the proposed dynamic
k=1 approach, two additional cases are studied; 1) the upper and
2266 IEEE TRANSACTIONS ON SMART GRID, VOL. 8, NO. 5, SEPTEMBER 2017

lower critical voltages are set to 1.034 p.u. and 0.964 p.u., [3] Generators Connected to the LV Distribution Network—Technical
respectively; and 2) the upper and lower critical voltages are Requirements for the Connection to and Parallel Operation With
Low-Voltage Distribution Networks, VDE Standard VDE-AR-N 4105,
set to 1.046 p.u. and 0.954 p.u., respectively. Compared to the 2011.
base case, no significant change in VFI and no change in the [4] Reference Technical Rules for Connecting Active and Passive Users to
number of tap operations was observed for both Case 1 and Networks Low-Voltage Electricity Networks of Energy Providers, CEI
Standard CEI 0-21, 2011.
Case 2. MVD decreased to 0.0539 p.u. and total power losses [5] E. Demirok et al., “Local reactive power control methods for over-
increased to 1468.132 kW for Case 1, but in contrast, MVD voltage prevention of distributed solar inverters in low-voltage grids,”
increased to 0.0549 p.u. and total power losses decreased to IEEE J. Photovolt., vol. 1, no. 2, pp. 174–182, Oct. 2011.
1449.961 kW for Case 2. However, these changes are very [6] T. Stetz et al., “Techno-economic assessment of voltage control strate-
gies in low voltage grids,” IEEE Trans. Smart Grid, vol. 5, no. 4,
small compared to the base case and do not have significant pp. 2125–2132, Jul. 2014.
impact on the control scheme. [7] T. Stetz, F. Marten, and M. Braun, “Improved low voltage grid-
integration of photovoltaic systems in Germany,” IEEE Trans. Sustain.
Energy, vol. 4, no. 2, pp. 534–542, Apr. 2013.
VII. C ONCLUSION [8] A. Samadi, R. Eriksson, L. Söder, B. G. Rawn, and J. C. Boemer,
“Coordinated active power-dependent voltage regulation in distribu-
This paper presents a dynamic PV inverter reactive power tion grids with PV systems,” IEEE Trans. Power Del., vol. 29, no. 3,
control scheme to facilitate high penetration of rooftop PV in pp. 1454–1464, Jun. 2014.
distribution systems under all-weather-condition. Three states [9] J. W. Smith, W. Sunderman, R. Dugan, and B. Seal, “Smart inverter
volt/var control functions for high penetration of PV on distribution
for operation of PV inverters are proposed using irradiance systems,” in Proc. IEEE/PES Power Syst. Conf. Expo., Phoenix, AZ,
variation and voltage measurement at the PCC. In the nor- USA, 2011, pp. 1–6.
mal state, the scheme is designed to provide loss reduction [10] P. M. S. Carvalho, P. F. Correia, and L. A. F. M. Ferreira, “Distributed
reactive power generation control for voltage rise mitigation in distribu-
support while continuing to support load demand when irradi- tion networks,” IEEE Trans. Power Syst., vol. 23, no. 2, pp. 766–772,
ance changes slowly. In the fluctuating state, reactive power is May 2008.
modulated in order to mitigate voltage fluctuations caused by [11] Y. Wang, P. Zhang, W. Li, W. Xiao, and A. Abdollahi, “Online overvolt-
transient cloud movement. Finally, the control offers reactive age prevention, control of photovoltaic generators in microgrids,” IEEE
Trans. Smart Grid, vol. 3, no. 4, pp. 2071–2078, Dec. 2012.
power support in order to mitigate voltage violation defined [12] R. Tonkoski, L. A. C. Lopes, and T. H. M. El-Fouly, “Coordinated
for the contingency state. A reactive power ramp rate limiter active power curtailment of grid connected PV inverters for overvoltage
is introduced in order to prevent additional voltage fluctu- prevention,” IEEE Trans. Sustain. Energy, vol. 2, no. 2, pp. 139–147,
Apr. 2011.
