Automatika

Journal for Control, Measurement, Electronics, Computing and

Communications

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/taut20

Bus selection index for distributed generators

placement and sizing in the electrical network

Muhammad Raza, Aurangzeb Rashid Masud, Ayaz Hussain & Ravi Mohan Lal

To cite this article: Muhammad Raza, Aurangzeb Rashid Masud, Ayaz Hussain & Ravi Mohan
Lal (2022): Bus selection index for distributed generators placement and sizing in the electrical
network, Automatika, DOI: 10.1080/00051144.2022.2130257

To link to this article: https://doi.org/10.1080/00051144.2022.2130257

© 2022 The Author(s). Published by Informa

UK Limited, trading as Taylor & Francis
Group.

Published online: 06 Oct 2022.

Submit your article to this journal

Article views: 50

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=taut20
AUTOMATIKA
https://doi.org/10.1080/00051144.2022.2130257

REGULAR PAPER

Bus selection index for distributed generators placement and sizing in the
electrical network
Muhammad Raza a , Aurangzeb Rashid Masud a , Ayaz Hussaina and Ravi Mohan Lalb
a Department of Electrical Engineering, Bahria University Karachi Campus, Karachi, Pakistan; b School of Science and Engineering,
University of Dundee, Dundee, UK

ABSTRACT ARTICLE HISTORY

Renewable energy sources, widely considered as distribution generation units, will be essential Received 20 December 2021
parts of the future electrical network. They will be integrated into the network at all voltage levels Accepted 24 September 2022
and act as the main controlling elements in the microgrid to stabilize the network. However, the KEYWORDS
objective of determining the suitable busbar to integrate the renewable generation source and Renewable energy;
its maximum capacity is still an indecisive problem. This article presents a methodology to select distribution generation;
suitable busbars to integrate the renewable generation unit. A bus selection index is proposed optimum placement and
based on three criteria, i.e. voltage deviation, active power loss, and grid energy infeed. The sizing; network operational
developed algorithm considers network busbar voltage limit and branch loading as constraints limits; analytical technique
and the solution is determined using analytical technique. Algorithm is implemented in MATLAB,
and MATPOWER is used for the power flow analysis. Furthermore, the proposed methodology is
analysed on CIGRE and IEEE medium voltage benchmark network.

1. Introduction performance if sited incorrectly and have an inappro-

Affordable and clean energy is one of the prime pillars priate size [5]. Distribution sources are placed and sized
of the United Nations (UN) sustainable development on the basis of certain technical and economical crite-
goals [1]. The importance of affordability with clean ria such as minimizing the network active power losses,
energy goals got precedence as the world economy improving network voltage profile, reducing the energy
and health system shattered with COVID-19 pandemic. import from the main grid, minimizing the investment
According to UN statistics, one of four hospitals has cost, and maximizing profit [6,7]. Depending on the
no reliable electricity in the developing countries [1]. system study, either all or any one of these criteria can
Traditional electrical network focus around central- be applied for bus selection. However, usually, the eco-
ized power generation, which produce more losses with nomical criteria is applied for the selection of DGs
increasing power flow and high voltage drops for longer among different types, i.e. wind turbine, PV system,
transmission lines. Without significant reinforcement fuel cells, biomass generator, energy storage system, and
in the electrical transmission and distribution capacity, micro-hydro turbines, while the DG placement is based
the consistently growing energy demand cannot be ful- on the technical constraints.
filled even with the availability of large power plants. Distribution generation units are usually integrated
Alternatively, power generation system spread across with low to medium voltage networks. The variation
the electrical network offer many benefits like energy in the energy infeed by the DGs have high impact on
generation where needed, lower losses, improved sys- the network operating conditions such as the impact of
tem stability, and convenient integration of renew- PV system on the rural and urban network is addressed
able generation system [2]. Many countries such as in Ref. [8]. In Ref. [9], the optimum location of DG
Denmark and Germany are shifting from central- in the radial network is based on voltage stability,
ized to decentralized electrical structure using renew- voltage deviation, and active power loss index. Simi-
able energy sources as distributed power generation larly, in Refs. [10–16] the deployment of energy stor-
system [3]. age units and the analysis of PV systems effects on
Renewable sources have intermittent nature, and the the network have been performed to determine the
renewable power plants (operating as distributed gener- need of network reconfiguration. Static load model
ation) posses small capacity which limit their capability is usually applied when considering only DG output
to provide ancillary services such as reactive power sup- power variation for the analysis, but some researches
port, frequency control, etc. [4]. Thus, distributed gen- have also considered the time-varying nature of the
eration (DG) can significantly deteriorate the network load [17–19].

CONTACT Muhammad Raza [email protected]

© 2022 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted
use, distribution, and reproduction in any medium, provided the original work is properly cited.
2 M. RAZA ET AL.

