See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/327435713

Optimal oversizing of utility-owned renewable DG inverter for voltage rise

prevention in MV distribution systems

Article  in  International Journal of Electrical Power & Energy Systems · February 2019

DOI: 10.1016/j.ijepes.2018.08.040

CITATIONS READS

9 303

3 authors:

Abdelfatah Ali David Raisz

South Valley University RWTH Aachen University
32 PUBLICATIONS   125 CITATIONS    66 PUBLICATIONS   344 CITATIONS   

SEE PROFILE SEE PROFILE

Karar Mahmoud
Aalto University
60 PUBLICATIONS   507 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Fault localization and management in MV networks View project

EU H-2020 RESERVE project View project

All content following this page was uploaded by Abdelfatah Ali on 20 April 2019.

The user has requested enhancement of the downloaded file.

 Electrical Power and Energy Systems 105 (2019) 500–513

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

journal homepage: www.elsevier.com/locate/ijepes

Optimal oversizing of utility-owned renewable DG inverter for voltage rise T

prevention in MV distribution systems
⁎
Abdelfatah Alia,b, , David Raiszb, Karar Mahmoudc
a
Department of Electrical Engineering, South Valley University, 83523 Qena, Egypt
b
Department of Electric Power Engineering, Budapest University of Technology and Economics, 1111 Budapest, Hungary
c
Department of Electrical Engineering, Aswan University, 81542 Aswan, Egypt

A R T I C LE I N FO A B S T R A C T

Keywords: The penetration of renewable distributed generations (DG) is recently being increased in distribution systems.
Distribution systems Due to the intermittent nature of these generation units, several technical problems are expected based on the
Distributed generation environmental conditions and load profiles, such as voltage rise and increasing losses. A traditional way to
Optimal inverter size alleviate these technical problems is to oversize their interfacing inverters for releasing their capacities to inject/
Voltage regulation
absorb further reactive power. In the literature, the DG inverter is normally oversized by a certain percentage
Active power curtailment
Reactive power capability
(e.g., 10%) for this purpose. In turn, an optimization-based method is proposed in this paper to calculate the
optimal oversize of the interfaced inverter employed in various utility-owned DG types to regulate voltages and
to reduce losses with minimum total costs. The proposed method considers the active power curtailment (APC)
feature in the DG inverter and the transformer taps. Different control schemes of the interfaced inverter are
considered and incorporated in the proposed optimization model. The simulation results demonstrate the ef-
fectiveness of the proposed method compared with existing methods.

1. Introduction These challenges have to be considered when deciding the DG pene-

As the annual electric energy demand grows, the penetration of
distributed generations (DG) in distribution systems has noticeably in-
creased throughout the world to fulfill the massive demand for elec-
tricity. The integration of DG units with an existing distribution system
has many benefits to several entities (e.g., owner, utility, and final
user).
The optimal allocation of the DG has been deeply studied in the
literature. In [1,2], the optimal size and location of DG are determined
to minimize the system losses and maximize the system voltage stabi-
lity. The simultaneous optimal placement of DG and optimal allocation
of tie-switch based on maximization of system load-ability is addressed
in [3]. The allocation of multiple DG units to minimize the system
losses, increase load-ability, and improve system voltage stability has
been addressed in [4–6].
High penetration of utility-owned DG into distribution system has
positive and negative effects on the normal operation of the system.
Improved voltage profile, high reliability of power supply to the con-
sumers, power loss reduction, support voltage stability and support of
ancillary services are major positive impacts [6–10], whereas negative
ones include several technical, economic and regulatory challenges.
tration level. Voltage rise is one of the main concerns that can limit the
allowed penetration of DG [11,12]. During high power generation of
DG and light load periods (extreme state), there is a high possibility of
reverse power flow, and therefore voltage rise, in the distribution feeder
[13–15]. The conventional voltage regulation devices such as on-load
tap changers (OLTCs) are not capable of treating these issues com-
pletely without a proper coordination with DG. Therefore, a supervisory
control and data acquisition-based two-stage voltage control scheme for
the coordination of the capacitor banks (CBs), OLTC transformer, and
DG has been presented in [16].
Several forms of utility-owned DG such as photovoltaic (PV), wind
turbine generation system (WTGS), microturbines, and fuel-cells are
interfaced to the medium voltage (MV) distribution system through
inverters. These power electronics devices are superior to the conven-
tional interface machines due to their flexibility in operation and con-
trol. Inverters of DG have the capability to provide reactive power
support to MV distribution systems locally at the point of common
connection (PCC). The reactive power capability of DG inverters is an
efficient way to compensate the fluctuating voltage, caused by the
fluctuating DG output (active) power [17–22]. Furthermore, the smart
voltage source inverters (VSI) of DG with reactive power compensation

⁎
Corresponding author at: Department of Electric Power Engineering, Budapest University of Technology and Economics, 1111 Budapest, Hungary.
E-mail address: abdelfatah.mohamed@vet.bme.hu (A. Ali).

https://doi.org/10.1016/j.ijepes.2018.08.040
Received 19 December 2017; Received in revised form 18 June 2018; Accepted 23 August 2018
0142-0615/ © 2018 Elsevier Ltd. All rights reserved.
A. Ali et al. Electrical Power and Energy Systems 105 (2019) 500–513

max
Nomenclature Pinv maximum active power of the inverter (MW)
max
Qinv maximum reactive power of the inverter (Mvar)
Abbreviations PPV dc power output from PV farm (MW)
Tij period between hours i and j (h)
DG distributed generation Tkm period between hours k and m (h)
APC active power curtailment Ts time duration of the worst-case period (h)
OLTC on-load tap-changer Tp time duration of the planning period (h)
CB capacitor banks VN nominal voltage (pu)
PV photovoltaic Vmax upper limit of the voltage (pu)
WTGS wind turbine generation system Vmin lower limit of the voltage (pu)
MV medium voltage VPCC voltage magnitude at PCC (pu)
PCC point of common coupling Vi,g voltage magnitude at bus i during state g
VSI voltage source inverter Srated rated size of the inverter (kVA)
MPPT maximum power point tracking Sos,k,g inverter oversize at bus k during state g (kVA)
KKT Karush–Kuhn–Tucker Sosmax maximum inverter oversize (kVA)
pdf probability distribution function Sos,k,o inverter oversize without considering OLTC (kVA)
Sos,k,N inverter oversize with considering OLTC (kVA)
Symbols Pcurt APC (MW)
Pcurt,k,g APC at bus k during state g (MW)
R solar irradiance (kW/m2) min
Pcurt , k, g minimum APC at bus k during state g (MW)
α, β parameters of Beta pdf max
Pcurt , k, g maximum APC at bus k during state g (MW)
fb (R) beta distribution function of R PMPPT maximum power point tracking power (MW)
probR (Gs ) probability of the solar irradiance being in state s Ctotal total cost ($)
Rs1, Rs2 solar irradiance limits of state s Cg1 cost of energy losses in the distribution system during state
PPVs output power of the PV module during state s g ($)
TCells cell temperature during state s (°C) Cg2 cost of energy losses in the inverter during state g ($)
TA ambient temperature (°C) Cg3 cost of inverter oversize during state g ($)
Kv voltage temperature coefficient (V/°C) Cg4 cost of curtailing active power during state g ($)
NOT nominal operating temperature of cell (°C) as, bs, cs cost coefficients of inverter oversize
Isc short circuit current (A) ap, bp, cp cost coefficients of active power curtailment
Voc open circuit voltage (V) ainv, inverter’s stand by losses
IMPP current at maximum power point (A) binv voltage dependent losses
VMPP voltage at maximum power point (V) cinv ohmic losses
Ravs average solar irradiance of state s Ctotal,k,o total costs without considering OLTC ($)
Icells cell current (A) Ctotal,k,N total costs with considering OLTC ($)
Vcells cell voltage (V) CL cost rate of energy losses in the distribution system
fw (v ) weibull distribution function of v ($/kWh)
v wind speed (m/s) Cinv,k cost rate of energy losses in kth inverter ($/kWh)
fr (v ) Rayleigh pdf ∅dg set of inverters-based DG locations
probWT (Gw ) probability of the wind speed being in state w ∅b set of buses of the system
vw1, vw2 wind speed limits of state w ∅r set of branches of the system
PWTw output power of the wind turbine during state w (MW) PLds,g active power losses of a distribution system during state g
vci cut-in speed (m/s) (MW)
vr rated speed (m/s) PLinv,g active power losses of Kth inverter during state g (MW)
vco cut off speed (m/s) tapij,g tap position of OLTC in the line i-j during state g
vavw average wind speed of state w tapmin minimum tap position of OLTC
Prated rated power (MW) tapmax maximum tap position of OLTC
Ψ complete annual generation-load model Gij conductance of the line i-j (Ω−1)
C a matrix that includes all possible combinations of the PV/ Bij susceptance of the line i-j (Ω−1)
WTGS output power state and the load states x,y binary variables
prob(Cg) one-column matrix that represents the probability corre- PDG,i rated power of the PV/WTGS connected at bus i (MW)
sponding to matrix C Pd,i active power demand at bus i (MW)
Ns total number of states in modelΨ Qd,i reactive power demand at bus i (Mvar)
Pgrid grid active power (MW) NB number of buses
Qgrid grid reactive power (Mvar) CRk cost reduction at DG location k (%)
Pinv active power output from the inverter (MW) OSRk oversize reduction at DG location k (%)
Qinv reactive power output from the inverter (Mvar)

