Active Voltage Control Using Distributed Generation

on Distribution Networks
He Yujun, Marc Petit
Dept. of Power and Energy Systems
Supélec E3S
Gif-sur-Yvette, France
[email protected]

Abstract— The increase of integration of distributed generations Besides voltage constraints, the reactive power support is
(DG) requires the network management to be more active. This also of concern to the operators in introducing massively DG.
paper investigates an active voltage control approach by [10] has investigated passive and active approaches of voltage
coordinating the reactive power of DG and the operation of on- control to provide reactive capability of DG. Both approaches
load tap changer (OLTC). A sensitivity model of the V-Q are able to minimize the reactive power but the active
coupling in distribution systems is used in the optimization of approach is more expensive in the implementation.
the coordinated voltage regulation. The performance of voltage
control has been verified and compared with the conventional In this paper an approach of active voltage control using
OLTC control. The results show that this approach is very the regulation of DG is investigated in order to improve the
promising. hosting of DG without violation of voltage constraint. In this
approach, the set point of the reactive regulation of DG is
Index Terms—Distributed generation, active voltage control, determined through an optimization process by using a
distribution network, sensitivity matrix. sensitivity model of the V-Q coupling in the distribution
network by taking into account the influence of line resistance
I. INTRODUCTION R, as the ratio of R/X is much higher than that in transmission
The distributed generation (DG) is usually connected systems (for MV 150 mm² cables, R is even greater than X).
within distribution systems in “fit and forget” mode. This
This paper is organized as follows: the coupling of V-Q of
operation mode is based on the fact that the traditional
a distribution feeder is presented in the first part of section II;
distribution network is passive. If the increase of DG affects
then the process of optimal regulation based on the sensitivity
the network safety, the distribution system operator (DSO) has
model is formulated in the rest part of this section. Section III
to disconnect some parks of DG and then a part of power
provides the validation of this approach by the simulations and
generation is wasted [1]. The growth of DG penetration
the discussions of their result. At the end, conclusions are
within actual power systems requires the network system
summarized in the section IV.
management to be smarter and more flexible. The
conventional voltage control using OLTC and banks of shunt II. SYSTEM DESCRIPTION
capacitor (ShC) is known as “voltage/var control”. However,
A. The Coupling of DG’s Power Output and Voltage Profile
it becomes less efficient to deal with the voltage fluctuation in
the condition of the heavy penetration of DG. One solution is 1) Voltage Droop Equation: Considering one DG is
to adopt an active voltage control by regulating continuously connected within a distribution feeder as Fig. 1 shows:
DG’s power generation [2]. This methodology can be realized
coordinately or locally. The coordinated manner is centralized
and needs to acquire the network state with the help of some
communication improvement and the implementation of a
state estimator [6, 7]. The local manner is decentralized and
doesn’t need information from the other nodes [3, 5]. Many
publications have studied the voltage control methodology in
these two approaches [1]-[8]. Some of these publications have
proposed for the distribution systems a methodology of
hierarchical voltage control which is wildly used for
transmission systems [8]. The coordination of regulation in
substation and DG in the feeder is also of interest to the
Figure 1. One single feeder connected with DG
operators [9].
where ܲ݅ ൅ ݆ ή ܳ݅ is the power injection of DG at the bus of '
connection, ܲܿ ൅ ݆ ή ܳܿ is the total load consumption in the ΔV = S ⋅ ΔQ (5)
vq
downstream buses, ܴ ൅ ݆ ή ܺ the resistance and reactance of
the feeder. Hence the active power is decoupled with the voltage
variation. This is an approximation in transmission systems
The voltage droop by the injection of DG is close to
where the ratio R/X is relatively weak. In the following part
( Pi − Pc) ⋅ R + (Qi − Qc) ⋅ X of this section, only the non-decoupled model is used for the
ΔV = (1) optimization process since it is more appropriate for our case.
(V1 + V2 ) / 2
B. Formulization of Optimization
For a distribution feeder, the ratio R/X is much greater The optimization of reactive power setpoint can be
than that of transmission network so the part with active
formulized as minimizing an objective function subject to
power in (1) is not negligible. If we consider a 150 mm² MV
some system constraints:
cable which parameters are R=0.26 Ω/km and X = 0.11 Ω/km,
the voltage increase due to 1MW injection can only be
compensated by reactive power absorption if the DG power
­°min f ( Pi , Qi ; V j , δ j )
factor is tanϕ=2.4, which will not be possible at full active ® g(P , Q ;V , δ ) ≤ 0
power injection. °̄ i i j j (6)