The problem of integrating DGs can be solved they are based on the network performance cri-
through analytical and optimization techniques. In ana- teria. This enables us to assess the network
lytical method, DG is connected to the network buses behaviour precisely.
sequentially and power flow is calculated recursively – Since voltage is the primary parameter to be influ-
over the range of DG active and reactive power. Results enced by the integration of the DG, the voltage
such as voltage deviation, active power loss, grid infeed index is quantified by three aspects rather than
and reactive power are recorded and compared to considering only DG impact on average voltage of
achieve the desired objectives without violating oper- the whole network.
ating limits [9,20–26]. This technique is easy to imple- – A novel probabilistic-based penalty factor is intro-
ment for a small network but mathematical formula- duced for busbars voltage and branches loading
tion for multiple objectives and result analysis become limits violation.
complex for large networks. On the other hand, opti- – The proposed index identifies the bus location and
mization algorithms such as particle swarm method, DG capacity for optimal network operation or
ant colony, genetic algorithm, tabu search, etc. are also maximum power integration.
being used by the researchers for the DG placement
and sizing [27–32]. In Refs. [33–35], the multi-objective Rest of the article is organized as follows: math-
optimization problem is formulated as mono-objective ematical formulation of the three sub-indices is pre-
function in which the energy storage system and DG sented in Section 2. Then, the voltage and loading limits
location is optimized according to their investment and penalty factor is defined using probabilistic approach.
network operational costs. In Ref. [7], a multi-objective In Section 3, an algorithm is presented to solve the
optimization problem is formulated to select DG size DG integration problem with an analytical approach
according to total energy loss and maximizing DG gen- using selected indices. The proposed method is anal-
eration profit. Literature shows that the mathematical ysed on CIGRE and IEEE medium voltage benchmark
and theoretical DG placement definition as optimiza- by implementing algorithm in MATLAB using MAT-
tion problem provides flexibility in implementing a POWER functions, and simulation results are discussed
large network, however, analysing the large solution in Section 4. Finally some significant conclusions are
data is complicated. Optimization results provide large drawn in Section 5.
data sets of voltages, power, and loading of the sys-
tem for different scenarios that need to be analysed and
2. Distribution generation placement and size
compared for the bus selection and DG sizing. Large
criteria
solution data problem is somewhat solved by defin-
ing evaluation indicator [36–38]. In Ref. [36], annual Distribution generation sources of small capacity are
energy loss indication is formulated considering only usually integrated in low and medium voltage networks
DG active power. In Ref. [6], an optimal locator index i.e below 50 MW. The selection of DG capacity and
is defined based on power loss sensitivity factor. placement depends on both economical and technical
Optimum bus selection through evaluation indica- feasibility. Economically, lower installation cost of DG
tor is relatively easy, however in existing literature, bus with high energy penetration is desirable to maximize
selection index is mostly based on single parameter the profit. From technical feasibility perspective, the
only, such as energy loss index or voltage stability index DG integration should not violate the voltage deviation
or emission reduction, etc., and they cannot be mod- limits, must not exceed the equipment thermal limits,
ified easily to incorporate other criteria [13,30,39,40]. and should not increase losses in the network. This arti-
Furthermore, the DG sizing in the literature mainly cle focuses only on technical parameters for the bus
address considering only DG active power. It is obvi- selection to integrate DG into the network. Primarily,
ous that DGs can contribute to voltage support, thus it the suitability of DG is determined through indexes of
is important to analyse DG effect on its connected bus- voltage, active losses, and grid power (either import or
bar as well as on all other busbars voltages individually export). Furthermore, busbars operating voltages and
with respect to reactive power infeed [3,4]. branches (line, transformer, etc.) thermal loading limits
In this article, bus selection index for DG integra- are accounted as a penalty factor.
tion is proposed as a function of DG active and reactive Consider a distribution network of Figure 1 hav-
power based on three sub-indices, i.e voltage index, grid ing n number of busbars. The network configuration
power index, and power loss index. For better illustra- could be a radial or meshed, and the proposed bus
tion, the contributions of the paper can be summarized selection methodology is applicable to both medium
as follows: and low voltage networks. The network may consist of
multiple feeding points from the national high voltage
– A new methodology is proposed to formulate the transmission grid, which is represented by the equiva-
indexes as non-conflicting objective function and lent grid model. Grid provides power balancing for any
 AUTOMATIKA 3

2.1. Voltage index (VI)

The voltage index indicates the amount of change in the
network voltages due to the addition of the DG. There
are three aspects related to the voltage change i.e.

1. Change in the ith busbar voltage when DG is con-

nected with it.
Figure 1. Distribution generation integration in the network. 2. Average change in the network voltages (exclud-
ing ith busbar) when the DG is connected with ith
busbar.
mismatch between distribution generation and load. 3. Average change in the ith busbar voltage when
Distribution generation systems mainly feed energy the DG is connected with other network busbars
into the grid through voltage source converter, which (excluding ith busbar).
can control the active and reactive power indepen-
dently [3,4]. Thus, DGs can be operated as either con- Here, i is the index of a bus at which VI is calculated.
stant reactive power, power factor, or voltage magnitude The function a fi (u) is calculated using Equation (2)
control device. In this research, DG reactive power is to analyse the sensitivity of the bus according to the
controlled to maintain the constant power factor at power infeed by the DG. The rise and drop in the
the connected busbar. Power infeed by the distribution voltage magnitude based on the amount of reactive
source changes the busbar voltages, varies the loading power infeed (either capacitive or inductive) while
of the lines and transformers that consequently effects active power infeed usually raise the bus voltage. At this
the losses, and influences the power infeed of the grid. stage, it is insignificant to determine whether the volt-
Thus, the selection indexes are functions of DGs active ages are improved or not rather the main objective is to
and reactive power. A bus selection index (BSI) is intro- identify the busbar at which maximum change in the
duced that depends on three indexes, i.e voltage index voltage can be achieved.
(VI), power loss index (PLI), and grid power index
(GPI). These indexes are calculated for each busbar y 
 ui − ui0 
and compared for best location and size of the DGs. a
y fi (u) =   i ∈ Z+ , i ≤ n
The indexes are formulated as minimizing function ui0 
1 a
and can be applied in both analytical and optimization w
a
algorithm to select suitable busbar to integrate DG and fi (u) = fi (u) (2)
w y=1 y
its capacity determination.
The maximum number of DGs that can be con-
nected in the grid is equal to the total number of bus- where y ui is the per unit voltage of the ith busbar with
bars. However, the network planner can choose the case index y, the index y is the combination when at
desired number of DGs to be connected in the net- least one DG is connected with ith bubar (Ai ), w is the
work. The total number of combination for DGs grid total number of combination elements of Ai , and ui0 is
connection is calculated by using Equation (1). the voltage at the ith busbar when DG is not connected
to any busbar.
n! The objective of DG integration is not only to
n Cr = =d (1)
r!(n − r)! improve the voltage level of a single busbar but to make
all busbars voltage level better. The average change in
Here, n is the total number of busbars, and r is the
the network voltage is calculated using Equation (3).
number of DGs that needs to be connected in the net-
work. Further, each case of the combination gives the
information of busbar at which DGs are connected. n  y uj −uj0 
j=1  uj 
For calculating BSI of ith busbar, the combinations are b
y fi (u) = 0
j ∈ Z+ , j = i
divided into two sets, i.e Ai and Bi . The set Ai has the n−1
index of combination cases when at least one DG is 1 
w
b b
connected to the ith busbar and the remaining indexes fi (u) = y fi (u) (3)
w
y=1
belong to the set Bi , thus Ai and Bi are disjoint sets.
Where,
  Here, j is the index of network buses. y uj is the voltage of
Ai : y1 , y2 , y3 , . . . , yw jth busbar when DG is connected to ith busbar, and uj0
  is the voltage of jth busbar when no DG is connected.
Bi : k1 , k2 , k3 , . . . , kd−w
Further, the possibility of improving the ith busbar
Ai ∪ Bi : {e1 , e2 , e3 , . . . , ed } voltage by connecting DG to other busbar is calculated
4 M. RAZA ET AL.