capability can mitigate the voltage rise problem [23]. For instance, at
the occasions of high PV generation, i.e., sunny moments, and low
consumption (extreme state), the spare capacity of the interfaced in-
verter can be employed to absorb reactive power, thereby regulating
the voltage at PCC. However, during such extreme state, the DG in-
verter is often fully loaded. For the purpose of allowing further reactive
power capability, many studies suggested to oversize the interfaced
inverter. For example, in [19,24], it is established that 10% inverter
oversize provides a sufficient release of the inverter capacity for loss
reduction and voltage improvement, while 60% is considered in [25].
Oversizing the inverter of DG with a certain level (e.g., 10% or 60%) for
preventing voltage rise is not an optimal approach. Oversized inverters

501
A. Ali et al. Electrical Power and Energy Systems 105 (2019) 500–513

have the capability to deliver reactive power without curtailing the real PV Farm
power of DG. Note that the cost of a DG project is expected to increase
in the case that the interfaced inverter is decided to be oversized. Thus,
PCC PPV
an optimal oversize of the interfaced inverter of DG must be determined OLTC Pgri Pinv
for preventing voltage rise while considering the oversize costs.
Another alternative way of regulating the overvoltage during the Grid d

Qgri Qinv
extreme state by the interfaced inverter is active power curtailment
d DC/AC
(APC) of DG [26–28], especially for distribution systems with high R/X AC inverter
ratios [29]. This latter control-based way seems to be cost-efficient as it Active power flow load
Reactive power flow
does not require oversizing DG inverter capacity. However, by using
this way, the DG units are forced not to operate at full capacity (i.e., Fig. 1. Schematic diagram of MV distribution system connected with DG.
waste energy) for voltage regulation during the extreme state.
The cost of inverter oversizing and the cost of curtained power must 2. Problem description
be considered when deciding the inverter oversize level. Accordingly,
three solutions are possible for the voltage rise problem by the DG in- 2.1. Voltage regulation problem with DG
verter: (1) Oversizing DG inverter without APC, (2) Enabling the APC
feature without oversizing the DG inverter, (3) Oversizing DG inverter The impact of DG units on voltage regulation is considered a serious
and enabling APC. The best solution is determined according to the cost
drawback and limitation for high DG penetration. Fig. 1 shows a
rates of inverter oversizing and curtained power. This optimal oversize schematic diagram of grid-connected inverter-based DG unit (PV as an
of the interfaced inverter for mitigating voltage rise is expected to vary
example). A local load is connected to PCC, which is placed at DG unit
according to different factors, including DG penetration, DG type, and terminals. To draw the maximum efficiency of DG unit, it tends to
DG location. Another significant factor is the presence of voltage con-
operate in the maximum power point tracking (MPPT) mode. In other
trol devices (e.g., transformer taps) which are required to be considered words, the output power of DG depends on the energy source conditions
as control variables in the optimization model for inverter oversizing.
rather than the power balance between supply and loads in the feeder.
The voltage regulation problem with DG can be effectively and Therefore, the required amount of the power should be complemented
economically treated if both reactive power capability and APC options
from the main grid.
are considered when oversizing the interfaced inverter. In this paper,
When the output power of DG exceeds the local load, reverse power
the optimal oversize of the interfaced inverter in different utility-owned
flow can occur due to this excess generation of DG. This reverse flow of
DG types is calculated for preventing voltage rise. Inverter oversizing
power can lead to voltage rise because the traditional distribution
cost, APC cost, and cost of energy losses are considered in the objective
systems are designed for one-way power flow. Due to the high pene-
function to be minimized. The reason for incorporating the cost of en-
tration of DG units in distribution systems, voltage rise can violate the
ergy losses is that it is demonstrated that allowing the DG inverter to
acceptable thresholds which lead to the negative effects of equipment.
inject/absorb reactive powers can increase/decrease system losses. To
Hence, the voltage rise problem is one of the foremost concerns for
the best of the authors’ knowledge, this is the first work to propose an distribution systems with high DG penetration. Fig. 2 illustrates the
optimization model for determining the optimal DG inverter oversize to
effect of high DG penetration on the voltage profile at PCC. As shown in
regulate voltage in distribution systems. The proposed optimization Fig. 2(a), the output power of DG exceeds the local load during the
model determines the most economical solution for voltage regulation
period Tij. During this period, the power difference between DG output
by considering inverter oversize and/or APC options. The tap position and load will flow back to the main grid. Therefore, the voltage at PCC
of the OLTC is also incorporated in the optimization model as a control
will rise over the nominal voltage (VN). As the generation of DG con-
variable to be optimized. To demonstrate the effectiveness of the pro- tinues to increase, voltage rise at PCC violates the upper limit (Vmax)
posed method for calculating the inverter oversizing to prevent voltage
during the period Ts (Ts = Tkm), as shown in Fig. 2(b). The period (Ts) in
rise, it is compared with several methods in the literature which the voltage violates the upper limit is the worst-case period of
[19,24,25,27,30]. The optimal inverter oversizing is comprehensively
this day.
analyzed for different DG types, penetrations, and locations. The pro-
posed method can compute the optimal inverter oversize for avoiding
the technical problems in MV distribution systems at high DG pene-
PV
tration. The main contributions of the paper are summarized as follows:
Load demand and

generation
PV generation

Local load
An optimization model is proposed to determine the optimal DG
inverter oversize for voltage regulation in the distribution system.
Different DG types are incorporated in the proposed optimization- Ts
based inverter oversizing model.
The tap position of the OLTC is considered in the optimization 0 Ti Tk Tm Tj 24
model as a control variable to be optimized. Time(h)
Comprehensive simulations are carried out with different DG types, (a)
penetrations, and locations.
Vmax
Voltage (pu)

The proposed method for the DG inverter oversizing is an efficient

tool that can help system operators and planners to decide the inverter VN
size of DG to prevent voltage problems. Vmin
The remaining part of the paper is organized as follows. Section 2
describes the voltage regulation problem with DG and its mitigation 0 Ti Tk Tm Tj 24
Time(h)
techniques. Modeling of renewable energy sources and load Data is (b)
described in Section 3. The optimization model for calculating the
oversize of the DG inverter is given in Section 4. Sections 5 and 6 de- Fig. 2. Daily profiles of DG output, local load, and voltage at PCC. (a) Local
scribe the results and conclusions, respectively. load and PV generation profile. (b) Voltage profile.

