Applied Energy 165 (2016) 36–47

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Benefits analysis of Soft Open Points for electrical distribution network

operation
Wanyu Cao a, Jianzhong Wu a,⇑, Nick Jenkins a, Chengshan Wang b, Timothy Green c
a
Institute of Energy, School of Engineering, Cardiff University, Cardiff CF24 3AA, UK
b
Key Laboratory of Smart Grid of Ministry of Education, Tianjin University, Tianjin 300072, China
c
School of Engineering, Imperial College London, London, UK

h i g h l i g h t s

An analysis framework was developed to quantify the operational benefits.

The framework considers both network reconfiguration and SOP control.

Benefits were analyzed through both quantitative and sensitivity analysis.

a r t i c l e i n f o a b s t r a c t

Article history: Soft Open Points (SOPs) are power electronic devices installed in place of normally-open points in elec-
Received 31 August 2015 trical power distribution networks. They are able to provide active power flow control, reactive power
Received in revised form 5 December 2015 compensation and voltage regulation under normal network operating conditions, as well as fast fault
Accepted 8 December 2015
isolation and supply restoration under abnormal conditions. A steady state analysis framework was
Available online 30 December 2015
developed to quantify the operational benefits of a distribution network with SOPs under normal net-
work operating conditions. A generic power injection model was developed and used to determine the
Keywords:
optimal SOP operation using an improved Powell’s Direct Set method. Physical limits and power losses
Soft Open Point
Back-to-back converters
of the SOP device (based on back to back voltage-source converters) were considered in the model.
Distribution network Distribution network reconfiguration algorithms, with and without SOPs, were developed and used to
Network reconfiguration identify the benefits of using SOPs. Test results on a 33-bus distribution network compared the benefits
of using SOPs, traditional network reconfiguration and the combination of both. The results showed that
using only one SOP achieved a similar improvement in network operation compared to the case of using
network reconfiguration with all branches equipped with remotely controlled switches. A combination of
SOP control and network reconfiguration provided the optimal network operation.
Ó 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://
creativecommons.org/licenses/by/4.0/).

1. Introduction aiming at reliable and cost-efficient operation of distribution net-

works [3], e.g., network reconfiguration and increasing use of
The widespread use of distributed energy resources, e.g., dis- power electronic devices.
tributed generators (DG), energy storage and controllable loads, In a distribution network, there are usually a number of
is being promoted by many countries. This can lead to operation normally-open points connecting adjacent feeders. These
problems including excessive fault level as well as violations of normally-open points (switches) are able to be closed (while open-
thermal and voltage limits [1,2]. Power utilities have convention- ing other switches) to reconfigure the network and achieve load
ally used expensive and time consuming approaches such as build- transfer between feeders. Under normal operating conditions,
ing new circuits to maintain the quality of power supply. However, extensive research has been conducted into network reconfigura-
there are also alternative operational measures being investigated, tion for optimal network operation (e.g., load balancing, loss min-
imization and improved voltage profiles) [4,5]. However, practical
applications of automatic network reconfiguration are presently
⇑ Corresponding author. very limited due to the high cost of remotely-controlled switches,
E-mail addresses: [email protected] (W. Cao), [email protected] (J. Wu), the associated ICT (information and communication technologies)
[email protected] (N. Jenkins), [email protected] (C. Wang), t.green@
infrastructure, and maintenance of hardware/software.
imperial.ac.uk (T. Green).

http://dx.doi.org/10.1016/j.apenergy.2015.12.022
0306-2619/Ó 2016 The Authors. Published by Elsevier Ltd.
This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
 W. Cao et al. / Applied Energy 165 (2016) 36–47 37

An alternative solution to improve distribution network opera- nected feeders and supply or absorb reactive power at its interface
tion, without requiring network topology changes, is the use of terminals under normal operation conditions.
power electronic devices. Power electronic devices enable more
efficient use of existing network capacity by controlling power 2.1. Modeling of Soft Open Points
flows in an accurate and flexible way [6]. A wealth of information
exists on the use of these devices in the transmission network for A generic power injection model of SOP was developed. This
bulk power transfer [7]. Recently, installing power electronic model considers SOP terminal power injections and hence enables
devices in place of normally-open points in a distribution network, straightforward incorporation of SOPs into existing power flow
namely ‘Soft’ Open Points (SOPs), has been investigated [8,9]. analysis algorithms without considering the detailed controller
Instead of simply opening/closing normally-open points, these design.
devices are able to control load transfer and optimize network Fig. 1b shows the representation of an SOP model with real and
voltage profile by providing fast, dynamic and continuous real/ reactive power, injecting into feeders I and J through both termi-
reactive power flow control between feeders [8]. nals. Taking these power injections as decision variables, the
Previous studies have investigated the use of power electronic power flow in feeder I is calculated by the following set of recursive
devices at the normally-open points to facilitate distribution net- equations [12]:
work operation [10,11]. In [10], a unified power flow controller

