You are on page 1of 5

Simulation of Integrated Volt/VAR Control for PV

Penetration Studies
Daixi Li and David L. Lubkeman
Department of Electrical and Computer Engineering
North Carolina State University
Raleigh, NC 27606

Abstract—Delivering voltage within suitable range to the cus- Quasi-static time series (QSTS) simulation, supported by
tomers is one of utility’s responsibilities. To keep voltage within software such as OpenDSS, facilitates a more detailed time-
ANSI C84.1 limits, utilities have different voltage regulation tech- domain distribution study by providing the capability to see the
niques, and Volt/VAR control is one commonly used approach in
distribution systems.The objective of this paper is to explore the dynamic interaction between control devices under different
topic related to the voltage control of electric power distribution timeframes. There have been a number of papers published
circuits with high penetration levels of distributed photovoltaic on PV impact studies, flicker problems and inverter control
(PV) power systems. An Integrated Volt/VAR Control (IVVC) based on QSTS simulation [6]–[12]. Yet, there is a lack of
model is developed for use with the Open Distribution System a viable Integrated Volt/VAR Control simulation scheme for
Simulator (OpenDSS) simulation platform for emulating the
effectiveness of centralized controls on mitigating distributed properly modeling Volt/VAR interaction with PV penetration,
photovoltaic system voltage impacts. The test results of the IVVC which could be implemented with OpenDSS time-series study.
control prove that the proposed control can successfully keep This paper proposes an Integrated Volt/VAR Control simu-
the voltage within the tight target voltage range, for applications lation that can be used in conjunction with OpenDSS which
such as Conservation Voltage Reduction (CVR). A utility-scale emulates the behavior of control schemes utilities might be
distribution feeder is used as the test circuit for this study.
Index Terms—Distribution Circuit Simulation, OpenDSS, PV
deploying in practice. The following sections describe the
System Penetration Analysis, Volt/VAR Control. additional logic needed for such a simulation. Examples are
provided using a typical utility distribution circuit.
I. I NTRODUCTION II. C ONTROL M ETHOD D ESCRIPTION
Most utilities in the United States follow the ANSI C84.1 The control utilizes a hybrid capacitor banks (CAPs)/Load
standard [1], which defines the allowable operational voltage Tap Changer (LTC)/Line Voltage Regulator (LVR) technology.
range for utilization voltage and service voltage. It is the The objective of the control is to minimize circuit losses,
utility’s responsibility to maintain the service entrance voltage substation VAR flow and keep circuit voltage profiles within
within service range [2]. Some utilities would keep the voltage the desired ranges. The proposed approach is to choose a
in an even tighter range for implementing CVR. To keep volt- control action at each timestep which minimizes the defined
age within acceptable tolerance, utilities have different voltage objective function. Equations (1) - (2) describe the objective
regulation techniques, and centralized Volt/VAR control is one function in the control, where A is the weighting factor for
commonly used approach in distribution systems. Researchers circuit loss, CircuitLoss is circuit kW loss at a given time
have proposed different kinds of Volt/VAR algorithms [3]– step, and B is the load weighting factor for the selected pilot
[5]. However, most of those controls require modeling the test points. Verror is an array calculated using the absolute voltage
circuits and integrating power flow models into the control deviations from the upper and lower limits. The expected
algorithm. outcome of the proposed control is that the voltages stay
Photovoltaics (PV) integration to the power system has within chosen bandwidth and substation VAR measurement
grown exponentially in recent years which raises technical to be close to zero.
concerns about adverse impacts of PV on distribution systems.
Firstly, PV power output is subject to irradiance fluctuations Obj=A · CircuitLoss + B · Verror (1)
due to cloud movements, which results in output power Verror =VBus − Vupper limit + Vlower limit − VBus (2)
variability. Highly variable PV power being injected into the
system will create intermittent voltage problems. Secondly, A. Volt/VAR Control Modeling Procedure
PV interconnection can cause over-voltage problems near PV The flow chart of the control is shown in Fig. 1. Results
connection buses. Then, there is also a problem with reverse from the power flow are collected at the beginning of each
flow of power. In a conventional distribution system, power timestep, including the circuit losses, the measurement of
flows unidirectionally from substation to customers. With high voltages and substation VARs, and the initial state of each
levels of PV integration, power may flow backwards which control device. Depending on the number of regulators, test
creates challenges with legacy circuit designs and control circuits can be divided into several voltage control zones. The
techniques. measured voltages from pilot points are processed and divided