2) Non-decoupled Model: The first-order derivative of i = 1, , Ng; j = 1, , Nc

power flow equations: where Pi and Qi are the power generation or consumption of
Ng, the set of controllable buses; Vj and įj are voltage profile
ª ∂P ∂P º of Nc, the set where the voltage profile is monitored.
ª Δ P º « ∂δ ∂V » ª Δ δ º
«¬ ΔQ »¼ = « ∂Q ∂Q » ⋅ «¬ ΔV »¼ (2) The set Ng can be comprised of all the loads where DG is
connected or some loads with thermal consumption and
«¬ ∂δ ∂V »¼ electric vehicle charging. In this paper, only the reactive
can be used to model the coupling between the power and power regulation of DG is considered. The set Nc consists of
voltage through a sensitivity matrix. By taking some the buses equipped with voltage sensor which is capable to
transformation of this matrix, the coupling of V-P and V-Q communicate with centralized controller. Considering the
can be obtained as follows: numerous buses in a typical distribution network, it is
impossible to monitor all voltage profile. Hence, the voltage
control strategy is to regulate the voltage at some pilot buses,
ΔV = S ⋅ ΔP + S ⋅ ΔQ (3) which might be comprised of the substation bus, DG-
vp vq
connected buses or buses with heavy loads. In our case, only
the voltage of the substation bus will be regulated since its
where S and S are respectively sensitivity matrices of voltage level has an influence on all feeders. The voltage at
vp vq
substation is conventionally regulated by OLTC with a voltage
coupling P-V and Q-V with this non-decoupled model.
regulator (VR). In the case without DG connection, the VR
Hence, the variation of the voltage magnitude at other buses
setpoint is close to the upper limit so as to minimize the line
affected by the power generation variation at the bus of DG losses. This setpoint has to be adjusted to respect the voltage
can be obtained. The details of these matrices are presented in constraints of some buses when DG is connected.
the appendix.
Therefore, the objective function of voltage control is to
3) Decoupled Model: The relationship of power and minimize the voltage variation of substation bus, which can be
voltage is usually decoupled as two coupling P-ș and Q-V in written as
most power system analysis. So (2) can be rewritten as this
form: 2
f (Q ) = (V − V − S ⋅ ΔQ ) (7)
g set m vq, g g
ª ∂P º where Vset is the setpoint of VR with OLTC, Vm is the
ª ΔP º « ∂δ 0 » ª Δδ º measurement in real time, and ΔQg the variation of the
«¬ ΔQ »¼ = « ⋅
∂Q » «¬ ΔV »¼
(4)
setpoint of reactive power of DG. The sensitivity matrix Svq,g,
«¬ 0 ∂V »¼ which is the sub-matrix of Svq at the column of generators,
Hence the variation of voltage becomes represents the coupling Q-V between the reactive power of
DG and the voltage at substation busbar presented in the
subsection II-A.
 The system constraints consist of the voltage limits and III. SIMULATION AND DISSCUSION
reactive power range of DG: The proposed method has been verified on a simplified
distribution network in Fig. 3. The MV busbar feeds three
V ≤V + S ⋅ ΔQ ≤ V (8)
min m vq, g g max feeders. The first feeder has both load and DG injection; the
second is a pure load feeder without DG; the last one feed
V ≤V ≤V (9) only a large DG farm. All DGs are windmills. The length for
min g max
each line is 5km in feeder 1, 2 and 10km in feeder 3, with a
Q ≤ Q + ΔQ ≤ Q (10) line impedance 0.26+j0.11Ÿ/km. The total active power of
min g g max DG is up to 12MW. The active and reactive consumption of
By using this sensitivity matrix, the objective function is each load is 3.0MW and 1.2MVar. The HV/MV transformer
quadratic and the constraints are linear. Hence, this voltage level is 63kV/20kV and the maximal capacity is
optimization problem can be solved using quadratic 36MVA as a typical French substation of distribution systems.
programming algorithms. The active set method is very The OLTC on the transformer is able to regulate the voltage
efficient to solve the quadratic programming and it is used in level from -10% to 10% in 17 steps, where one step is equal to
our optimization. As this algorithm works in real time, it is 1.25%. The substation is fed by an HV network through two
unacceptable if there is too much computation. The feature of 40km transmission lines, with a line impedance
this algorithm is of interest especially when the size of 0.15+j0.4Ÿ/km.
distribution network is huge.
The process of voltage control plays on both the reactive
regulation of DG and the switches of OLTC. The OLTC is
switched by a voltage regulator (VR). The switching of OLTC
is decided by the set value Vset, the deadband DB and the
delay time Td of VR. In the coordination, the controller will
firstly make use of the reactive support of DG and block the
function of OLTC when Vm is deviated from Vset. If Vm is
still not in the range of deadband during a period Td after the
reactive adjustment, then OLTC will be switched. The reactive
power of DG will be also adjusted to the new tap position of
OLTC. This process is illustrated in Fig. 2.