using Equation (4). indicates the reduction in active and reactive power
  infeed by the HV substation.
d−w  k ui −k ui0 
k=1  k ui  y 
1
w
+ pi − pi0 yq − q
i i0
c
fi (u) = k ∈ Z , k ∈ Bi
0
(4) gi (p, q) = + (9)
d−w w y=1 pi0 qi0
Here, the case index k is the combination when no DG
is connected with the ith busbar (Bi ). The effect of the Here, pi0 , and qi0 are the active and reactive power of the
DGs at ith busbar voltage is determined by combining grid without DG, whereas y pi , and y qi are the active and
a
fi (u), b fi (u), and c fi (u) as given by Equation (5). reactive power considering combination cases from set
Ai . The grid power index minimizing function with
fi (u) = a fi (u) + b fi (u) − c fi (u) (5) penalty factor is expressed by Equation (10).

GPIi = gi (p, q) + |gi (p, q) · (2 − u hi − th hi )| (10)

High value of a fi (u) and b fi (u) indicate that the busbar
is suitable for the placement of the DG. While higher
value of c fi (u) implies that the ith bus can be improved 2.4. Network operational limits
by placing DG on other buses, thus it lowers the DG
The energy infeed by the DG’s changes the thermal
placement suitability at ith bus.
loading of the branches (cables and transformers) and
The fi (u) is a maximizing function. To define voltage
voltage level of busbars. These changes must not violate
index (VIi ) as a minimizing function, fi (u) is multiple
the network operational limits. Any violation in the net-
by −1 as given by Equation (6).
work operational limits is considered as penalty factor,
VIi = −fi (u) + |fi (u) · (2 − u hi − th hi )| (6) which reduces the suitability of a busbar as a poten-
tial candidate for connecting DG. Penalty factors are
The power infeed by the DG may violate operating lim- calculated as a cumulative normal distribution func-
its of busbars voltage and branches thermal loading. tion. Using the cumulative distribution function, the
These limits are applied as penalty factors defined by probability that the busbars and branches are having
u
hi and th hi , as explained in later section. values less than or equal to the operational limits is
determined.
Two penalty factors are defined i.e u hi and th hi which
2.2. Power loss index (PLI) indicates the violation of the voltage and thermal load-
The integration of DGs also reduces the losses in the ing limits violation, respectively. To obtain the value
network. The function ri (p) in Equation (7) determines of u hi , absolute change in the bus voltage from the
change in the network active power loss at ith busbar rated value of all network busbars is calculated using
compared to network losses without DG integration. Equation (11) for the combination cases of set Ai .

y  +
j = | uj − 1.0| j ∈ Z : j ≤ n
 y urt y
(11)
1
w
Plossi − Plossi0
ri (p) = (7)
w y=1 Plossi0 Here, j is the index of network buses and n is the total
number of network buses.
Here, Plossi0 is the network total active power loss when Bus voltage varies with the infeed of the DG power.
no DG is connected, and y Plossi is the net active power There might be some buses in the network which are
loss of the network considering combination cases from operating below the operational limits before adding
set Ai . The negative value of ri (p) indicates that the DG power but their voltages get improved afterward.
losses have been reduced, thus power loss index as a Similarly, there could be some buses that may violate
minimizing function with penalty factor can be defined the limits after the addition of the DG power. Limit vio-
as Equation (8). lation can be identified by comparing  y urt j with the
threshold value. Using Equation (11), the suitability of
PLIi = ri (p) + |ri (p) · (2 − u hi − th hi )| (8) connecting DG at a busbar can be compared with others
considering the number of violating buses and voltage
magnitude deviation from the rated value. However,
2.3. Grid power index (GPI)
this information would be complicated to comprehend
Distribution sources support the local energy man- for a large network. Thus, the normal distribution can
agement system and reduce the reliance on feeding be applied on the data acquired using Equation (11) to
grid. The power generation in the distribution network determine the cumulative probability for a given value.
increases the revenue by feeding the power into the This given value would be the limit value. The cumula-
high voltage grid. Further, DGs provide reactive power tive distribution function (CDF) determines the prob-
support to the network as well. The function gi (p, q) ability of the network voltages being less than or equal
 AUTOMATIKA 5