r i1
was developed to regulate network voltage with minimum line Pi ¼ Pi1  P Lossði1;iÞ  PL;i ¼ Pi1  2
 P2i1 þ Q 2i1  PL;i
losses in a simple loop network (with two feeders), and experimen- jV i1 j
tal results were presented to verify its effectiveness. Field tests of ð1:iÞ
installing back-to-back converters between adjacent feeders were
xi1

reported in [11]. Power flow was balanced which in turn led to
Q i ¼ Q i1  Q Lossði1;iÞ  Q L;i ¼ Q i1  2
 P2i1 þ Q 2i1  Q L;i
reduced line losses and improved network voltages. Although the jV i1 j
benefits of installing individual SOP for network operation have ð1:iiÞ
been investigated in a simple two-feeder network together with
ðr2i1 þ x2i1 Þ
2 
the controller design and simulation, methodologies for benefit
quantification, i.e., steady state analysis of distribution networks jV i j2 ¼ jV i1 j2  2  ðr i1 Pi1 þ xi Q i Þ þ 2
 Pi1 þ Q 2i1
with SOPs were not addressed and the advantages of the more jV i1 j
widespread use of these devices in distribution networks have ð1:iiiÞ
not been explored.
with boundary conditions:
To fill this gap, a method to quantify the operational benefits of
 
a distribution network with SOPs was developed, for power loss  I 
Pn þ PS inj  ¼ PLossðn;sIÞ ð2Þ
minimization, feeder load balancing and voltage profile improve-
ment. A generic model of an SOP for steady state analysis was  
 I 
developed, which takes into account both physical limitations Q n þ Q S inj  ¼ Q Lossðn;sIÞ ð3Þ
and internal power losses of the back-to-back voltage-source con-
verters of a typical SOP device. Based on the SOP model, an where P represents active power, Q reactive power, V nodal voltage,
improved Powell’s Direct Set method was developed to obtain r resistance and x reactance. Subscript Loss denotes line losses, and L
the optimal SOP operation. This method determines a good initial denotes load. The variables and parameters are shown in Fig. 1b.
approximation of the SOP operation based on simplified power Similar recursive power flow equations with boundary conditions
flow equations, which significantly reduces the computation bur- are applied to feeder J.
den. The performance of traditional network reconfiguration was To consider the internal losses of the SOP equipment, the fol-
compared to using SOPs. A method that combines SOP control lowing equality constraint of power balance is used:
and network reconfiguration was developed to identify the bene-
fits. In addition, the benefits of using SOPs in distribution networks PIS inj þ PJS inj þ P SOP;Loss ¼ 0 ð4Þ
with DG connections were also investigated.
where PSOP;Loss denotes the internal power losses of the whole SOP
device.
2. Steady state analysis of Soft Open Points Various types of power electronic devices can be implemented
as an SOP, such as unified power-flow controllers, back to back
Fig. 1a shows a typical location of an SOP which allows the and multi-terminal voltage-source converters [9]. In this paper,
power electronic device to control active power flow between con- the back-to-back voltage-source converters (back-to-back) were

(a) (b)

SOP

Fig. 1. (a) Simple distribution network with an SOP and (b) power injection model of SOP for distribution network power flow control.
38 W. Cao et al. / Applied Energy 165 (2016) 36–47

used as an SOP. Constraints of such device due to the physical lim-

itations and the internal power losses are formulated as follows.