502
A. Ali et al. Electrical Power and Energy Systems 105 (2019) 500–513

2.2. Mitigation techniques of voltage rise using DG inverter μ × (1 + μ) ⎞

β = (1−μ) × ⎜⎛ −1
⎝ σ2 ⎠ (2)
The voltage rise seen at PCC can be alleviated by controlling active
and/or reactive power injections from the DG inverters. Besides miti- μ×β
α=
gating voltage rise, inverters can also alleviate under-voltage by re- 1−μ (3)
active power injection. The smart VSI has a reactive power capability in
To incorporate the output power of the PV system as a multi-state
which they can inject/absorb reactive power. Fig. 3 shows the P-Q
variable in the planning formulation, the pdf is divided into several
capability curve of an inverter [25], where the semicircle represents the
states. In each state, the solar irradiance is within specified limits. The
limit of the inverter’s operating range. Note that the radius of this
number of the states should be carefully selected because a small
semicircle denotes the rated size of the inverter (Srated), Pinv is the active
max number of states affects the accuracy, whereas a large number increases
power output of the inverter, and Qinv is the maximum reactive power
the problem complexity. In this paper, the number of states is chosen to
of the inverter. At high power generation of DG, the DG inverters are
max be 12 states in which the step is adjusted to be 100 W/m2 [32,33].
often fully loaded (Pinv = Pinv ) (i.e., work at unity power factor), where
max The solar irradiance probability of each state can be determined as
Pinv is the maximum active power that can be injected by DG, so there
follows:
is no spare capacity to inject/absorb reactive power to alleviate the
Rs2
voltage rise problem. If the inverter is not oversized, it will not be ap-
plicable to absorb reactive power during voltage rise period, and so probR (Gs ) = ∫ fb (R). dR
Rs1 (4)
voltage rise problem cannot be fixed.
The output power of the PV module for each state can be calculated
by using the average value of the solar irradiance at this state as fol-
2.3. Proposed optimization model of inverter oversizing
lows:
To regulate the voltages and reduce the power losses with minimum VMPP × IMPP
PPVs = N × × Vcells × Icells
total costs, the inverters oversizing (Sos) and APC (Pcurt) should be op- Voc × Isc (5)
timally and simultaneously computed. Moreover, the tap position of in which
OLTC must be computed optimally. In this paper, the optimal oversize
of the utility-owned DG inverter is calculated considering APC option to NOT −20
TCells = TA + Ravs ⎛ ⎞
regulate the voltages and reduce the power losses. Also, the effect of ⎝ 0.8 ⎠ (6)
optimal tap position of OLTC on the inverter oversize and APC is stu-
Icells = Ravs (Isc + Ki (TCells−25)) (7)
died. Based on the reactive power capability and APC of the inverters,
the following control schemes are considered in this paper, as illu- Vcells = Voc−Kv TCells (8)
strated in Fig. 4 [25].

Control scheme A: Only the option of APC is available. The DG 3.2. Wind speed modeling
inverter has no capability to inject/absorb reactive power, i.e., the
inverter operates at unity power factor (IEEE 1547 standard [31]). To model the behavior of wind speed, Weibull pdf is often re-
Control scheme B: The inverter has the capability to inject/absorb commended [32,34]. The Weibull pdf which relates the wind speed and
reactive power, and there is no option for APC. The size of DG in- the output power of WTGS is given by:
verter could be optimally oversized by Sos for further reactive power (k − 1) k
k v ⎡ v
capability. fw (v ) = ⎛ ⎞ ⎛ ⎞
⎟ exp ⎢−⎛ ⎞ ⎤
⎝ c ⎠⎝ c ⎠ c ⎥
Control scheme C: Both options of control schemes A and B are ⎣ ⎝ ⎠ ⎦ (9)
available. where k and c are the shape index and scale index, respectively. If the
shape index is 2, the pdf is called a Rayleigh pdf. Hence, Rayleigh pdf is
3. Modeling of renewable energy sources and load a special case of Weibull pdf which can mimic most wind speed profiles.
In this work, Rayleigh pdf is utilized to model the speed for each state as
In this section, the probability of the generated power from the follows:
renewable energy sources (RESs) is considered for optimally oversizing 2
2v ⎡ v
their inverters. Two types of RESs (PVand WTGS) are considered in this fr (v ) = ⎛ 2 ⎞ exp ⎢−⎛ ⎞ ⎤
⎝c ⎠ c ⎥
study. A study period of one year is selected to represent the random ⎣ ⎝ ⎠⎦ (10)
behavior of the different RESs. Beta and Weibull probability density As in the case of PV, the pdf for wind speed modeling has been
functions (pdf) are used to model hourly solar irradiance and wind divided into 12 states in which the step is adjusted to be 1 m/s. The
speed data, respectively. The load model is generated by utilizing the
hourly load shape of the IEEE-RTS system. P

3.1. Solar irradiance modeling max

Pinv (Rated Power)

To describe the random phenomenon of the solar irradiance, a beta Pinv

pdf is used which is formulated as follows:

Γ(α + β )
⎧ Γ(α )Γ(β ) × R(α − 1) × (1−R)(β − 1) , 0 ≤ R ≤ 1, α, β ≥ 0 Srated
fb (R) =
⎨ 0, otherwise -Q Q
⎩ (1)
max
Qinv Pinv tanș max
Qinv Pinv tan ș
The parameters of the beta pdf (α, β) can be calculated by utilizing
the mean (µ) and standard deviation (σ) of the random variable R as
follows: Fig. 3. Inverter P-Q capability curve.

503
A. Ali et al. Electrical Power and Energy Systems 105 (2019) 500–513

Control scheme A ⎧ 0, 0 ⩽ vavw < vci

P
Control scheme B ⎪ Prated × (vavw − vci )
, vci ⩽ vavw < vr
PWTw = (vr − vci )
Control scheme C ⎨ Prated, vr ⩽ vavw < vco
max
Pinv
⎪
(Rated Power) ⎩ 0, vco ⩽ vavw (12)
Srated+Sos
Pcurt
3.3. Load modeling
Pinv Srated
The load profile will be assumed to follow the IEEE-RTS [35]. Based
on this assumption, a clustering technique has been used to divide the
load into 10 states utilizing the central centroid sorting process [36,37].
-Q Q Fig. 5 shows the 10 load states accompanied by their probabilities
( probLoad (L y ) ) which are used in this paper [34].
Fig. 4. Different control schemes of the inverter.
3.4. Combined generation-load model
0.2
0.18 0.163 0.163 0.165 0.165 The modeling of PV/WTGS output power and the load are used to
0.16 generate the combined PV-load/WTGS-load model. In this study, solar
probability of the load

0.14 irradiance states, wind speed states, and load states are assumed to be
0.12
independent. Based on this assumption, the combined probability of the
0.106
0.091 load and solar irradiance/wind speed (prob{Cg}) can be computed by
0.1
convolving the two probabilities, as described in the following:
0.08
0.056
0.06 0.047
(a) Combined solar irradiance-load model
0.04 0.033

0.010 prob {Cg } = probR (G ) × probLoad (L) (13)

0.02
0
35.1 40.6 45.1 51 58.5 65 71.3 77.4 85.3 100 (b) Combined wind speed-load model
% of peak load
prob {Cg } = probWT (G ) × probLoad (L) (14)
Fig. 5. The yearly probability of the load.
Based on this concept, a generation–load model is computed by
Stage I: Data Collection listing all possible combinations of the PV/WTGS output power and the
load. The complete generation–load model can be calculated as follows:
Define the type and penetration Ψ = [{Cg , prob (Cg )}: g = 1: Ns]
Read system data (15)
System Data

DG Data

of DG connected to the system

where prob(Cg) is a vector that represents the combined probability
Define the constraints of the Read the historical/long-term based on matrix C; and Ns is the total number of states in the model Ψ ,
distribution system forecast data of DG
which is equal to the product of the solar irradiance/wind speed states
and the load states (in this study, Ns = 12 × 10 = 120 states).

Stage II: Data structure 4. Problem formulation

Estimate the probabilities of DG power and load during the planning period 4.1. Modeling of Inverter-based DG

Construct the optimization model for inverter oversizing considering parameters DG is sometimes connected to a distribution grid by asynchronous
of the cost function and constraints generators or synchronous generators. Some DG types, such as fuel
cells, PV, WTGS, and microturbines are connected to distribution
Stage III: Problem solving system through electronic power devices (DC/AC or AC/AC con-
verters). In such inverter-based DG, the active power output, the re-
Solve the optimization problem (19)-(32) to get the optimal inverter oversize that
active power output, and the oversize of the DG inverter can be de-
regulates the voltages and minimizes the total costs
scribed as follows,
Stage IV: Results printing PMPPT −Pcurt VPCC ⩾ V max
Pinv = ⎧
⎩ PMPPT
⎨ VPCC < V max (16)
Print optimal inverter oversize, optimal APC, optimal tap position, total costs,
active losses, and voltages at all buses
max 2
Qinv = ± (Srated + Sosmax )2 −Pinv (17)
Fig. 6. Block diagram of the proposed method.
⎧ 0 V min ⩽ VPCC ⩽ V max
Sos = 2 2
⎨ Pinv + Qinv −Srated V min > VPCC > V max (18)
wind speed probability of each state can be given as follows: ⎩

v w2 Eq. (16) implies that if the voltage at PCC is greater than the upper
probWT (Gw ) = ∫ fr (v). dv voltage limit, the curtailed power has a real value (0 < Pcurt ≤ PMPPT),
v w1 (11) otherwise it is zero. The constraint in (17) defines the maximum re-
active power injection/absorption of the DG inverter considering the
The output power of wind turbine for each state can be calculated rated inverter capacity and the maximum inverter oversize value. The
by using the average value of the wind speed at this state as follows: optimal inverter oversize is described in (18) where Qinv is the optimal

504
A. Ali et al. Electrical Power and Energy Systems 105 (2019) 500–513

36 37 38 39 40 41 42 43 44 45 46

47 48 49 50 51 52 66 67 68 69

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
OLTC

Slack Bus 53 54 55 56 57 58 59 60 61 62 63 64 65

28 29 30 31 32 33 34 35

Fig. 7. Single line diagram of 69-bus distribution system.