Figure 3. Test network

Figure 4. Wind generation and load profiles

Figure 2. Process of voltage control
 The 24-hour simulation has been performed considering a busbar, then the voltage at bus 7 will be under the lower limit.
daily load and wind generation profile in Fig. 4 extracted from On the other hand, the operation of OLTC in order to raise the
the website of RTE who is the French transmission system voltage level on the feeder 2 will introduce the overvoltage on
operator. The DG’s reactive ranges that the operator can feeder 1 and 3 in case of heavy DG injection. The results of
regulate are from -0.35pu to 0.4pu (the negative value means active control show that the voltage problem on one feeder can
absorption). The presented active voltage control methodology be regulated by the DG in this feeder and the influence on
has been compared with a conventional OLTC control which other feeders can be limited.
is used in actual distribution systems. The reference of OLTC
in the two cases is always at a fixed level, but it is possible to Regarding the voltage violations on the feeder without DG
run an OPF algorithm to set the reference hourly in order to injection, the voltage control can be realized not only by the
minimize active power losses. The bank of ShC has not been OLTC regulation but also by an active mean, for example the
considered since they are normally operated in a manual direct load control. The electric vehicle charging can be
manner and not very frequently (once a day when changing applied in such ancillary services as well as the heating system
from peak to off-peak demand). or air conditioner. This issue will be the future work but not
presented in this paper.
From Fig. 3, it can be predicted that the overvoltage
problem might happen on feeder 1 and 3 by DG in heavy The maximal and minimal voltage levels during a daily
generation condition, while the low-voltage problem might period of the two methods are shown in the table 1.
happen at the end of feeder 2 in heavy load condition without TABLE I. MAXIMAL AND M ININAL VOLTAGE
DG. The two kinds of voltage violation problem could occur
simultaneously given that DG’s power generation is
Active control Conventional OLTC
intermittent and unpredictable. This peak condition is taken
Vmax 1.05pu 1.064pu
into consideration in order to compare the performance of
active voltage control and conventional OLTC control. Vmin 0.942pu 0.944pu
The voltage profiles at some buses are shown in Fig. 5.
The voltage limit in the constraints is from 0.95pu to
1.05pu. The maximal voltage level in active voltage control is
lower than that in conventional OLTC control. Moreover, the
minimal voltage level in active control is higher than that in
conventional OLTC control. Both the upper and lower limits
are violated in the conventional OLTC control. The
performance of active voltage control is better on this aspect
of voltage limit. However, the minimal voltage of active
voltage control is still under the lower limit because there is no
DG in feeder 2 available to increase the voltage level.
The regulations of OLTC and reactive power of DG are
shown in Fig. 6 and Fig.7 respectively.