to the limits. The CDF will approach zero when bus- Moreover, the probability that network busbars and
bars violate the operational limit, while the CDF value series branch elements will have voltages and thermal
1.0 shows that no busbar is violating the voltage limit. loading above the operational limits can be calculated
The average, standard deviation, and CDF equations using Equation (15).
are defined as Equation (12).
P(X > ulim ) = 1 − u hi ∵ X =  y urt
j
1 
w
1 
n
u
μi =  y urt
j P(Y > ilim ) = 1 − th hi ∵ Y =  y Ilrt (15)
w y=1
n j=1
These two probabilities are mutually exclusive and
1 1  y rt u 2
w n
u
σi = ( uj − μi ) added together to be applied as a factor of index. For
w y=1 n j=1 example, if f (z) is a minimizing function than the best
result would be at its least value. However, the solu-
 2 
u 1 ulim 1 u − u μi tion would not be suitable if at this least value there
hi = √ exp − du
u σ 2π
i 0 2 uσ
i exists any violation of operational limit. So, f (z) is
(12) increased in a positive direction as a factor of the oper-
The operational limit is applied on the absolute volt- ational limits violation probabilities to worsen its value.
age deviation from the rated value because the busbars Mathematically, it can be expressed as Equation (16).
must not violate both the upper and lower limits. Here,
the magnitude of the deviation is more important rather f̂ (z) = f (z) + |f (z) · (2 − u hi − th hi )| (16)
than the direction of the change i.e. either rise or drop
For no violation situation, the value of u hi = th hi =
from the rated value. Since the deviation value varies
1.0 and the last term of above equation becomes zero
from zero to positive real number, the lower integral
hereby no penalty factor is applied.
limit in the CDF can be set to zero and the upper
limit would be the voltage operational limit. Typically,
voltage deviation limit is ±10% in MV and LV distribu- 2.5. Bus selection index (BSI)
tion network thus, ulim would be 0.1 p.u. Furthermore,
thermal loading of the series elements such as cables Bus selection index of ith busbar is defined as
and transformer is calculated using Equation (13) for Equation (17). Any network busbar having the least
combination cases from set Ai . value of BSI would be the best candidate for integrating
the DG.
yI
l
 y Ilrt = rt (13) BSIi = VIi + PLIi + GPIi (17)
Il
The proposed selection index can be used to find more
Here, l is the index of the element, Ilrt is the element than one suitable busbar by ranking them in ascending
rated current value in ampere, and y Il is the magnitude order based on their BSI value. If more than one busbars
of actual current flowing through l element in ampere. have the same BSI then the decision can be made based
Similarly, CDF is calculated to apply thermal limit. on either observing the power to be feed-in by DGs
The average, standard deviation, and CDF equations (maximum would be preferable), or based on the indi-
are defined as Equation (14) for calculating the proba- vidual index. The developers can decide if they prefer
bility of elements operating below thermal limit. Gener- more voltage to be improved or give priority to losses.
ally, all elements loading must not exceed 100% thus the
thermal limit ilim would be 1.0 p.u. Also, CDF value 1.0
shows that no element has violated the thermal loading 2.6. Performance index (PI)
limit. The value of the element loading varies from zero The performance index indicates the probability of the
to positive real number, thus the lower limit in CDF busbar or the series branch element of violating the
equation is set to zero. network operational limit while connecting DG at all
network busbars. This is expressed by two variables,
1  1  y rt
w m
th
μi =  Il i.e u hrt th rt
i , and hl . To calculate performance index at
w y=1 m the busbars, an absolute voltage deviation of the ith
l=1
busbar voltage from the rated value is calculated using
1 1  y rt th 2
w m
th Equation (18) for all combination cases.
σi = ( Il − μi )
w y=1 m
l=1
⎛  e urt e
i = | ui − 1.0| e ∈ (Ai ∪ Bi ) (18)
2 ⎞
1 ilim 1 I − μi ⎠
th
th
hi = √ exp ⎝− th σ
dI Here, i is the busbar at which performance is calculated,
th σ 2π 0 2 i
i and e is the index of all combination cases. The mean,
(14) standard deviation, and CDF equation of ith busbar
6 M. RAZA ET AL.

voltage is calculated using Equation (19). The voltage is to be ensured that the power flow should converge
operation limits (ulim ) are also set here as 0.1 p.u. at all active power set-points.
• Step 3: At least 100 steps change in the DG out-
1  e rt
d
u rt put power between minimum and maximum val-
μi =  ui
d ues are defined to acquire the sufficient number of
e=1
sample data. However, this step size can be varied
1  e rt u rt 2 considering the algorithm computation time. The
d
u rt
σi = ( ui − μi ) step change in active power can be calculated as
d e=1
(Pd_max − Pd_min )/No. of Sample required.
 2  • Step 4: The DG reactive power capability is set to 0.9
u rt 1 ulim 1 u − u μrti
hi = √ exp − u σ rt
du power factor (capacitive and/or inductive) to fulfill
u σ rt 2π 0 2
i i the minimum requirement set by grid operators.
(19) • Step 5: 10% bus voltage magnitude limit and 100%
Similarly, loading of the lth series branch element is branches loading limit are set as network constraints.
calculated for all combination cases. • Step 6: Power flow calculation is performed with-
eI out DG connection at any bus and results are
l
 e Ilrt = (20) recorded.
Ilrt
• Step 7: Define the maximum number of DGs
The probability of the series branch element of violat- required to be connected with the grid. DGs are con-
ing the thermal limits is calculated using Equation (21). nected at all busbar according to the combinations
The thermal limit ilim is equal to 1.0 p.u. as defined by Equation (1). Power flow calculation is
performed over the range of DG output active power
1  e rt
d
th rt as defined in step 2. Results are recorded.
μl =  Il
d e=1 • Step 8: Step 6 and 7 is performed for DG output
power at 0.9 capacitive, 0.9 inductive, and 1.0 power
1  e rt th rt 2
d factor.
th rt
σl = ( Il − μl ) • Step 9: VI, PLI, GPI, and BSI are calculated for all
d
e=1 network buses.
⎛ 2 ⎞ • Step 10: Busbar and branches are identified for any
1 ilim 1 I − th μrt
th rt
hl = √ exp ⎝− th σ rt
l ⎠ dI violation. Furthermore, performance index are cal-
th σ rt 2π 0 2
l i culated to determine the network elements proba-
(21) bility of deviating from rated value.

3. Implementation of bus selection index in 4. Network analysis and simulation results

analytical technique
The proposed bus selection index is implemented and
The proposed distribution source integration method- analysed on the CIGRE medium voltage benchmark
ology is based on the analytical technique and deter- network and IEEE 33 bus benchmark network consid-
mines the suitable busbar location according to the ering both single DG integration as well as integrating
network capability. The presented algorithm identifies multiple DG simultaneously. The algorithm is imple-
the DG active power at which network will operate in mented in MATLAB using MATPOWER functions.
the most optimized manner. The solution also indicates The loads are considered constant at their most prob-
the maximum capacity of the DG that can be infeed able values for the power flow analysis, and generator-
into the network considering network constraints. Fur- oriented sign convention is used in the power flow
thermore, the proposed method is applicable for both results.
single or multiple DGs integration. Following are the
steps which algorithm performs to acquire the desired
4.1. DG integration analysis on CIGRE network
results:
The CIGRE medium voltage benchmark network is
• Step 1: The network model is built in the MAT- shown in Figure 2 and the network parameters are
POWER to perform the power flow computation. taken from Ref. [41]. The network is modelled as
Distribution generation unit is modelled as a con- meshed network and the power flow calculation with-
stant PQ source. out any DG in the network shows that the mini-
• Step 2: Active power range of distribution generation mum busbar voltage would be 0.95 p.u. The trans-
unit is set from zero to estimated maximum power formers (Br1 and Br2) have the maximum loading
that can be feed into the network. Any arbitrary i.e 72%. The active power loss is 0.109 MW and the
value can be set for the maximum power, however, it reactive power demand of the network branches is
 AUTOMATIKA 7

Figure 3. Bus selection index at 0.9 capacitive power factor.

Figure 4. Comparison of bus selection index at 15 MW DG

Figure 2. CIGRE MV benchmark network.
infeed power considering different power factors.