‹,(((
into groups based on the zone of the buses. In reality, multiple TABLE I
voltage regulate devices will have different time delays in VOLTAGE C ONSTRAINTS USED IN THE C ONTROL
order to coordinate with one another. In this control, time delay Voltage Limit Lower Limit Upper Limit
is modeled as priority of a voltage regulation device. When CVR bandwidth 122V 124V
a tap change is in need for multiple regulators, the upstream Power Quality bandwidth 0.95 p.u. 1.05 p.u.
device will have higher priority over downstream device.
Control actions for voltage regulators are determined by
comparing voltage measurements with the chosen voltage
constraints. If voltages exceed the upper limit, the controller
will send a command to reduce a tap position. When the
lower voltage limit is violated, the control will indicate an
increase in the tap position. The three phases of the voltages
are being controlled independently for both substation LTC
and LVR. The tap which needs to be changed is selected by
a lookup table based on the phase of the voltage violation
and the control zone of the violation bus. Whenever a control
decision is determined and executed, the power flow will be
performed again, and a new value for objective function will
be calculated based on the new results from the power flow.
Capacitor banks are controlled in a slightly different way. VAR
flow at the substation is measured and used to compare with
control criteria. When Q > 12 Qrated , CAPs will receive
command for switch-in. When Q < − 34 Qrated , a CAP is
required to switch out. Unlike the voltage control, CAPs have
no pre-defined priority. Therefore, each available CAP will be
switched in/out (based on chosen control decision) to calculate
a new value for the objective function. At the end of each
timestep, all the possible values from previous calculations for
the objective function will be compared and the control action
which has the best impact on the circuit will be chosen. This Fig. 1. Flow Chart for the Proposed Volt/VAR Control
control procedure is called the base control procedure.

B. Additional Steps for Voltage Constraints comparing these two values, the control will decide whether
Table I specifies the two ranges for voltage constraints. to reduce or increase the tap position. The higher result means
Keeping voltages within CVR bandwidth is the primary target a more severe voltage violation, so it will be taken care first.
of the proposed control. Since CVR bandwidth is a relatively For instance, suppose bus k has 5 kW load and exceeds 124V
tight range for voltages, there are cases where base control by 0.05 p.u., and another bus m has 3 kW with 0.02 p.u.
procedure cannot keep the voltage within the CVR bandwidth deviation from 122V. Given the results of bus k is larger (i.e.
range; voltages at pilot points violate the two boundaries of 0.25 > 0.06), one can come to the conclusion that the upper
the CVR bandwidth simultaneously. Power Quality bandwidth limit violation is more severe. Thus, the chosen tap position
(i.e. the ANSI C84.1 standard) is introduced to deal with these need to be reduced.
cases. Additional steps are designed and shown in Fig. 2. Unlike the previous base control procedure, voltage reg-
When voltages are higher than 124V and lower than 122V, ulators are not selected by lookup table. Instead, every tap
we consider ANSI criteria (i.e. Power Quality bandwidth as controller will execute the chosen control decision which
shown in Table I) as the new voltage constraints. If only one results in a new set of values for the objective function.
limit is exceeded under the new constraints, voltage regulator The minimum of this set of values will be found and then
control follows the same steps as before: make corresponding compared with the values of objective function resulted from
tap changes and calculate a new value for the objective the base control procedure. Whichever value is the minimum,
function. If the voltages at the pilot points violate the CVR the corresponding action will be taken as the action of this
bandwidth but are still within the range of Power Quality timestep. In brief, only one control (or even none) is chosen at
bandwidth, the load weighting factors of the buses that violated each timestep. We choose the control action which minimizes
the CVR bandwidth will be determined. The load weighting the objective function, because it brings the biggest benefits to
factors times the voltage deviation from the upper limit of the system. There is no intentional delay for voltage regulators’
CVR bandwidth will results in one value, and the product of tap change or CAP switch action.
the load weighting factors and the voltage derivation from the To ensure a fair comparison, the circuit is always reset
lower limit of CVR bandwidth will be a different value. By to an initial state after a new value of objective function is