ations due to reactive power variation caused by switching [13] R. F. Yan, B. Marais, and T. K. Saha, “Impacts of residential photovoltaic
between the states. A coordination strategy is proposed to power fluctuation on on-load tap changer operation and a solution using
switch control between the states and manage interaction DSTATCOM,” Elect. Power Syst. Res., vol. 111, pp. 185–193, Jun. 2014.
between fast PV inverter controllers and slow OLTC for volt- [14] M. J. E. Alam, K. M. Muttaqi, and D. Sutanto, “A multi-mode control
strategy for VAr support by solar PV inverters in distribution networks,”
age regulation. Numerical tests on modified IEEE 37-node test IEEE Trans. Power Syst., vol. 30, no. 3, pp. 1316–1326, May 2015.
feeder demonstrated that the proposed approach is advanta- [15] Generating Dollars Where There’s No Sun. (Dec. 2014). [Online].
geous compared to the unity power factor, the cosϕ(P), and the Available: http://www.pvmagazine.com/archive/articles/beitrag/
generating-dollars-where-theres-no-sun-100003490/329/#axzz3z77tGFAt
Q(V) reactive power control strategies and that the proposed [16] Y. Meng and X. Wang, “Dynamic modeling of grid integrated solar and
approach improves the operational performance of distribution battery energy system for cloud transient impact study,” in Proc. IEEE
systems with high rooftop PV penetration. Results demonstrate Energytech, Cleveland, OH, USA, 2013, pp. 1–6.
[17] E. Abdelkarim, M. M. Aly, M. Abdel-Akher, Z. Ziadi, and T. Senjyu,
superior performance of the proposed approach for reducing “Supersession of large penetration photovoltaic power transients using
the number of OLTC operations, decreasing power losses, storage batteries,” in Proc. IEEE 10th Int. Conf. Power Electron. Drive
smoothing out voltage fluctuations, and mitigating voltage vio- Syst. (PEDS), Kitakyushu, Japan, 2013, pp. 78–83.
lation in the LV grid. The results are promising and they [18] Industry Guide 2014. (May 2014). Inverter, Storage and PV System
Technology. [Online]. Available: http://www.pv-system-tech.com/
can be extended to study coordination between inverters for technology/storage-and-energy-management/
mutual inferencing and implementing the approach to three- [19] A. R. Malekpour and A. Pahwa, “Reactive power and voltage control in
phase inverters in the future. Additionally, grid connection was distribution systems with photovoltaic generation,” in Proc. North Amer.
Power Symp. (NAPS), Champaign, IL, USA, 2012, pp. 1–6.
assumed in this research. Implementation of the scheme in [20] G. C. Christoforidis, A. Chrysochos, G. Papagiannis, M. Hatzipanayi,
islanded mode and night time operation are topics for future and G. E. Georghiou, “Promoting PV energy through net metering opti-
research. mization: The PV-NET project,” in Proc. Int. Conf. Renew. Energy Res.
Appl. (ICRERA), Madrid, Spain, 2013, pp. 1117–1122.
[21] M. S. Arifin, M. M. S. Khan, A. Haque, and N. Al-Masood,
ACKNOWLEDGMENT “Improvement of load-margin and bus voltage of Bangladesh power
system with the penetration of PV based generation,” in Proc. Int. Conf.
The authors would like to thank NSF for providing financial Inf. Electron. Vis. (ICIEV), Dhaka, Bangladesh, 2013, pp. 1–5.
support for the project. The views expressed in this paper are [22] Limited Export and Reactive Power Control. (Jan. 2015). [Online].
Available: http://soldistribution.com.au/sma-solar-training-limited-
of the authors. export-and-reactive-control
[23] Solar Integration Technology. (Oct. 2014). [Online]. Available: http://
www.emmos.org/prevconf/2010/12-2-SMA_America_Preserntion.pdf
R EFERENCES
[24] M. H. Rashid, Alternative Energy in Power Electronics, 1st ed. Oxford,
[1] IEEE 1547 Standard for Interconnecting Distributed Resources With U.K.: Butterworth-Heinemann, 2014.