3.19 MVar. Moreover, the ‘Grid’ supplies the total power

of 34.29 MW and 7.14 MVar to fulfil the network
load demand.
The response without DG indicates that the busbar
voltages are lower than the rated values and network
demands capacitive reactive power. Thus, bus selection
index is calculated for the active power range from 0.0
to 20 MW at 0.9 capacitive power factor as shown in
Figure 3. The most suitable bus for the DG integra-
tion is at the lowest BSI value at any given DG active
power. Although bus ‘B1’, ‘B2’, and ‘B3’ appears suitable
for high DG active power output, the network would
not operate efficiently which means the network volt-
ages and loading are not improved and their values are
closed to ‘no DG’ response. DG placement other than
‘B1’, ‘B2’, ‘B3’ at high power would worsen the net- Figure 5. Individual index at 15 MW DG infeed power with 1.0
power factor.
work response as the losses in the network will increase
significantly. The average increase in the total network
losses would be up to 0.65 MW at 15.0 MW DG active that the active power losses are increased significantly
power infeed. The positive sign of the BSI index indi- regardless at which busbar DG is connected, and the
cates that the network response becomes worse com- improvement in the bus voltages and energy infeed into
pared to without DG connected condition as shown the grid is minimum.
in Figure 4. Figure 4 shows that the network cannot The maximum DG active power output is limited
be improved despite providing reactive power support by the constraints of the network. The response of at
by the DG. Furthermore, DG needs to be operated least one busbar voltage limit violation is shown in
near 1.0 power factor at 15.0 MW DG output power. Figure 6. Similarly, Figure 7 shows the response of at
The individual index, as shown in Figure 5, indicates least one branch loading limit violation. The responses
8 M. RAZA ET AL.

Figure 6. Voltage limit violation at 0.9 capacitive power factor. Figure 9. At least one branch loading limit violation compari-
son at different power factors.

Figure 7. Branch loading limit violation at 0.9 capacitive power

factor.
Figure 10. BSI response of Bus 7 at different power factors.

Figure 8. At least one busbar voltage limit violation compari-

son at different power factors. Figure 11. Individual index comparison at DG power 4.0 MW
and 0.9 capacitive power factor.

imply that the maximum output of the DG can only

be 13 MW if connected to bus ‘B15’ since at least considering all three indexes. The buses can be ranked
one busbar violate the limit. On the other hand, there in ascending order for DG integration suitability. At
will be no voltage limit violation up to 20 MW if 4.0 MW, the bus ‘B7’ is the most suitable candidate
DG is connected with the buses ‘B1’ to ‘B4’ ; how- to integrate DG while ‘B11’ would be next suitable
ever, branch loading limit will be violated at ‘B4’ for candidate.
DG active power 15 MW and above. The maximum The response of ‘B7’ BSI at different power factor
infeed power at different power factor can be deter- is shown in Figure 10. The response shows that the
mined by analysing Figures 8 and 9. The range of DG DG needs to be operated with capacitive power factor
active power output is increased from 20 to 40 MW in order to achieve optimized network operation while
in the simulation. Although the active power infeed inductive power factor will deteriorate the network per-
capability can be increased with respect to voltage limit formance. The comparison of the voltage, active power
by controlling DG reactive power, the loading of the loss, and grid power indexes at 4.0 MW DG power
branches cannot be improved significantly as shown in is shown in Figure 11. Primarily, active power loss
Figure 9. At bus ‘B3’ and ‘B4’ the loading of the trans- improved significantly when DG is connected at ‘B7’.
former would be the main limiting factor and the DG Further it also enables the maximum energy transfer to
power can be increased up to 32 MW, whereas DG the higher grid from the distribution generation unit.
connected at bus ‘B1’ (grid bus) provides no benefit Although the voltage improvement rank of ‘B7’ is 4.0,
in term of network voltages and active power losses its performance is comparable to the highest ranked
improvement. bus (‘B15’). The VI at ‘B15’ is −0.056 and VI at ‘B7’
The minimum BSI is at 4.0 MW as shown in Figure 3. is −0.050. The analysis concludes that the ‘B7’ is high-
This is the most suitable DG active power infeed loca- est ranked busbar based on BSI and most suitable for
tion to achieve the best-optimized network operation integrating the DG.
 AUTOMATIKA 9

In the above analysis, suitable busbar is determined

for the placement and sizing of a single DG. Similarly,
suitable location for multiple DGs integration can also
be determined using the proposed methodology. The
required number of DGs is the input parameter and
it can be set up to the maximum number of busbars
i.e 15 DGs on CIGRE network. However, for simplifi-
cation, the bus selection index and sizing comparison
analysis is performed up to 6 DGs integrated simulta-
neously. The minimum BSI of each busbar and the net
active power feed by the DGs is shown in Figure 12. The
plot shows the comparison of the busbars suitability for
integrating multiple DGs simultaneously. In case of 2
DGs connecting in the grid, the most suitable bus to
connect at least one DG is B3 and B2, and the net opti- Figure 12. Comparison of BSI and sizing at each busbar for
simultaneously connected DGs.
mium power of the 2 DGs could be between 8.84 and
9.33 MW. Bus selection index at B2 and B3 is minimum
since large DGs can be integrated on these buses com-
pare to other busbars, consequently more energy can be
feed into the grid. Similarly, improvement in the net-
work voltages and losses increase with the increase in
the number of DGs. With 6 DGs, network operational
level will be improved significantly while connecting
them on busbars from 4 to 15. Obviously, DGs con-
nection at busbar 1 to 3 do not provide significant
improvements in the network voltage and losses since
they are close to the grid feeding point. The net opti-
mum active power for 6 DGs integration is around
6.0 MW having each DGs approximately 1.0 MW of
capacity. Furthermore, it is also evident that the size of
the DGs integrated far from the grid feeding point hav- Figure 13. Comparison of at least one busbar voltage deviation
ing much greater power than nearby loads is not ben- for simultaneously connected DGs.
eficial. In Figure 13 network busbar voltage limits are
compared. The result indicates net DGs active power at
which at least one busbar voltage violates limit while the
corresponding busbar must be connected with a DG.
For instance, considering the case of 5 DGs integra-
tion at busbar 11 to 15, the net DG power should be
less then 15.0 MW while the maximum voltage viola-
tion at 15.0 MW would be 0.107 p.u i.e. 10.7%. Sim-
ilarly, the thermal loading limit violation results with
respect to the net DGs power is shown in Figure 14.
Further in this case, the first branch loading violation
would occur when the net DG power would exceed
20.0 MW, so in this case voltage violation parame-
ter will be given priority for selecting the maximum
size of the DG.
The algorithm is tested on the standard Intel (R) Figure 14. Comparison of at least one branch loading limit
violation for simultaneously connected DGs.
Core (TM) i9-10850K CPU@ 3.6 GHz based system.
The execution time of the algorithm for multiple DGs
integration analysis is shown in Table 1. Note that the 4.2. DGs integration analysis on IEEE 33 bus
total iteration does not include the power flow iteration, network
i.e Newton Raphson iteration. The algorithm provides a
deterministic solution, having fast execution time, and The proposed methodology is also analysed on the
covers wide range of DG power with sufficient step IEEE medium voltage 33 bus standard network as given
size. in Figure 15. The network is in radial configuration
10 M. RAZA ET AL.