‹,(((
Fig. 4. PV Output Power: Sunny Day and Cloudy Day

case with no PV, 2. Sunny day with two MW-scale PV plants,


3. Cloudy day with two MW-scale PV plants. These cases
are daily simulation which utilize 15-minute load data and 1-
second resolution PV data. To have a better observation of
the control, 10-sec simulation resolution is chose for all the
test cases which means the system status is updated every
Fig. 2. Additional Steps when CVR bandwidth is violated 10 second and control makes decision accordingly. There are
8640 timesteps in each case.
Fig. 4 shows the PV profiles used in the simulation which
are 1-second resolution data from EPRI website. Simulation
is conducted using sunny day PV profile and cloudy day PV
profile respectively [13]. Load data are the peak summer day of
2014 taken from the yearly profile of the test feeder. Before
running any test cases, pilot points must be determined by
running a voltage study. The results of the voltage study show
that there are many low-voltages on Phase A.

IV. T EST R ESULTS


A. Case 1: Base Case

Fig. 3. Test feeder with the location of the LTC, LVR, Pilot Points and PV
The base case has no PV interconnect to the test feeder.
Fig. 5 demonstrates that the Volt/VAR control has a good
performance, since mostly the voltages are kept between
calculated. Note that: 122V and 124V. There are times when voltage exceeds CVR
1) Regulator control: For each phase, the corresponding bandwidth, but even in those cases, the overall voltages always
phase of the voltage regulator is being controlled. stay within Power Quality bandwidth.
2) Choose only the best result of the Objective Function at The number of control actions for each case are summarized
each step. If no control is needed, go to next timestep. in Table II. Phase A of the LVR has the most control actions
during the day. This is expected because Phase A has the most
III. T EST C IRCUIT low-voltage buses in the voltage study. Based on the loading
A representative utility feeder circuit is used as testbed for on each phase, phase B and phase C have similar loading,
the proposed control scheme. It is an actual distribution circuit whereas Phase A is more heavily loaded.
with a nominal voltage of 12.47 kV. The main part of the Fig. 6 plots the voltage at bus 2041 (Phase A) which is a
feeder is approximately 3 miles, with a summer peak demand pilot point near the end of the feeder. In this plot, we can
of 6.7 MW. The test feeder has a mix of commercial load and have a better look at voltage variation due to different control
residential load, while the majority of the load is residential. movements. We can find in Fig. 6 that the voltages are shifting
Four shunt capacitors, with the same rating of 600 kVAR, are up and down because of tap changes. For example, around 6
connected to the feeder, and can be controlled to switch in am, there is a voltage drop at bus 2041 which is the result from
and out based on need. a tap position reduction on Phase A of the LTC. Also, some
There is one substation LTC and one line voltage regulator spikes in voltage can be observed in Fig. 6 which are due to
in the test feeder. Two voltage regulating devices results in interactions between different control devices. Fig. 7 illustrates
two voltage control zones: Zone 1 is the voltage zone being that the VAR measurement at the substation is following the
controlled by the substation LTC, while Zone 2 is monitored changes in the CAP operations. Whenever a CAP is switched
by the LVR. The location of two devices are marked in Fig. 3. in, the CAP injects VARs into the circuit and the need of VARs
Daily QSTS simulation is set up for testing the Volt/VAR from the substation is reduced. That’s why we can see a drop
control. Simulation scenarios are listed as follows: 1. Base in substation VAR around noon.