2.1.1. Physical limitations of back-to-back converters

Back-to-back voltage-source converters consist of two insulated
gate bipolar transistor (IGBT)-based voltage-source converters
(VSCs). The two VSCs are series-connected on their DC sides, i.e.,
sharing a common capacitor, as shown in Fig. 1b. Both VSCs build
their own voltage waveforms with desired amplitude and phase
angle. This, in turn, gives a full (four-quadrant) control of the active
power flowing through the dc link as well as independent reactive
Fig. 2. Comparison between the quadratic and approximate loss estimation
power supply or absorption at both interface terminals. A full
function.
description of the back-to-back converters properties can be found
in [13].
The terminal power injections of back-to-back converters Comparing the internal power losses derived from the two
implemented as a SOP are controlled directly by each VSC, the models, as shown in Fig. 2, it is noted that when IIVSC 6 IIVSC;rate ,
operational limits of VSC capacity and terminal voltage are consid-
the device power losses using the approximate model are higher
ered as:
than those using the quadratic one, i.e., more conservative conclu-
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 2 sions are drawn. The approximate model was used in this paper.
PIS inj þ Q IS inj 6 SIVSC;rate ð5Þ
2.2. Optimal operation of Soft Open Points
V sI 6 V IVSC;rate ð6Þ
The benefits of SOPs for both system loss reduction and feeder
where SIVSC;rate is the power rating of the VSC connected to feeder I; load balancing were investigated. To quantify these benefits, the
amounts of real and reactive power injections of the SOPs were
V sI and V IVSC;rate denote the actual and maximum ac terminal voltage.
determined through solving a combinational nonlinear con-
Similar capacity and terminal voltage constraints are applied to the
strained optimization problem.
other side VSC connected to feeder J.
2.2.1. Problem formulation
2.1.2. Internal power losses of back-to-back converters Two objective functions are used separately to quantify the
Internal power losses of the whole back-to-back device are optimality of the SOP operation.
made up of the losses of its individual components, including the
semiconductors (conduction and switching losses), passive compo-
2.2.1.1. Active power loss minimization. The active power losses con-
nents (DC link capacitor, filter, AC line choke), transformers and the
sist of two components: the feeder losses and the internal power
cooling system.
losses of the SOP devices. The active power loss minimization
As suggested in [14], these losses can be categorized into three
problem is formulated as:
components: no load losses, linear and quadratic losses depending
on the converter current, which is a function of the active and reac- X
nl
P2k þ Q 2k
tive power exchanged with the ac network: Minimize PT;Loss ¼ rk  þ P SOP;Loss ð11Þ
k¼1 jV k j2
2 I
PIVSC;Loss ¼ aI  IIVSC þ b  IIVSC þ cI ð7Þ where r k ; Pk ; Q k ; V k are resistance, active power, reactive power, and
voltage of branch k; and nl is the total number of branches in the
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi,
2 2 network.
IIVSC ¼ PIS inj þ Q IS inj jV sI j ð8Þ
2.2.1.2. Feeder load balancing. A branch loading index LIk is used to
where PIVSC;Loss and IIVSC represent the power losses and the ac current measure the loading level of each branch in the network, which is
expressed as:
of the VSC connecting to feeder I. Similar equations are applied to
the other side VSC. Thus all equations shown in the following sec-  2
Ik
tion only illustrate the VSC connected to feeder I. LIk ¼ 8 k 2 nl ð12Þ
Ik;rate
As shown in (7), the total losses are determined by the coeffi-
I where Ik and Ik;rate are the actual and the rated branch current of
cients, aI ; b and cI . These coefficients are difficult to obtain due
to the limited information available from the open literature or branch k.
manufacturers. Usually only the internal power losses (or effi- Feeder load balancing is achieved by minimizing the load bal-
ciency) under nominal conditions and the no load loss, cI are given. ance index LBI, which is defined as the sum of branch load balanc-
A linear function is used to approximate the quadratic one in ing indices, as shown in (13):
(7): X
nl

I
Minimize LBI ¼ LIk ð13Þ
PIVSC;Loss ¼ k  IIVSC þ cI ð9Þ k¼1

.
I
k ¼ P IVSC;Loss;rate  cI IIVSC;rate ð10Þ 2.2.1.3. Constraints. Constraints of the distribution network and the
SOP devices are considered. The device constraints of each SOP are
where P IVSC;Loss;rate and IIVSC;rate are power losses and ac current of the shown in (4)–(6). The network constraints are expressed as
VSC connected to feeder I under nominal condition. follows:
 W. Cao et al. / Applied Energy 165 (2016) 36–47 39

gðx; sÞ ¼ 0 ð14Þ N-dimensional quadratic problem [15]. The efficacy has also
been demonstrated for any other function form, as pre-
jV i;min j 6 jV i j 6 jV i;max j8 i 2 nb ð15Þ sented in [16,17], but may require more iterations.
(iii) The mutually conjugate directions are generated after each
jSk j 6 jSk;max j 8 k 2 nl ð16Þ iteration, as illustrated in the following part.
where gðx; sÞ; V i;min ; V i;max and Sk;max are the power flow equations,
the minimum and the maximum voltage of bus i and the maximum
capacity allowed in branch k. nb are the total number of buses of the 2.2.2.2. PDS method for optimal SOP operation. The process of deter-
network. mining optimal SOP operation using the PDS method is shown in
Fig. 3:
2.2.1.4. Penalty function. Constrains of mentioned above are
included into the objective function by using the penalty function Step 1: Initialization. Based on (19), the initial approximate SOP
ð1Þ
method. In this way, unconstrained optimization methods are used operation S 0 , the initial search direction set with 3n linear
ð1Þ ð1Þ ð1Þ
directly by solving the transformed objective function (with pen- independent search directions, fngð1Þ ¼ fn1 ; . . . ; ni ; . . . ; n3n g
alty terms): and the convergence criterion e are specified. The 3n directions
Minimize F obj ðSÞ ¼ fPT;Loss or LBIg þ k1  f SOP þ k2  f V þ k3  f S ð17Þ in fngð1Þ are initially chosen to be the co-ordinate directions
(linear independent). This means only one decision variable in
where k1 ; k2 ; k3 are the penalty constants; f SOP ; f V ; and f S are the (19) will be changed when searching along one direction.
penalty functions for violations of the SOP device constraints, the Step 2: Generate a new mutually conjugate direction within one
network voltage and capacity constraints; S denotes the decision iteration. There are two sub steps:
variable vector. ðkÞ
– starting from S 0 , k indicates the iteration number (one itera-
Based on the power injection model of SOP in Section 2.1, the
tion includes searching along 3n directions, k = 1 initially),
real and reactive power injections at both terminals are specified ðkÞ
as the decision variables: with one SOP installation, sequentially find S i which gives the minimum of the objec-
h iT tive function (17) along each search direction, ni
ðkÞ
in fngðkÞ ,
S ¼ PISinj ; Q ISinj ; Q JSinj . PJS inj is not included in S since it is deter-
which is expressed as:
mined directly after S is specified according to the equality con-
straint equation of (4). Therefore, the coordinates of n SOPs are