Table 1. Computational time of the algorithm for a 15 bus CIGRE network.

No. of step change
No. of DGs in power No. of cases Total iteration Execution time (s)
1 101 15 1515 1
2 101 105 10,605 12
3 101 455 45,955 53
4 101 1365 137,865 162
5 101 3003 303,303 371
6 101 5005 505,505 634

Figure 15. IEEE medium voltage 33 bus standard network.

having 12.66 kV voltage level, and the parameters are plots with the BSI, it can be observed that the load-
taken from Ref. [42]. The cables resistance and reac- ing of the branches are violated at the lower DG size.
tance values in Ohms are given in the diagram, respec- For example, the BSI of B1, B2, and B19 is −3.276,
tively. Also, the current carrying capability of the cables −3.188 and −3.233, respectively, for 3 DGs simultane-
are sets as 0.3 kA. All the loads have lagging power ously integration. These BSI are achieved with net DG
factor. It is obvious that the capacitive reactive power size of 7.822 MW while the loading limits is violated at
infeed by the DGs would improve the network volt- 9.9201 MW for B1 & B2, and 6.6234 MW B19. Simi-
ages. Thus, the DG power factor is set as 0.90 leading larly, the busbar voltage limits is violated at 3.3267 MW
to acquire the maximum voltage support. The analysis while connecting DGs at B1, B2, and B19 thus the DGs
has been performed to integrate multiple DGs simul- with net active power of 7.822 MW cannot be con-
taneously and determines the suitable busbars and the nected due to voltage limit violation. In this case, the net
DG power. active power of 3 DGs should be less than 3.3267 MW.
Without any DG integration, the grid is required It is to be noted that the BSI is lowest despite the oper-
to supply the energy of 3.92 MW and 2.44 MVar. The ational limit violation, this implies that the network
maximum voltage drop is at B18 having 0.913 p.u volt- performance could be improved outside its boundary
age level. The total active power loss of the network is limits. The solution shown in Figure 17 said to be a
0.203 MW and the network cables requires 0.14 MVar global optimum point while BSI and DGs net active
inductive power. Further, the maximum active power power values defined by the voltage and loading limits
loss occur in cable ‘LN 2-3’ i.e 0.05 MW. are local minima. The BSI value at net DGs active power
Bus selection index up to 3 DG integration is shown values based on voltage limits is shown in Figure 19.
in Figure 16. It can be observed that the most suit- Considering the local minima point, B1 is suitable for
able bus for single DG integration is B2 and the size single DG integration, B3 and B23 are the most suit-
of the DG is 10.0 MW. B2 BSI is −4.3806 and the sec- able busbar for 2 DGs simultaneous integration, and
ond suitable bus is B1 with −4.3037 BSI. Figure 16 B23, B24, and B25 are suitable for 3 DGs simultaneous
shows the minimum BSI possible at each busbar with integration.
respective DG size while Figures 17 and 18 show at As mentioned previously, the BSI is based on three
least one limit violation of branch loading and busbar index and the indexes are the indicators of how much
voltage magnitude. By comparing the limit violation a voltage level and active power losses in the network
 AUTOMATIKA 11

Figure 16. Bus selection index of IEEE medium voltage 33 bus standard network for integrating multiple DGs.

Figure 17. Comparison of branches loading limit violation on IEEE medium voltage 33 standard network.

Figure 18. Comparison of voltage limit violation on IEEE medium voltage 33 standard network integrating multiple DGs.

can be improved as well as the energy that can be feed- been increased due to reverse energy flow while volt-
in by DGs into the grid after fulfilling the network load age level is improved and more energy is feed into the
demands. These indexes are non-conflicting in nature; upper-level grid.
however, the improvement in the each index varies It is also important to determine the global opti-
location-to-location. In certain case, one index might mum point of each individual index. This would help
be improving but others don’t. This usually occurs at to determine how much the individual indexes can be
the global optimum point since it is disregarding the improved in the network. The results shown in Fig-
operational limits. For example, the global optimum ures 20–22 quantify this analysis. For single DG place-
point at B19 for single DGs placement is −3.75153 ment, the most suitable busbar from voltage improve-
(BSI) as given in Figure 16 while VI is −0.0331, PLI ment perspective is B18 having VI value −0.2758 and
is equal to 1.688, and GPI value is −6.10647. It is the size of the DG would be 3.7063 MW. The second
clear that the active power losses in the network have most suitable busbar would B33 with −0.2626 VI value
12 M. RAZA ET AL.

Figure 19. BSI values at busbar voltage violation limit for IEEE medium voltage 33 standard network.