Table 1 than zero even the voltage is within limits.

Probabilities of PV and WTGS powers.
PV power probabilities WTGS power probabilities 4.2. Optimization model for optimal inverter oversizing

State No % of rated Probability State No % of rated Probability In this work, the optimal inverter oversize for voltage regulation in
power power
MV distribution systems is aimed to be computed. Minimizing the total
1 0 0.052 1 0 0.206 costs (Ctotal) for all possible combinations of load and DG output power
2 9.091 0.173 2 4.998 0.066 are considered in the objective function. The proposed optimization
3 18.182 0.106 3 14.995 0.112 model is mathematically formulated as follows:
4 27.273 0.115 4 19.993 0.104
Ns
5 36.364 0.106 5 34.988 0.112
6 45.455 0.11 6 44.985 0.091 MinimizeCtotal = ∑ (Cg1 + Cg2 + xCg3 + yCg 4) × prob (Cg )
7 54.545 0.099 7 54.982 0.077 g=1 (19)
8 63.636 0.088 8 64.979 0.050
9 72.727 0.068 9 74.976 0.045 in which
10 81.818 0.056 10 84.972 0.033
Cg1 = Tp CL PLds, g (20)
11 90.909 0.025 11 94.969 0.025
12 100 0.002 12 100 0.078
Cg 2 = Tp ∑ Cinv, k PLinv, k, g
k ∈∅dg (21)
20% penetration 40% penetration 60% penetration
Cg3 = ∑ as, k + bs, k Sos, k, g + cs, k Sos2 , k, g
80% penetration 100% penetration
k ∈∅dg (22)
12
Inverter oversize (%)

10
8
6
4 (i, j ) ∈∅r
2
0 (25)
2 5 8 11 14 17 20 23 26 29 32 35 38 41 44 47 50 53 56 59 62 65 68
DG location
Fig. 8. Optimal oversize of the PV inverter at different locations of the studied
system with five different PV penetrations (69-bus distribution system).
value of output reactive power. It is important to note that Pcurt and Sos
are required to be simultaneously computed so as to prevent the voltage
rise problem with DG. Furthermore, it is worth to mention that the
expressions (16)–(18) are valid only if the inverter oversize and/or APC
are/is optimally computed to regulate the voltage without considering
any other factors to be minimized. On the other hand, if any other
factors are added to the objective function (e.g., energy losses of the
system, inverter losses), Pcurt and/or Sos may have real values greater
2
Cg 4 = ∑ ap, k + bp, k Pcurt , k, g + cp, k Pcurt , k, g
k ∈∅dg (23)
where
PLds, g = ∑ [Gij (Vi2, g tapij2, g + V j2, g−2Vi, g Vj, g tapij, g cosδij, g )]
(24)
PLinv, k, g = ainv, k + binv, k (Srated, k + Sos, k, g ) + Cinv, k (Srated, k + Sos, k, g )2
The cost functions are given in Eqs. (22) and (23) are assumed as
quadratic equations. Quadratic equations are utilized instead of linear
equations to represent the general case. The choice of the coefficients of
these quadratic equations is based on the prices of the available com-
mercial inverters and energy prices [38,39]. In [40], a price list of
different inverter sizes is given which can be used to calculate the
parameters of the cost function based on the curve fitting technique. In
this paper, we have used the parameters given in Table 6 in the
Appendix. The quadratic equation in (25) represents the total active
power losses in the inverter in which the active power loss in an in-
verter can be approximated by a quadratic function of its apparent
power. Where ainv,k, binv,k, and cinv,k are the inverter’s standby losses,
voltage-dependent losses over the power electronic components

505
A. Ali et al. Electrical Power and Energy Systems 105 (2019) 500–513

20% penetration 40% penetration 60% penetration PG,1, g + C (g , 1) PDG, i−C (g , 2) Pd, i
80% penetration 100% penetration NB

12
−Vi, g ∑ Vj,g tapij,g [Gijcos(δij,g ) + Bij sin(δij,g )] = 0 ∀ i ∈ ∅b , g
j=1
Inverter oversize (%)

10 (26)
8 NB
QG,1, g−Qinv, i, g−C (g , 2) Qd, i−Vi, g ∑ Vj,g tapij,g [Gijsin(δij,g ) + Bij cos(δij,g )]
6 j=1

4 =0 ∀ i ∈ ∅b , g (27)
2 max max
−Qinv , k, g ≤ Qinv, k, g ≤ Qinv , k, g ∀ k ∈ ∅dg , g (28)
0
2 5 8 11 14 17 20 23 26 29 32 35 38 41 44 47 50 53 56 59 62 65 68 V min ≤ Vk, g ≤ V max ∀ k ∈ ∅dg , g (29)
DG location
min max
Pcurt , k , g ≤ Pcurt , k , g ≤ Pcurt , k , g ∀ k ∈ ∅dg , g (30)
Fig. 9. Optimal oversize of the WTGS inverter at the different locations of the
studied system with five different WTGS penetrations (69-bus distribution Sosmin ≤ Sos, k, g ≤ Sosmax ∀ k ∈ ∅dg , g (31)
system).
tapmin ≤ tapij, g ≤ tapmax ∀ (i, j ) ∈ ∅r , g (32)
20% penetration 40% penetration 60% penetration
Expressions (28) and (30) give a capability curve presented in Fig. 4.
80% penetration 100% penetration It is worth to mention that different constraints can be considered in the
8000 optimization model according to the utility requirements.
Curtailed energy (MWh)

7000
6000 4.3. Solution process
5000
The block diagram of the proposed method is shown in Fig. 6. As
4000
shown in the figure, the block diagram consists of four stages. The first
3000
stage is data collection, where the algorithm collects all required data
2000
from the system and DG units (e.g., system data, system constraints, DG
1000
types, and historical/long-term forecast of DG production). These data
0
are structured in the second stage where the probabilities of the gen-
2 5 8 11 14 17 20 23 26 29 32 35 38 41 44 47 50 53 56 59 62 65 68
erated power from DG and the load through the whole planning period
DG location
are estimated, and the optimization model is constructed. The optimi-
Fig. 10. Optimal curtailed energy from PV at the different locations of the zation model described in this section is solved in the third stage. Both
studied system with five different PV penetrations (69-bus distribution system). the reactive power of the DG inverter (represented by Qinv) and the
active power of the DG inverter (represented by Pinv where it is a
20% penetration 40% penetration 60% penetration function in Pcurt) have been optimized in the optimization model. All
80% penetration 100% penetration results are printed in the fourth stage.
The objective function in (19)–(32) is solved by using the interior-
8000 point nonlinear optimization method [42]. This method was first pro-
Curtailed energy (MWh)

7000 posed for linear programming by John von Neumann; afterward, it was
6000 modified to include the nonlinear programming. Because of con-
5000 tinuously differentiable of the objective function and constraints in
4000 (19)–(32), solutions satisfied the Karush–Kuhn–Tucker (KKT) condi-
3000 tions will imply the optimality [43]. There are many solvers of the
2000 interior-point such as KNITRO [44], LOQO [45], and IPOPT [46]. These
1000 are highly efficient and are the only solvers can manage very large-scale
0 problems with very reduced spaces [47]. It is important to note that any
2 5 8 11 14 17 20 23 26 29 32 35 38 41 44 47 50 53 56 59 62 65 68 other nonlinear programming method can be applied to solve the op-
DG location timization problem. A Mathematical Programming Language (AMPL)
[48] is applied in this paper to model the objective function (19)–(32),
Fig. 11. Optimal curtailed energy from WTGS at the different locations of the
while the nonlinear problem is solved via KINTRO solver interfaced
studied system with five different WTGS penetrations (69-bus distribution
system). with MATLAB®, where the Interior/Direct algorithm is used in the
solver [49].