‹,(((
Fig. 5. Voltage Profile at Sample Pilot Points (Case 1) Fig. 8. Voltage Profile at Sample Pilot Points (Case 2)

Fig. 6. Voltage at Bus 2041 and Tap Changes of Phase A (Case 1)


Fig. 9. Voltage at bus 2041 and Tap Changes of Phase A (Case 2)

TABLE II
N UMBERS OF C ONTROL ACTIONS IN E ACH T EST C ASE

LTC LVR
Case # CAP A CAP B CAP C CAP D
A B C A B C
1 2 2 2 14 3 3 2 2 0 1
2 1 4 3 89 3 6 0 2 0 3
3 1 0 1 79 3 8 0 2 0 3

C. Case 3: Cloudy Day with PV


Fig. 7. Substation VAR Measurement and CAP Switching Actions (Case 1)

This case utilizes the cloudy day PV profile. The results


B. Case 2: Sunny Day with PV of this case show oscillations in tap controls and voltages
Case 2 utilizes sunny day PV profile to simulate the control. as well. There is no control cycling in the LTC. Still, Phase
Fig. 8 shows the good performance of the voltage control A of the LVR has more than 70 tap changes. In Table II, a
where voltages are mostly within the desired constraints. Yet, total of 97 control actions take place in this single day. The
there are some voltage oscillations during 16:00 and 20:00 oscillations in tap controllers and voltages need to be fixed to
which are caused by oscillations of regulators tap changes. avoid unnecessary wear and tear of the tap changing devices.
There are 89 tap operations at Phase A of the LVR (from We can also observe some voltage variation during 12:00 to
Table II). The excessive tap changes are due to implementing 16:00 (see Fig. 10). These variations in voltages are due to
the tight voltage constraints of “122V to 124V”. Further cloud movement during the same period. Unlike voltages, the
discussion into these unnecessary tap changes is discussed VAR flow through the substation is not greatly affected by PV.
in subsection D. Voltage spikes due to control devices’ in- In conclusion, the proposed control simulation managed to
teractions can still be observed from Fig. 9. Unlike voltage, keep the voltages and substation VAR close to the target.
the VAR measurement shows the same trend without any Voltage variation caused by cloud is clearly shown in this
oscillation which is because PV injects only real power into the case. The oscillations in voltage regulator control still exists
circuit. Thus, reactive power is less affected by PV integration. for Case 3 and will be discussed in subsection D.