Figure 20. Optimum VI values at each busbar of IEEE medium

Figure 21. Optimum PLI values at each busbar of IEEE medium
voltage 33 standard network.
voltage 33 standard network.

and 4.106 MW DG active power. For 2 DGs simulta-

neous integration, B17 and B18 are most suitable and
DGs with net active power of 1.9081 MW can be con-
nected i.e 0.9540 MW each. For 3 DGs simultaneous
integration, B17, B18, and B33 are suitable with net
active power of 1.47 MW, i.e. approximately 0.5 MW
each. It can be observed that the size of an individual
DGs is reduced with the increase in the number of the
DGs connected simultaneously, and the improvement
in the network voltage would be lower.
Similarly, the suitable busbar from active power loss
improvement perspective is B26 for single DG inte-
gration having PLI −0.6756 with 3.416 MW capac-
Figure 22. Optimum GPI values at each busbar of IEEE medium
ity. B29 and B30 are suitable for 2 DGs integration voltage 33 standard network.
having net DG capacity of 1.5185 MW i.e. 0.76 MW
each. For 3 DGs integration, busbars B29, B30, and
B31 are suitable having PLI value of −0.7301, −0.738, grid which is evident from Figure 22 that B1 and B2 are
and −0.728, respectively. The optimum net DG power suitable for multiple DGs integration.
is 1.2188 MW. The analysis indicates that more dis- The proposed algorithm provides the deterministic
tributed generations are suitable for improving the net- solution and has the ability to analyse both global and
work active power losses. Furthermore, the DGs inte- local optimum points with fast execution time. The exe-
gration near to the feeder is suitable to maximize profit cution time of algorithm for the IEEE 33 bus medium
by feeding the energy back to the higher voltage level voltage standard network is given in Table 2. The total
 AUTOMATIKA 13

Table 2. Computational time of the algorithm for a 33 bus IEEE of the load and DG input power for simultaneously
network. integrating multiple DGs in the network.
No. of step
change in Total Execution
No. of DGs power No. of cases iteration time (s) Disclosure statement
1 101 33 3333 4.1 No potential conflict of interest was reported by the author(s).
2 101 528 53,328 65.29
3 101 5456 551,056 695.2
ORCID
Muhammad Raza http://orcid.org/0000-0002-4385-5020
number of iteration for single DG placement is 3333,
Aurangzeb Rashid Masud http://orcid.org/0000-0001-
which is executed in approx 4.1 s. Also, the execution 5225-2662
time is fast for 2 and 3 DGs simultaneous placement, i.e
65.29 and 695.2 s, respectively.
References
[1] Jensen L. The sustainable development goals report
5. Conclusion 2020. New York (NY): United Nation; 2020. (Tech.
Rep.).
The proposed bus selection methodology caters multi- [2] Adefarati T, Bansal R. Integration of renewable distri-
ple performance parameters for the selection of suitable buted generators into the distribution system: a review.
busbar to integrate multiple DGs with medium voltage IET Renew Power Gener. 2016 Aug;10(7):873–884.
network. The solution provides the maximum power [3] Blaabjerg F, Yang Y, Yang D, et al. Distributed power-
generation systems and protection. Proc IEEE. 2017
that a DG can infeed into the grid and the power at
Jul;105(7):1311–1331.
which network will operate in most optimized manner. [4] Olek B, Wierzbowski M. Local energy balancing and
The analysis shows that the global optimum point could ancillary services in low-voltage networks with dis-
exist outside the boundary limits, and the network can tributed generation, energy storage, and active loads.
be operated at this global point by making operational IEEE Trans Ind Electron. 2015 Apr;62(4):2499–2508.
limits less strict. In certain cases, global optimum point [5] Galvan E, Mandal P, Haque AU, et al. Optimal
placement of intermittent renewable energy resources
exists outside the boundary limits in multi-objective and energy storage system in smart power distri-
index when one or more index is maximizing while bution networks. Electr Power Compon Syst. 2017
others are minimizing. Further, the analysis shows that Aug;45(14):1543–1553.
more number of DGs in the network are suitable in all [6] Lee SH, Park JW. Optimal placement and sizing
the cases. Such as, one DG provides more optimized of multiple DGs in a practical distribution system
by considering power loss. IEEE Trans Ind Appl.
operation for network voltage improvement perspec-
2013;49(5):2262–2270.
tive and should be connected far from the feeders. [7] Radosavljevic J, Arsic N, Milovanovic M, et al. Optimal
While, more number of DGs distributed in the network placement and sizing of renewable distributed genera-
are suitable to improve the network losses however the tion using hybrid metaheuristic algorithm. J Mod Power
size of the DGs would be small. On the other hand, DG Syst Clean Energy. 2020;8(3):499–510.
of larger size should be connected near to the feeders [8] Hernández JC, Medina A, Jurado F. Impact comparison
of PV system integration into rural and urban feeders.
to be able to feed energy into upstream network with Energy Convers Manag. 2008;49(6):1747–1765.
ease. Furthermore, the utilization of converter in dis- [9] M’hamdi B, Teguar M, Tahar B. Optimal DG unit
tribution system enables the control of both active and placement and sizing in radial distribution network for
reactive powers thus the need of static VAR compen- power loss minimization and voltage stability enhance-
sator is reduced and the reactive power compensation ment. Period Polytech Electr Eng Comput Sci. 2020
Jan;64(2):157–169.
can be done through DGs. Therefore, DG sizing has
[10] Santos SF, Fitiwi DZ, Cruz MR, et al. Impacts of opti-
been done in the proposed algorithm considering both mal energy storage deployment and network reconfig-
active and reactive powers. It is obvious that the DG size uration on renewable integration level in distribution
is limited by the network operational limits, however, systems. Appl Energy. 2017 Jan;185:44–55.
the suitability of a busbar must be analysed considering [11] Emmanuel M, Rayudu R. The impact of single-phase
the number of network elements that are approaching grid-connected distributed photovoltaic systems on the
distribution network using P-Q and P-V models. Int J
the limit value. If more number of network elements Electr Power Energy Syst. 2017 Oct;91:20–33.
approach limit value, the busbar becomes less suit- [12] Watson JD, Watson NR, Santos-Martin D, et al. Impact
able. This methodology is implemented by introducing of solar photovoltaics on the low-voltage distribution
probabilistic-based penalty factor. Although the pro- network in New Zealand. IET Gener Transm Distrib.
posed index is implemented through analytical tech- 2016 Jan;10(1):1–9.
[13] Reza M, Schavemaker P, Slootweg J, et al. Impacts of
nique, the index can also be implemented as optimiza-
distributed generation penetration levels on power sys-
tion problem. The future extension of this work would tems transient stability. In: IEEE Power Engineering
be to implement the proposed bus selection index as Society General Meeting, 2004. Vol. 2. IEEE; 2004. p.
optimization problem considering the stochastic nature 2151–2156.
14 M. RAZA ET AL.