(proportional to the current), and ohmic losses over the impedances
(proportional to the square of the current) at bus k, respectively [41].
It is interesting to note that this model is efficient to choose the
proper control scheme which achieves the goal of minimum total costs.
This selection is modeled using the binary variables x and y (x,
y ∈ {0,1}), where the values of x, y depend on the used control scheme
as follows: (1) x = 0, y = 1 for the control scheme A, (2) x = 1, y = 0
for the control scheme B, and (3) x = y = 1 for the control scheme C.
The following constraints are considered in the optimization model
5. Results and discussions
A 69-bus distribution system shown in Fig. 7 has been used to test
the proposed method. This system is widely used in the literature for
DG integration studies. The system comprises a single 12.66 kV supply
point, 69-buses, 7 laterals, and 68 branches with a radial structure [50].
The total active and reactive loads are 3.802 MW and 2.695 Mvar, re-
spectively. The active and reactive power losses of the system at
nominal loading are 0.230 MW and 0.104 Mvar, respectively. The lower
and upper voltage thresholds should be 0.95 p.u. and 1.05 p.u., re-
spectively [51]. The planning period is assumed to be one year, and the

506
A. Ali et al. Electrical Power and Energy Systems 105 (2019) 500–513

20% penetration 40% penetration 60% penetration 20% penetration 40% penetration 60% penetration
80% penetration 100% penetration 80% penetration 100% penetration
1.2 1.2

1.15 1.15
Voltage (pu)

Voltage (pu)
1.1 1.1
Vmax Vmax
1.05 1.05

1 1

0.95 0.95

0.9 0.9
2 5 8 11 14 17 20 23 26 29 32 35 38 41 44 47 50 53 56 59 62 65 68 2 5 8 11 14 17 20 23 26 29 32 35 38 41 44 47 50 53 56 59 62 65 68
DG location DG location
(a) without applying the proposed method. (b) with applying the proposed method.
Fig. 12. Voltage profiles at PCC with installing PV system in the 69-bus distribution system (extreme state).

20% penetration 40% penetration 60% penetration 20% penetration 40% penetration 60% penetration
80% penetration 100% penetration 80% penetration 100% penetration
1.2 1.2

1.15 1.15
Voltage (pu)

1.1 Voltage (pu) 1.1

Vmax 1.05
Vmax
1.05

1 1

0.95 0.95

0.9 0.9
2 5 8 11 14 17 20 23 26 29 32 35 38 41 44 47 50 53 56 59 62 65 68 2 5 8 11 14 17 20 23 26 29 32 35 38 41 44 47 50 53 56 59 62 65 68
DG location DG location
(a) without applying the proposed method. (b) with applying the proposed method.
Fig. 13. Voltage profiles at PCC with installing WTGS in the 69-bus distribution system (extreme state).

parameters of the cost functions are given in the Appendix. Note that
only one DG is connected to the system, but its location is changed. The
probabilities of PV power and WTGS power are given in Table 1
[33,34].
5.1. Optimal inverter oversize of different DG types
In this subsection, the proposed method has been applied to two
different DG types (PV and WTGS) without considering OLTC. The
following DG penetrations with respect to the total load of the studied
distribution system are considered here:
• Low DG penetration (20%)
• Medium DG penetration (40%, 60%)
• High DG penetration (80%, 100%)
The calculated inverter oversize as a percent of its rated size and
total curtailed energy during the planning period for different DG types
and at the different DG locations with low, medium, and high DG pe-
netrations are shown in Figs. 8–11. As shown in these figures, the in-
verter oversizing is the optimal selection of the optimization algorithm
in some locations instead of APC or the combination of them (e.g., buses
59 and 60 with low and medium penetration of PV, and buses 57 and 58
with low and medium penetration of WTGS) and only APC option is
selected in other locations (e.g., buses 34 and 35 with all different
penetrations of PV, and buses 32 and 33 with all different penetrations
of WTGS). Furthermore, in some locations, neither inverter oversizing
nor APC is required to minimize the objective function (e.g., buses 3
and 4 with all different penetrations of PV, and buses 41 and 42 with all
different penetrations of WTGS). On the other hand, both options of
APC and inverter oversizing are the optimal selection of the optimiza-
tion algorithm in some locations (e.g., buses 62–64 with medium and
high penetrations of PV, and buses 57–59 with the high penetration of
WTGS). It is important to note that in the case of calculating real values
for both APC and inverter oversize, the required injected/absorbed
reactive power is supported by the spare capacity of the inverter due to
APC. If this spare capacity is not sufficient for the reactive power sup-
port, the inverter will be oversized to cover the required reactive power
injection/absorption. It is worth to mention that with increasing the DG
penetration the calculated inverter oversize is decreased while the
calculated APC is increased.
The voltage profiles at PCC during the extreme state (i.e., state 120,
in which the DG power is maximum, whereas the load is minimum)
without and with applying the proposed method for the different DG
types are given in Figs. 12 and 13. It is clear that, without applying the
proposed method, the voltages at PCC violate the upper limits in many
locations as shown in the figures. The proposed method regulates the
PCC voltages of all the violated cases.
This analysis demonstrates that the optimal inverter oversize and
APC values vary significantly depending on the DG type, DG location,
and DG penetration. Therefore, it is important to compute the inverter
oversize optimally using the proposed method.
5.2. Comparison of the proposed method with existing methods
To demonstrate the effectiveness of the proposed method, compar-
isons have been performed against the following methods.

507
A. Ali et al. Electrical Power and Energy Systems 105 (2019) 500–513

Method 1 Method 2 Method 3 Method 4 Proposed Method 1 Method 2 Method 3 Method 4 Proposed

4.4 4.4
4 4
3.6 3.6
3.2

Normalized cost
3.2
Normalized cost

2.8 2.8
2.4 2.4
2
2
1.6
1.6
1.2
1.2
0.8
0.8
0.4
0.4 2 5 8 11 14 17 20 23 26 29 32 35 38 41 44 47 50 53 56 59 62 65 68
2 5 8 11 14 17 20 23 26 29 32 35 38 41 44 47 50 53 56 59 62 65 68
DG location
DG location
(a) 20% penetration (b) 40% penetration
Method 1 Method 2 Method 3 Method 4 Proposed Method 1 Method 2 Method 3 Method 4 Proposed

4.4 4.4
4 4
3.6 3.6
3.2 3.2

Normalized cost
Normalized cost

2.8 2.8
2.4 2.4
2 2
1.6 1.6
1.2 1.2
0.8 0.8
0.4 0.4
2 5 8 11 14 17 20 23 26 29 32 35 38 41 44 47 50 53 56 59 62 65 68 2 5 8 11 14 17 20 23 26 29 32 35 38 41 44 47 50 53 56 59 62 65 68
DG location DG location
(c) 60% penetration (d) 80% penetration

Method 1 Method 2 Method 3 Method 4 Proposed

4.4
4
3.6
3.2
Normalized cost

2.8
2.4
2
1.6
1.2
0.8
0.4
2 5 8 11 14 17 20 23 26 29 32 35 38 41 44 47 50 53 56 59 62 65 68
DG location
(e) 100% penetration
Fig. 14. Normalized costs with installing PV at different locations of the 69-bus distribution system.

DG penetrations, the total costs are much high when Methods 1–4 have
Method the inverter oversize is optimally computed to minimize been applied and these total costs are increased with the increasing of
1: the active power loss without considering APC [19] DG penetration. On the contrast, the total costs are significantly de-
Method the inverter is oversized by 60% of its rated capacity [25] creased when the proposed method has been applied.
2: Based on this analysis, the optimal DG location for the different
Method the inverter is oversized to keep the inverter pf constant methods with the different DG penetrations as well as the corre-
3: [27] (e.g., equal to the load pf) sponding inverter oversize and total costs are compared in Tables 2 and
Method the inverter is oversized by 10% of its rated capacity 3 for PV system and WTGS, respectively. Pent., Prop., M1, M2, M3, M4
4: [24,30] denote the DG penetration, proposed method, Method 1, Method 2
The total costs obtained by applying those methods have been com- Method 3, and Method 4, respectively. From those tables, it is clear that
pared with the total costs obtained by applying the proposed method, as the optimal DG location and inverter oversize depend on the used
shown in Figs. 14 and 15. In the case of Method 1, only active power method, DG type, and DG penetration. Moreover, the proposed method
losses are considered to compute the optimal inverter oversize. The is better than the other methods for different DG types and DG pene-
inverter is oversized by fixed percentages of the rated inverter size in tration in minimizing the total costs, especially at high DG penetration.
Methods 2–4. These percentages are not the optimal oversize. It can be It is worth to mention that, Method 2 is not recommended for high DG
seen from Figs. 14 and 15 that at all DG locations and with the different penetration due to high total costs.