Electric Power Systems, IEEE Standard 1547, 2003. [25] Description of the Operating Parameters SUNNY TRIPOWER 60.
[2] IEEE 1547 Standard for Interconnecting Distributed Resources With (Apr. 2014). [Online]. Available: http://files.sma.de/dl/25977/STP60-10-
Electric Power Systems, IEEE Standard 1547 Amendment 1, 2014. Parameter-TI-en-10.pdf
MALEKPOUR AND PAHWA: DYNAMIC OPERATIONAL SCHEME FOR RESIDENTIAL PV SMART INVERTERS 2267

[26] T.-F. Wu, C.-L. Kuo, K.-H. Sun, and H.-C. Hsieh, “Combined unipolar Ahmad Reza Malekpour (S’06) received the B.S. and M.S. degrees from
and bipolar PWM for current distortion improvement during power com- Shiraz University, Shiraz, Iran, in 2003 and 2006, respectively. He is cur-
pensation,” IEEE Trans. Power Electron., vol. 29, no. 4, pp. 1702–1709, rently pursuing the Ph.D. degree with Kansas State University, Manhattan.
Apr. 2014. His research interests include volt/var control, integration of wind and solar
[27] R. Tonkoski and L. A. C. Lopes, “Voltage regulation in radial distribution power, and stochastic and distributed optimization in power system.
feeders with high penetration of photovoltaic,” in Proc. IEEE Energy
2030 Conf., Atlanta, GA, USA, 2008, pp. 1–7.
[28] R. Aghatehrani and R. Kavasseri, “Reactive power management of
a DFIG wind system in microgrids based on voltage sensitivity analysis,”
IEEE Trans. Sustain. Energy, vol. 2, no. 4, pp. 451–458, Oct. 2011.
[29] R. Walling and Z. Gao, “Eliminating voltage variation due to
distribution-connected renewable generation,” in Proc. DistribuTECH
Conf. Expo., San Diego, CA, USA, Feb. 2011, pp. 1–6.
[30] A. R. Malekpour, A. Pahwa, and S. Das, “Inverter-based var con-
trol in low voltage distribution systems with rooftop solar PV,” in
Proc. North Amer. Power Symp. (NAPS), Manhattan, KS, USA, 2013,
pp. 1–5.
[31] R. Tonkoski, D. Turcotte, and T. H. M. El-Fouly, “Impact of
high PV penetration on voltage profiles in residential neighbor- Anil Pahwa received the B.E. (Hons.) degree from BITS, Pilani, India, in
hoods,” IEEE Trans. Sustain. Energy, vol. 3, no. 3, pp. 518–527, 1975; the M.S. degree from the University of Maine, Orono, in 1979; and the
Jul. 2012. Ph.D. degree from Texas A&M University, College Station, in 1983, all in
[32] R. Caldon, M. Coppa, R. Sgarbossa, and R. Turri, “A simplified algo- electrical engineering. Since 1983, he has been with Kansas State University,
rithm for OLTC control in active distribution MV networks,” in Proc. Manhattan, where he is currently Professor and holds the Logan-Fetterhoof
AEIT Annu. Conf., 2013, pp. 1–6. Chair in the Electrical and Computer Engineering Department. The National
[33] eGauge Systems. (Mar. 2013). [Online]. Available: Academy selected him as a Jefferson Science Fellow in 2014 to serve as
http://out.easycounter. comlextemallegauge360.egaug.es a Senior Science Advisor in the U.S. State Department for one year. He
[34] Oahu Solar Measurement Grid. (Mar. 2013). [Online]. Available: worked in the East Asian and Pacific Affairs Bureau on international policies
http://www.nrel.gov/midc/oahu_archive/ to facilitate higher deployment of renewable energy. His research inter-
[35] Electric Power Research Institute (EPRI). (Jul. 2014). Open- ests include distribution automation, distribution planning, renewable energy
DSS Simulating Environment Software. [Online]. Available: integration into power systems, and intelligent computational methods for
http://smartgrid.epri.com/ distribution system applications.