Figure 5. Voltage profile at buses 1,7 and 8

From Fig. 5, the DG’s power generation raises the voltage

level at the buses of feeder 1 and 3. Meanwhile, the heavy
load condition on feeder 2 leads the voltage level at the end of Figure 6. The step changement of OLTC
feeder (bus 7) under the lower limit of the voltage constraint.
If OLTC’s tap is switched to lower the voltage at substation
 Figure 8. Active power losses of network

The active power losses in the network of these two kinds

of voltage control are presented in Fig. 8. The losses of active
voltage control are lower than of an OLTC control in this case.
This is because the OLTC set voltage can be higher in active
voltage control without violation of voltage limit. Considering
that most buses in the feeders are PQ type, their power losses
will decrease if the voltage profile is great. The conventional
OLTC control, performed in this condition with the presence
of DG, has to reserve more voltage margin against the voltage
rise due to DG. Hence, the set voltage of VR is lower and the
power losses increase. Moreover, the voltage level at the end
Figure 7. Reactive power of DG (Qg>0 means injection, of feeder is also raised by DG’s reactive injection and then the
Qg<0 means absorption) power losses can be reduced.
With regards to the OLTC operation, Fig.6 shows that the
active control needs less operation of OLTC during one day.
This is due to the defined objective function of optimization.
Since the voltage variation at substation is minimized, it is
reasonable that the operation of OLTC will be lowered. As
known, the frequent switching of OLTC will wear the devices
and thus shorten its life cycle. Reducing the switching number
means less maintaining cost of OLTC for the operator, which
can be a benefit of this active control.
In Fig. 7, it can be observed that the DG in the feeder has a
unity power factor according to the contract in which the
aggregator usually requires to the owner of DG. In the case of
active voltage control, the DG provides the reactive support
and the setpoint of reactive power is optimized by the
controller that makes the decision based on the network state.
During the peak condition of load and wind generation
(typically from 12h to 22h), DG3 is forced to absorb reactive Figure 9. Reactive power import at the substation
power in order to mitigate the voltage rise due to its active
power generation. For the other situations, the DGs are able to In French power grids, the reactive power is possibly
inject reactive power in the network to support the voltage offered from transmission system to substation (for a
level at substation. substation with 63kV/20kV transformer, it allows power
Besides limiting the voltage fluctuation, the performance factor in the range : − 0.4 ≤ tanϕ ≤ 0.4 ). The inductive power
of these voltage control methodologies on the network is also factor of DG will raise the reactive import. Considering that it
evaluated from two aspects: the active power loss in the is difficult to transmit much reactive power from upstream bus,
feeders and the reactive power import at the substation. it is desired that reactive power is compensated by local
devices. Fig. 9 shows that the reactive power import at the
substation with the active voltage control can be lowered due −1
to the reactive generation of DG. This result would not be ∂P ª ∂Q º
and S =M⋅ ⋅
always true since DG can absorb as well reactive power vq ∂δ «¬ ∂δ »¼ .
depending on its voltage setpoint. However, this voltage