508
A. Ali et al. Electrical Power and Energy Systems 105 (2019) 500–513

Method 1 Method 2 Method 3 Method 4 Proposed Method 1 Method 2 Method 3 Method 4 Proposed

4.4 4.4
4
4
3.6
3.6
3.2

Normalized cost
3.2
Normalized cost

2.8
2.8
2.4
2.4
2
2
1.6
1.6
1.2
1.2
0.8
0.8
0.4
0.4 2 5 8 11 14 17 20 23 26 29 32 35 38 41 44 47 50 53 56 59 62 65 68
2 5 8 11 14 17 20 23 26 29 32 35 38 41 44 47 50 53 56 59 62 65 68
DG location
DG location
(a) 20% penetration (b) 40% penetration

Method 1 Method 2 Method 3 Method 4 Proposed Method 1 Method 2 Method 3 Method 4 Proposed

4.4 4.4
4 4
3.6 3.6
3.2

Normalized cost
3.2
Normalized cost

2.8 2.8
2.4 2.4
2 2
1.6 1.6
1.2 1.2
0.8 0.8
0.4 0.4
2 5 8 11 14 17 20 23 26 29 32 35 38 41 44 47 50 53 56 59 62 65 68 2 5 8 11 14 17 20 23 26 29 32 35 38 41 44 47 50 53 56 59 62 65 68
DG location DG location
(c) 60% penetration (d) 80% penetration

Method 1 Method 2 Method 3 Method 4 Proposed

4.4
4
3.6
3.2
Normalized cost

2.8
2.4
2
1.6
1.2
0.8
0.4
2 5 8 11 14 17 20 23 26 29 32 35 38 41 44 47 50 53 56 59 62 65 68
DG location

(e) 100% penetration

Fig. 15. Normalized costs when installing WTGS at the different locations of the 69-bus distribution system.

Table 2
Optimal DG location, inverter oversize, and total costs with installing PV in the 69-bus distribution system.
Pent. Optimal DG location Oversize % Total normalized cost

M1 M2 M3 M4 Prop. M1 M2 M3 M4 Prop. M1 M2 M3 M4 Prop.

20% 35 61 61 61 56 0.05 60 22.55 10 0 0.937 1.334 1.086 0.981 0.916

40% 35 35 35 35 35 0.04 60 22.55 10 0 1.019 1.802 1.455 1.295 1.002
60% 32 35 35 35 27 0.02 60 22.55 10 0.01 1.023 1.992 1.563 1.365 1.000
80% 31 27 27 27 27 0.02 60 22.55 10 0 1.027 2.052 1.576 1.356 0.961
100% 30 27 27 27 27 0.01 60 22.55 10 0 1.032 1.978 1.511 1.295 0.907

5.3. Effect of OLTC analysis, optimal inverter oversize, optimal APC, and optimal tap po-
sition of OLTC are computed simultaneously. The tap ratios of OLTC are
In this subsection, the effect of OLTC (at the substation) operation taken as 33 tap positions (16 taps up, 16 taps down, and neutral tap).
on the total costs and inverter oversize has been studied. In this The transformer secondary voltage is regulated at 1.0 pu with the

509
A. Ali et al. Electrical Power and Energy Systems 105 (2019) 500–513

Table 3
Optimal DG location, inverter oversize, and total costs with installing WTGS in the 69-bus distribution system.
Pent. Optimal DG location Oversize % Total normalized cost

M1 M2 M3 M4 Prop. M1 M2 M3 M4 Prop. M1 M2 M3 M4 Prop.

20% 35 61 61 61 19 0.05 60 22.55 10 0 1.012 1.461 1.195 1.085 1.000

40% 35 35 35 35 35 0.04 60 22.55 10 0 1.019 1.792 1.450 1.291 1.001
60% 32 35 27 27 27 0.02 60 22.55 10 0.005 1.023 1.970 1.542 1.341 0.982
80% 31 27 27 27 27 0.02 60 22.55 10 0.003 1.027 1.954 1.506 1.299 0.927
100% 30 27 27 27 27 0.01 60 22.55 10 0.002 1.031 1.870 1.431 1.229 0.864

PV system WTGS PV system WTGS

16
16
14
14
8 (bus 56)
Optimal tap position

12
7 (bus 19)

12

Cost reduction (%)

9.10
10
10

7.77
8
3 (bus 35)

3 (bus 27)
3 (bus 27)
3 (bus 35)

8
2 (bus 27)

2 (bus 27)
2 (bus 27)

2 (bus 27)
6
6

2.95
4

2.58
2.57
2.18
4

1.89
1.89

1.54
1.54
2 2
0 0
20 40 60 80 100 20 40 60 80 100
DG penetration (%) DG penetration (%)
(a) (b)

PV system WTGS
120
100
100

100

100
Oversize reduction (%)

100

80

60

40

20
0
0

0
0

0
20 40 60 80 100
DG penetration (%)
(c)
Fig. 16. Results with considering OLTC: a) Optimal tap position of OLTC; b) CR with considering OLTC; c) OSR with considering OLTC.

19 20 21 22 26 27 28 29 30 31 32 33 position of OLTC reaches to eighth position (1.05) in case of PV system

while it reaches to seventh position (1.04375) WTGS. On the other
hand, at medium DG penetrations (40% and 60%), the tap position of
OLTC reaches the third position (1.01875) for both DG types. The tap of
OLTC reaches to the second position (1.0125) at high DG penetration
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
OLTC (80% and 100%) for both DG types. It is clear that, for different DG
types and different DG penetrations, the tap position of OLTC is greater
than the neutral tap (1.00), unlike the analysis described in subsections
Slack Bus 23 24 25 5.1 and 5.2, in which the tap position of OLTC was fixed at the neutral
tap. From this analysis, the optimal tap position of OLTC differs de-
pending on the DG location and penetration.
Fig. 17. Single line diagram of 33-bus distribution system.
The percentages of cost reduction (CR) and oversize reduction
(OSR) of the inverter with considering OLTC can be described as fol-
lows:
neutral tap and changes in steps of order 0.625%. The results of con-
sidering OLTC is shown in Fig. 16. Fig. 16(a) shows the optimal tap Ctotal, k, o−Ctotal, k, N
CRk (%) = × 100
positions of OLTC for different DG types at optimal DG locations with Ctotal, k, o (33)
different DG penetrations.
Sos, k, o−Sos, k, N
As shown in Fig. 16(a), at low DG penetration (20%), the tap OSRk (%) = × 100
Sos, k, o (34)

510
A. Ali et al. Electrical Power and Energy Systems 105 (2019) 500–513

20% penetration 40% penetration 60% penetration 27), respectively. Therefore, considering the operation of OLTC in the
80% penetration 100% penetration optimization algorithm has a significant benefit in reducing the total
30 cost.
Fig. 16(c) illustrates the OSR values at different DG locations and
Inverter oversize (%)

25 DG penetrations for different DG types. It is noted that there is no re-

20 duction in the inverter oversize at some locations and penetrations of
DG because the inverter is not oversized with/without considering
15 OLTC, while OSR is 100% at the locations and penetrations of DG in
10 which the inverters is oversized without considering OLTC.

5 5.4. Results of the IEEE 33-bus distribution system

0
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 To evaluate the effectiveness of the proposed method when applied
DG location to a different test system, the IEEE 33-bus distribution system, 12.66 kV
with total demand of 3.715 MW and 2.3 Mvar was selected for this
Fig. 18. Optimal oversize of the PV inverter at different locations of the studied
purpose as shown in Fig. 17. The details of this distribution system are
system (33-bus system) with five different PV penetrations.
available in [52]. To avoid results repetition, only one DG type (PV) is
considered in this analysis while the results for the other DG type follow
20% penetration 40% penetration 60% penetration the same trend.
80% penetration 100% penetration The inverter oversize and the total curtailed energy during the
8000 planning period at the different DG locations of the studied system with
7000
Curtailed energy (MWh)