[14] Elphick S, Drury G, Fixter B. The impact of small scale improved spotted hyena optimizer. J Clean Prod. 2021
solar PV on power quality: an empirical study. In: 2020 Mar;286:124941.
19th International Conference on Harmonics and Qual- [29] Arasteh A, Alemi P, Beiraghi M. Optimal allocation of
ity of Power (ICHQP); 2020 Jul. IEEE. 2020. p. 1–6. photovoltaic/wind energy system in distribution net-
[15] Al Momani T, Harb A, Amoura F. Impact of photo- work using meta-heuristic algorithm. Appl Soft Com-
voltaic systems on voltage profile and power losses of put. 2021 Sep;109:107594.
distribution networks in Jordan. In: 2017 8th Interna- [30] Bazrafshan M, Gatsis N, Dallanese E. Placement and
tional Renewable Energy Congress (IREC). IEEE; 2017 sizing of inverter-based renewable systems in multi-
Mar. p. 1–6. phase distribution networks. IEEE Trans Power Syst.
[16] Baghban-Novin S, Mahzouni-Sani M, Hamidi A, et al. 2019 Mar;34(2):918–930.
Investigating the impacts of feeder reforming and dis- [31] El-Ela AAA, El-Sehiemy RA, Abbas AS. Optimal
tributed generation on reactive power demand of dis- placement and sizing of distributed generation and
tribution networks. Sustain Energy Grids Netw. 2020 capacitor banks in distribution systems using water
Jun;22:100350. cycle algorithm. IEEE Syst J. 2018 Dec;12(4):3629–
[17] Zhu D, Broadwater RP, Tam KS, et al. Impact of 3636.
DG placement on reliability and efficiency with time- [32] Beskirli M, Koc I, Haklı H, et al. A new optimization
varying loads. IEEE Trans Power Syst. 2006 Feb;21(1): algorithm for solving wind turbine placement problem:
419–427. binary artificial algae algorithm. Renew Energy. 2018
[18] Essallah S, Khedher A, Bouallegue A. Integration of dis- Jun;121:301–308.
tributed generation in electrical grid: optimal placement [33] Valencia A, Hincapie RA, Gallego RA. Optimal loca-
and sizing under different load conditions. Comput tion, selection, and operation of battery energy stor-
Electr Eng. 2019 Oct;79:106461. age systems and renewable distributed generation in
[19] Hung DQ, Mithulananthan N, Lee KY. Determining medium-low voltage distribution networks. J Energy
PV penetration for distribution systems with time- Storage. 2021 Feb;34(2020):102158.
varying load models. IEEE Trans Power Syst. 2014 [34] Fazliana Abdul Kadir A, Mohamed A, Shareef H, et al.
Nov;29(6):3048–3057. An improved gravitational search algorithm for opti-
[20] Zhang C, Li J, Zhang YJ, et al. Optimal location plan- mal placement and sizing of renewable distributed
ning of renewable distributed generation units in dis- generation units in a distribution system for power
tribution networks: an analytical approach. IEEE Trans quality enhancement. J Renew Sustain Energy. 2014
Power Syst. 2018 May;33(3):2742–2753. May;6(3):033112.
[21] Ehsan A, Yang Q. Optimal integration and planning of [35] Ibrahim K, Sirjani R, Shareef H. Performance assess-
renewable distributed generation in the power distribu- ment of pareto and non-pareto approaches for the opti-
tion networks: A review of analytical techniques. Appl mal allocation of DG and DSTATCOM in the distri-
Energy. 2018 Jan;210:44–59. bution system. Tehnicki Vjesnik. 2020 Oct;27(5):1654–
[22] Gözel T, Hocaoglu MH. An analytical method for the 1661.
sizing and siting of distributed generators in radial sys- [36] Ma C, Dasenbrock J, Töbermann JC, et al. A novel
tems. Electr Power Syst. 2009 Jun;79(6):912–918. indicator for evaluation of the impact of distributed gen-
[23] Hung DQ, Mithulananthan N, Lee KY. Optimal place- erations on the energy losses of low voltage distribution
ment of dispatchable and nondispatchable renewable grids. Appl Energy. 2019 May;242:674–683.
DG units in distribution networks for minimizing [37] Ishak R, Mohamed A, Abdalla AN, et al. Optimal place-
energy loss. Int J Electr Power Energy Syst. 2014 ment and sizing of distributed generators based on a
Feb;55:179–186. novel MPSI index. Int J Electr Power Energy Syst. 2014
[24] Selim A, Kamel S, Alghamdi AS, et al. Optimal place- Sep;60:389–398.
ment of DGs in distribution system using an improved [38] Memarzadeh G, Keynia F. A new index-based method
harris hawks optimizer based on single- and multi- for optimal DG placement in distribution networks.
objective approaches. IEEE Access. 2020;8:52815–52829. Eng Rep. 2020 Oct;2(10):1–15.
[25] Kizito R, Li X, Sun K, et al. Optimal distributed gen- [39] Hassan AS, Othman EA, Bendary FM, et al. Opti-
erator placement in utility-based microgrids during mal integration of distributed generation resources in
a large-scale grid disturbance. IEEE Access. 2020;8: active distribution networks for techno-economic ben-
21333–21344. efits. Energy Rep. 2020 Nov;6:3462–3471.
[26] Hung DQ, Mithulananthan N, Bansal R. Analytical [40] Li Z, Xu Y, Fang S, et al. Optimal placement of heteroge-
strategies for renewable distributed generation inte- neous distributed generators in a grid-connected multi-
gration considering energy loss minimization. Appl energy microgrid under uncertainties. IET Renew
Energy. 2013 May;105:75–85. Power Gener. 2019 Oct;13(14):2623–2633.
[27] Erdinc O, Tascikaraoglu A, Paterakis NG, et al. Compre- [41] Strunz K, Abbasi E, Abbey C, et al. Benchmark systems
hensive optimization model for sizing and siting of DG for network integration of renewable and distributed
units, EV charging stations, and energy storage systems. energy resources. CIGRE; 2014. (Tech. Rep.).
IEEE Trans Smart Grid. 2018 Jul;9(4):3871–3882. [42] Dolatabadi SH, Ghorbanian M, Siano P, et al. An
[28] Naderipour A, Nowdeh SA, Saftjani PB, et al. Deter- enhanced IEEE 33 bus benchmark test system for dis-
ministic and probabilistic multi-objective placement tribution system studies. IEEE Trans Power Syst. 2021
and sizing of wind renewable energy sources using May;36(3):2565–2572.