regulation renders distribution systems to be more flexible by
providing a possibility to respond the reactive power demand REFERENCES
from transmission grid.
[1] G. Strabac, N. Jenkins, M. Hird, P. Djapic,G. Nicholson, "Integration
By the comparison of the two methodologies, the active of operation of embedded generation and distribution networks,"
voltage control can provide promising performances University of Manchester Institute of Science and Technology (UMIST),
2002.
especially in a peak scenario. Moreover, the active voltage
[2] T. Xu, P. C. Taylor, "Voltage control techniques for electrical
control applies less number of switches to reduce the wear of distribution networks including distributed generation, " in Proc. of the
OLTC. 17th World Congress, IFAC, Seoul, Korea, July, 2008.
[3] P. N. Vovos, A. E. Kiprakis, A. R. Wallace,s, G. P. Harrison,
IV. CONCLUSION "Centralized and distributed voltage control: impact on distributed
This paper investigates an active voltage control generation penetration, " IEEE trans. Power Systems, vol 22, pp 476-
483, 2007.
methodology that could be applied in distribution networks [4] A. Keane, M. O’Malley, "Optimal allocation of embedded generation
with the presence of massive DG injection. A coordinated on distribution networks," IEEE trans. Power Systems, vol 20, No. 3,
approach is used among the regulation of DG and OLTC. In August 2005.
this coordination of voltage regulation, the reactive power of [5] T. Sansawatt, L. F. Ochoa, G. P. Harrison, "Integrating distributed
DG is controlled by a process of optimization based on the generation using decentralized voltage regulation," unpublished.
Presented at IEEE PES General Meeting, July 2010.
network state at pilot node in order to avoid excessive [6] F. Bignucolo, R. Caldon, V. Prandoni, "Radial MV networks voltage
operations of OLTC. The strategy of voltage control has been regulation with distribution management system coordinated
verified in a daily simulation and compared with the controller," Electric Power Systems Research, 78, pp634-645, 2008.
conventional OLTC control. The results show its promising [7] A.G. Madureira, J.A.P. Lopes, “Coordinated voltage support in
performances especially in a peak condition. distribution networks with distributed generation and microgrids,” IET
Renewable Power Generation, Vol. 3, Iss. 4, pp. 439–454, 2009.
V. APPENDIX [8] O. Richardot, A. Viciu, Y. Besanger, N. Hadjsaid, C. Kieny,
"Coordinated voltage control in distribution networks using distributed
The Sensitivity Matrix Calculation generation," Transmission and Distribution Conference and Exhibition,
The matrix form of (2) can be written as two equations, 2005/2006 IEEE PES , vol., no., pp.1196-1201, 21-24 May 2006.
[9] D. Cortinas, P. Justin, "Methods and tools developed by EDF for the
∂P ∂P
⋅ Δδ +
connection of distributed generators," Distributech Europe 99, October,
ΔP = ⋅ ΔV (A.1) 1999.
∂δ ∂V [10] L. F. Ochoa, A. Keane, G. P. Harrison, "Minimizing the Reactive
∂Q ∂Q Support for Distributed Generation: Enhanced Passive Operation and
ΔQ = ⋅ Δδ + ⋅ ΔV (A.2) Smart Distribution Networks," IEEE trans. Power Systems, vol 26, No.
∂δ ∂V 4, November 2011.
(A.2) can be transformed to the following form,
−1
Δδ =
ª ∂Q º §
⋅ ¨ ΔQ −
∂Q ·
⋅ ΔV ¸ (A.3)
«¬ ∂δ »¼ © ∂V ¹
Then inject (A.2) into (A.3):
∂P ª ∂Q º − 1 § ∂Q · ∂P ⋅ ΔV (A.4)
ΔP = ⋅ « » ⋅ ¨ ΔQ − ⋅ ΔV ¸ +
∂δ ¬ ∂δ ¼ © ∂V ¹ ∂V
Thus it results:

ΔV = S ⋅ ΔP + S ⋅ ΔQ (A.5)
vp vq
−1
ª ∂P ∂P ª ∂Q º − 1 ∂Q º
where S = « − ⋅ ⋅ »
vp « ∂V ∂δ «¬ ∂δ »¼ ∂V »
¬ ¼