different DG penetrations (low, medium, and high) are shown in

6000
5000
4000
3000
2000
1000
0 with applying the proposed method are given in Fig. 20. Without ap-
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32
DG location
Fig. 19. Optimal curtailed energy from PV at the different locations of the
studied system (33-bus system) with five different PV penetrations.
CR formulated in (33) is shown in Fig. 16(b) at optimal DG locations
with different DG penetrations for different DG types. It is observed
from this figure (Fig. 16(b)) that the value of CR varies with the DG
type, DG location, and DG penetration. In case of the PV system, the
maximum CR is 9.10% at bus 56 (optimal location of 20% penetration)
with 20% DG penetration, while the minimum CR is 1.54% with 100%
DG penetration at bus 27 (optimal location of 100% penetration). The
value of CR at other DG locations and DG penetrations varies between
those maximum and minimum values. On the other hand, in the case of
WTGS, the maximum and minimum CR are 7.77% and 1.54% with 20%
and 100% DG penetrations at their optimal locations (bus 19 and bus
Figs. 18 and 19, respectively. As shown in these figures, like the first
test system (69-bus distribution system), the optimal selection of the
optimization algorithm varies among inverter oversizing, APC, or both
of them according to the DG location and DG penetration. At certain
locations, neither inverter oversizing nor APC is selected by the opti-
mization algorithm.
The voltage profiles at PCC during the extreme state without and
plying the proposed method, the voltage violates the upper limit of the
voltage at many locations with different DG penetrations. In turn, when
applying the proposed method, the voltages at PCC are kept within
acceptable limits at all DG locations with different DG penetrations as
shown in Fig. 19(b). Therefore, the proposed method can solve the
voltage rise problem at the high penetration of DG.
The optimal DG location, the corresponding inverter oversize, and
the corresponding total costs by applying the different methods to the
studied distribution system with the different DG penetrations are given
in Table 4. This table shows that the proposed method gives the
minimum total cost for different DG penetration, especially at high DG
penetration.
The previous results demonstrate the effectiveness of the proposed
method for calculating the optimal inverter oversize and optimal APC
to prevent the voltage rise problem with minimum cost. Furthermore,
the proposed method is applicable to different DG types, DG locations,
DG penetrations, and distribution systems. Therefore, it is

20% penetration 40% penetration 60% penetration 20% penetration 40% penetration 60% penetration
80% penetration 100% penetration 80% penetration 100% penetration
1.2 1.2

1.15 1.15
Voltage (pu)

Voltage (pu)

1.1 1.1
Vmax Vmax
1.05 1.05

1 1

0.95 0.95

0.9 0.9
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32
DG location DG location
(a) without applying the proposed method (b) with applying the proposed method
Fig. 20. Voltage profiles at PCC with installing PV in the 33-bus distribution system (extreme state).

511
A. Ali et al. Electrical Power and Energy Systems 105 (2019) 500–513

Table 4
Optimal DG location, inverter oversize, and total costs with installing PV in the 33-bus distribution system.
Pent. Optimal DG location Oversize % Total normalized cost

M1 M2 M3 M4 Prop. M1 M2 M3 M4 Prop. M1 M2 M3 M4 Prop.

20% 30 22 30 30 30 95.37 60 17.65 10 25.91 1.556 1.506 1.208 1.132 0.905

40% 22 22 22 22 5 0.70 60 17.65 10 0.00 2.188 1.895 1.443 1.327 0.998
60% 18 22 18 18 5 2.70 60 17.65 10 0.00 1.805 2.162 1.553 1.395 0.999
80% 18 18 18 18 18 1.53 60 17.65 10 0.48 1.373 2.088 1.482 1.327 0.984
100% 18 18 18 18 18 0.98 60 17.65 10 0.31 1.060 1.934 1.370 1.226 0.889

recommended to use the proposed method for computing the inverter
oversize optimally.
6. Conclusions
In this paper, an optimization method is proposed for determining
the optimal oversize of DG inverters for voltage regulation by con-
sidering reactive power capability and APC. The costs of inverter
oversizing and APC are incorporated in the optimization model.
Additionally, the active losses in MV distribution systems and DG in-
verter are considered. The optimal oversizing of DG inverters for re-
active power control can be accurately computed for voltage regulation.
The results confirmed that considering APC in the proposed inverter
oversizing method can effectively regulate voltages with minimum total
costs, especially at high DG penetration. It is also demonstrated that the
optimal inverter oversize differs depending on DG type, location, and
penetration. The proposed method is a helpful tool for system planners
to calculate the optimal inverter oversize for voltage regulation at dif-
ferent DG penetration levels.

A. Appendix

See Tables 5 and 6.

Table 5
Cost rates.
Cost rate $/kWh

CL 0.06
Cinv 0.06

Table 6
Cost function parameters.
Variable Value

ainv 3.8 × 10−4

binv 7.6 × 10−4
cinv 7.6 × 10−4
as 0
bs 25
cs 0.25
ap 0
bp 50
cp 0.5

Appendix B. Supplementary material

Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.ijepes.2018.08.040.

References
[1] Aman MM, Jasmon GB, Bakar AHA, Mokhlis H. A new approach for optimum DG
placement and sizing based on voltage stability maximization and minimization of
power losses. Energy Convers Manag 2013;70:202–10. https://doi.org/10.1016/J.
ENCONMAN.2013.02.015.
[2] Mahmoud K, Abdel-Akher M, Ahmed A-FA. Sizing and locating distributed gen-
erations for losses minimization and voltage stability improvement. In: PECon2010 -
2010 IEEE Int. Conf. Power Energy; 2010. http://doi.org/10.1109/PECON.2010.
5697652.
[3] Mokhlis H, Jasmon GB, Aman MM, Bakar AHA. Optimum tie switches allocation
and DG placement based on maximisation of system loadability using discrete ar-
tificial bee colony algorithm. IET Gener Transm Distrib 2016;10:2277–84. https://
doi.org/10.1049/iet-gtd.2015.0303.
[4] Vatani M, Gharehpetian GB, Sanjari MJ, Solati Alkaran D. Multiple distributed
generation units allocation in distribution network for loss reduction based on a
combination of analytical and genetic algorithm methods. IET Gener Transm Distrib
2016;10:66–72. https://doi.org/10.1049/iet-gtd.2015.0041.
[5] Aman MM, Jasmon GB, Bakar AHA, Mokhlis H. A new approach for optimum si-
multaneous multi-DG distributed generation Units placement and sizing based on
maximization of system loadability using HPSO (hybrid particle swarm optimiza-
tion) algorithm. Energy 2014;66:202–15. https://doi.org/10.1016/J.ENERGY.
2013.12.037.
[6] Mahmoud K, Yorino N, Ahmed A. Optimal distributed generation allocation in
distribution systems for loss minimization. IEEE Trans Power Syst 2016;31:960–9.
https://doi.org/10.1109/TPWRS.2015.2418333.
[7] Bollen MHJ. Integration of distributed generation in the power system. Wiley; 2011.
[8] Mahmud N, Zahedi A. Review of control strategies for voltage regulation of the
smart distribution network with high penetration of renewable distributed gen-
eration. Renew Sustain Energy Rev 2016;64:582–95. https://doi.org/10.1016/j.
rser.2016.06.030.