‹,(((
TABLE III
N UMBERS OF C ONTROL ACTIONS FOR THE I MPROVED C ASES

LTC LVR
Case # CAP A CAP B CAP C CAP D
A B C A B C
4 1 0 3 11 3 4 0 2 0 3
5 1 0 1 9 1 2 0 2 0 3

lifetime of a voltage regulation device. Thus, it is worthwhile


to relax the bandwidth with a small margin value in exchange
for the avoidance of those unnecessary control actions.
Fig. 10. Voltage at bus 2041 and Tap Changes of Phase A (Case 3) V. C ONCLUSION
The proposed emulation of the Integrated Volt/VAR Control
is tested on three different cases. The results proved that the
control is working as expected, and successfully augmenting
OpenDSS for quasi-static simulation of Volt/VAR control.
While the control successfully enables a tight bandwidth
(i.e. 122V to 124V) for voltages and keep power factor
at substation close to unity, it is observed that cycling in
regulator control exists when PV plants are interconnected.
An improvement is suggested to solve the problem, and the
test results for the altered control show that the number of
tap changes is reduced by approximately 82%. The proposed
control can provide a good starting point for us to simulate the
Fig. 11. Voltage at Bus 2041 and Tap Changes of Phase A (Case 4) impact on PV penetration for distribution planning purposes.
R EFERENCES
D. Improvement for PV Integrated Cases [1] ANSI C84.1-2011. American National Standard for Electric Power
Systems and Equipment-Voltage Ratings (60 Hz), 2011.
There are several unnecessary tap changes on Phase A of the [2] W. H. Kersting, Distribution System Modeling and Analysis, Third Edi-
LVR when PV systems are integrated into the feeder. Further tion. CRC Press, Jan. 2012.
study shows that the cycling in control devices is due to the [3] I. Roytelman, B. K. Wee, and R. L. Lugtu, “Volt/VAR control algorithm
for modern distribution management system,” IEEE Transactions on
control’s immediate response to tiny derivations in voltages. Power Systems, vol. 10, no. 3, pp. 1454–1460, Aug. 1995.
These derivations are mostly smaller 10 × 10−5 p.u., however [4] X. Feng and F. Yang, “Voltage regulation optimization,” Patent US
based on control logic they are still considered to be out of 12/613,265, 2013.
[5] Z. Shen and M. E. Baran, “Gradient based centralized optimal Volt/Var
bandwidth. Thus, tap positions will be changed accordingly. control strategy for smart distribution system,” in Innovative Smart Grid
Yet this new tap position will incur a violation on the other Technologies ISGT, IEEE PES. Washington, DC: IEEE, Feb. 2013, pp.
voltage boundary, which causes the tap to move back to its 1–6.
[6] R. F. Arritt and R. C. Dugan, “Value of sequential-time simulations
initial setting. After this tap reset, the small deviations will in distribution planning,” IEEE Transactions on Industry Applications,
trigger a tap change again and this process will continue until vol. 50, no. 6, pp. 4216–4220, Dec. 2014.
the control reaches a steady state. To solve this problem, the [7] J. E. Quiroz, M. J. Reno, and R. J. Broderick, “Time series simulation of
voltage regulation device control modes,” 2013 IEEE 39th Photovoltaic
voltage control should be made to be less sensitive to small Specialists Conference (PVSC), pp. 1700–1705, 2013.
voltage violations. A margin of 8 × 10−4 is added to the tap [8] R. J. Broderick, J. E. Quiroz, M. J. Reno, and A. Ellis, “Time Series
control logic, so that the control will not react to voltage Power Flow Analysis for Distribution Connected PV Generation,” Tech.
Rep. SAND2013-0537, 2013.
violations less than 8 × 10−4 p.u. Essentially, the cycling in [9] M. Seethamraju, “Simulation of Smart Functionalities of Photovoltaic
voltage control happens because the voltage bandwidth is too Inverters by Interfacing Both OpenDSS and Matlab,” in 2015 IEEE 16th
tight. The problems will be gone if a wider bandwidth is used. Workshop on Control and Modeling for Power Electronics (COMPEL),
Vancouver, BC, Jul. 2015, pp. 1–6.
The sunny and cloudy PV cases are simulated again (Case [10] J. W. Smith, R. Dugan, and W. Sunderman, “Distribution modeling and
4 & 5). Results in Table III shows that the number of control analysis of high penetration PV,” Energy Society General Meeting, pp.
actions on Phase A of the LVR is greatly reduced. Fig. 11 also 1–7, 2011.
[11] R. Dugan, W. Sunderman, and B. Seal, “Advanced inverter controls for
proves that the cycling problems are solved in Case 4. Similar distributed resources,” 22nd International Conference and Exhibition on
findings can be obtained for the cloudy case with the added Electricity Distribution (CIRED 2013), pp. 1–4, 2013.
control margin (Case 5). [12] W. Sunderman and R. C. Dugan, “Open source modeling of advanced
inverter functions for solar photovoltaic installations,” in 2014 IEEE PES
Overall, we can conclude that the proposed amendment for T&D Conference and Exposition, Chicago, Apr. 2014, pp. 1–5.
the voltage control is a good solution for the cycling effect [13] “EPRI Distributed PV Monitoring and Feeder Analysis,” http://dpv.epri.
caused by PV integration. Reducing tap actions can extend the com/measurement data.html, Accessed: 2016-07-05.

‹,(((

You might also like