512
A. Ali et al.
[9] Mohammadi P, Mehraeen S. Challenges of PV integration in low-voltage secondary
networks. IEEE Trans Power Deliv 2017;32:525–35. https://doi.org/10.1109/
TPWRD.2016.2556692.
[10] Bayod-Rújula AA. Future development of the electricity systems with distributed
generation. Energy 2009;34:377–83. https://doi.org/10.1016/J.ENERGY.2008.12.
008.
[11] Masters CL. Voltage rise: the big issue when connecting embedded generation to
long 11 kV overhead lines. Power Eng J 2002;16:5–12. https://doi.org/10.1049/
pe:20020101.
[12] Agalgaonkar YP, Pal BC, Jabr RA. Distribution voltage control considering the
impact of PV generation on tap changers and autonomous regulators. IEEE Trans
Power Syst 2014;29:182–92. https://doi.org/10.1109/TPWRS.2013.2279721.
[13] Elrayyah AY, Wanik MZC, Bouselham A. Simplified approach to analyze voltage rise
in LV systems with PV installations using equivalent power systems diagrams. IEEE
Trans Power Deliv 2017;32:2140–9. https://doi.org/10.1109/TPWRD.2016.
2594242.
[14] De Carne G, Buticchi G, Zou Z-X, Liserre M. Reverse power flow control in a ST-fed
distribution grid. 1-1 IEEE Trans Smart Grid 2017;1. https://doi.org/10.1109/TSG.
2017.2651147.
[15] Bouhouras AS, Sgouras KI, Gkaidatzis PA, Labridis DP. Optimal active and reactive
nodal power requirements towards loss minimization under reverse power flow
constraint defining DG type. Int J Electr Power Energy Syst 2016;78:445–54.
https://doi.org/10.1016/J.IJEPES.2015.12.014.
[16] Mehmood KK, Khan SU, Lee S-J, Haider ZM, Rafique MK, Kim C-H. A real-time
optimal coordination scheme for the voltage regulation of a distribution network
including an OLTC, capacitor banks, and multiple distributed energy resources. Int
J Electr Power Energy Syst 2018;94:1–14. https://doi.org/10.1016/J.IJEPES.2017.
06.024.
[17] Farivar M, Neal R, Clarke C, Low S. Optimal inverter VAR control in distribution
systems with high PV penetration. In: 2012 IEEE Power Energy Soc. Gen. Meet.,
IEEE; 2012. p. 1–7. http://doi.org/10.1109/PESGM.2012.6345736.
[18] Lelis IS, Drougakis IA, Pompodakis EE, Alexiadis MC. Photovoltaic systems in low-
voltage networks and overvoltage correction with reactive power control. IET
Renew Power Gener 2016;10:410–7. https://doi.org/10.1049/iet-rpg.2014.0282.
[19] Turitsyn K, Sulc P, Backhaus S, Chertkov M. Options for control of reactive power
by distributed photovoltaic generators. Proc IEEE 2011;99:1063–73. https://doi.
org/10.1109/JPROC.2011.2116750.
[20] Varma RK, Rahman SA, Vanderheide T. New control of PV solar farm as STATCOM
(PV-STATCOM) for increasing grid power transmission limits during night and day.
IEEE Trans Power Deliv 2015;30:755–63. https://doi.org/10.1109/TPWRD.2014.
2375216.
[21] Ali A, Raisz D, Mahmoud K. Adaptive multi-objective optimization for power loss
minimization and voltage regulation in distribution systems. In: 2016 Eighteenth
Int. Middle East Power Syst. Conf., IEEE; 2016. p. 573–8. http://doi.org/10.1109/
MEPCON.2016.7836949.
[22] Jashfar S, Esmaeili S. Volt/var/THD control in distribution networks considering
reactive power capability of solar energy conversion. Int J Electr Power Energy Syst
2014;60:221–33. https://doi.org/10.1016/J.IJEPES.2014.02.038.
[23] Rafi FHM, Hossain MJ, Lu J. Hierarchical controls selection based on PV penetra-
tions for voltage rise mitigation in a LV distribution network. Int J Electr Power
Energy Syst 2016;81:123–39. https://doi.org/10.1016/J.IJEPES.2016.02.013.
[24] Turitsyn K, Sulc P, Backhaus S, Chertkov M. Local control of reactive power by
distributed photovoltaic generators. In: 2010 First IEEE Int. Conf. Smart Grid
Commun., IEEE; 2010. p. 79–84. http://doi.org/10.1109/SMARTGRID.2010.
5622021.
[25] Su X, Masoum MAS, Wolfs PJ. Optimal PV inverter reactive power control and real
power curtailment to improve performance of unbalanced four-wire LV distribution
networks. IEEE Trans Sustain Energy 2014;5:967–77. https://doi.org/10.1109/
TSTE.2014.2313862.
[26] Tonkoski R, Lopes LAC, El-Fouly THM. Coordinated active power curtailment of
grid connected PV inverters for overvoltage prevention. IEEE Trans Sustain Energy
2011;2:139–47. https://doi.org/10.1109/TSTE.2010.2098483.
[27] Ghosh S, Rahman S, Pipattanasomporn M. Distribution voltage regulation through
513
View publication stats
Electrical Power and Energy Systems 105 (2019) 500–513
active power curtailment With PV inverters and solar generation forecasts. IEEE
Trans Sustain Energy 2017;8:13–22. https://doi.org/10.1109/TSTE.2016.2577559.
[28] Ugranlı F, Karatepe E. Multi-objective transmission expansion planning considering
minimization of curtailed wind energy. Int J Electr Power Energy Syst
2015;65:348–56. https://doi.org/10.1016/J.IJEPES.2014.10.031.
[29] Madureira AG, Peças Lopes JA. Coordinated voltage support in distribution net-
works with distributed generation and microgrids. IET Renew Power Gener
2009;3:439. https://doi.org/10.1049/iet-rpg.2008.0064.
[30] AlKaabi SS, Khadkikar V, Zeineldin HH. Incorporating PV inverter control schemes
for planning active distribution networks. IEEE Trans Sustain Energy
2015;6:1224–33. https://doi.org/10.1109/TSTE.2015.2422305.
[31] Institute of Electrical and Electronics Engineers. IEEE 1547standard for inter-
connecting distributed resources with electric power systems; 2008.
[32] Atwa YM, El-Saadany EF, Salama MMA, Seethapathy R. Optimal renewable re-
sources mix for distribution system energy loss minimization. IEEE Trans Power
Syst 2010;25:360–70. https://doi.org/10.1109/TPWRS.2009.2030276.
[33] Swief R, Abdel-Salam T, El-Amary N. Photovoltaic and wind turbine integration
applying cuckoo search for probabilistic reliable optimal placement. Energies
2018;11:139. https://doi.org/10.3390/en11010139.
[34] Atwa YM, El-Saadany EF. Probabilistic approach for optimal allocation of wind-
based distributed generation in distribution systems. IET Renew Power Gener
2011;5:79. https://doi.org/10.1049/iet-rpg.2009.0011.
[35] Pinheiro JMS, Dornellas CRR, Schilling MT, Melo ACG, Mello JCO. Probing the new
IEEE reliability test system (RTS-96): HL-II assessment. IEEE Trans Power Syst
1998;13:171–6. https://doi.org/10.1109/59.651632.
[36] Singh C, Lago-Gonzalez A. Improved algorithms for multi-area reliability evaluation
using the decomposition-simulation approach. IEEE Trans Power Syst
1989;4:321–8. https://doi.org/10.1109/59.32495.
[37] Singh C, Kim Y. An efficient technique for reliability analysis of power systems
including time dependent sources. IEEE Trans Power Syst 1988;3:1090–6. https://
doi.org/10.1109/59.14567.
[38] Giannakis GB. Optimal Dispatch of Photovoltaic Inverters in Residential
Distribution Systems 2014;5:487–97. http://doi.org/10.1109/TSTE.2013.2292828.
[39] Stetz, TJ, Künschner J, Braun M, Engel B. Cost optimal sizing of photovoltaic in-
verters - influence of new grid codes and cost reductions. 25th Eur PV Sol, vol. 2;
2010. p. 5022–4.
[40] Retail Price List 2018. https://www.rpc.com.au/pricelists/pricelist-retail.pdf (ac-
cessed June 17, 2018).
[41] Braun M. Reactive power supplied by PV inverters - Cost-Benefit-Analysis. In: 22nd
Eur Photovolt Sol Energy Conf Exhib 3 2007:7p.
[42] Forsgren A, Gill PE, Wright MH. Interior methods for nonlinear optimization. SIAM
Rev 2002;44:525–97. https://doi.org/10.1137/S0036144502414942.
[43] Kuhn HW, Tucker AW. Nonlinear programming. In: Proc 2nd Berkeley Symp; 1951.
p. 481–92.
[44] Waltz RA, Plantenga TD. User’s Manual Knitro User’s Manual; 2010.
[45] Vanderbei RJ. LOQO user’s manual — version 3.10. Optim Methods Softw
1999;11:485–514. http://doi.org/10.1080/10556789908805760.
[46] Wächter A, Biegler LT. On the implementation of an interior-point filter line-search
algorithm for large-scale nonlinear programming. Math Program 2006;106:25–57.
https://doi.org/10.1007/s10107-004-0559-y.
[47] Benson HY, Vanderbei RJ, Shanno DF. Interior-point methods for nonconvex non-
linear programming: filter methods and merit functions. Comput Optim Appl
2002;23:257–72. https://doi.org/10.1023/A:1020533003783.
[48] A Mathematical Programming Language (AMPL) n.d. https://ampl.com/.
[49] https://www.artelys.com/tools/knitro_doc/2_userGuide/algorithms.html n.d.
[50] Baran ME, Wu FF. Optimal capacitor placement on radial distribution systems. IEEE
Trans Power Deliv 1989;4:725–34. https://doi.org/10.1109/61.19265.
[51] Australian Standard AS/NZS 61000.3.100:2011, Electromagnetic compatibility
(EMC) – Limits – Steady state voltage limits in public electricity systems; 2011.
[52] Baran ME, Wu FF. Network reconfiguration in distribution systems for loss reduc-
tion and load balancing. IEEE Trans Power Deliv 1989;4:1401–7. https://doi.org/
10.1109/61.25627.