ðkÞ ðkÞ ðkÞ
Min F obj S i ¼ F obj S i1 þ ki  ni ; i ¼ 1; 2; . . . ; 3n
combined together sequentially to obtain a point of S in a 3n-
dimentional space, thus ð19Þ
h iT In this paper, a one-dimensional search method, the Golden
S ¼ P IS inj;1 ;Q IS inj;1 ;Q JS inj;1 ;P IS inj;2 ;Q IS inj;2 ;Q JS inj;2 ;...; P ISinj ;n ;Q ISinj ;n ;Q JSinj ;n Ratio Rule method [19], was adopted to calculate the optimal
ð18Þ step size, ki .
The load flow analysis method introduced in [5] combined
with the SOP injection model formulated in (1)–(4) was
2.2.2. Method of determining optimal SOP operation implemented as a subroutine to calculate the objective func-
ðkÞ
Based on the Powell’s Direct Set (PDS) method presented in tion F obj ðS i Þ in (19).
[15], an improved PDS method to optimize the SOP operation
was developed. The performance improvement is achieved by – after searching down the 3n directions, a conjugate direction is
obtaining a good initial approximated SOP operation. generated by (20)

2.2.2.1. Powell’s direct set method. Most mathematical optimization ðkÞ ðkÞ ðkÞ
nconju: ¼ S 3n  S 0 ð20Þ
methods require explicit expressions of the derivatives to define
the direction of movement, i.e. search direction approaching to
the optimum (from a starting point). Step 3: Update the search direction set for the next iteration
The Powell’s Direct Set method is a direct search method pro- fngðkþ1Þ . Two scenarios are considered:
posed by Powell [15]. It defines the search directions in a direct ðkÞ ðkÞ ðkÞ
– If n = 1, discard the first direction n1 in fni g by adding nconju:
manner, i.e., solely depending on the objective function itself.
to the end, as shown in (21)
Therefore, this method is easy to implement and is not limited
by the existence of derivatives of the objective function. It has been n o
ðkÞ ðkÞ ðkÞ ðkÞ
successfully applied to solve problems for which it is difficult or fngðkþ1Þ ¼ n2 ; n3 ; . . . ; n3n ; nconju: ð21Þ
impossible to calculate the derivatives [16–18].
A comprehensive review of PDS method as well as the mathe-
– If n > 1, a ‘smarter’ updating procedure is required to ensure a
matical proof of its convergence were given in [15]. Three key
reasonable rate of convergence, i.e., replace the direction in
properties of this method are highlighted below:
fngðkÞ , which shows the worst performance. The general proce-
(i) For an N-dimensional problem, minimization is achieved by dure is presented as follows [15]:
an iterative procedure that searches down N linear indepen-
Find nðkÞ
m that gives maximum reduction among the previous
dent directions within each iteration, i.e. starting from the one-dimensional searching processes in Step 2, as shown in
best known approximation to the optimum. (22):
(ii) Fast convergence to the optimum is achievable by only n

o
ðkÞ
searching down N mutually conjugate directions. The opti- Dmax ¼ max F obj S m1  F obj S ðkÞ
m ð22Þ
16m63n
mal solution of a quadratic function has been proved achiev-
ðkÞ
able by searching along those mutually conjugate directions
Replace nðkÞ
m by nconju: giving more efficient convergence if the
once only. Hence only N iterations are required to solve the following two criterions are satisfied
40 W. Cao et al. / Applied Energy 165 (2016) 36–47

(1) Initialization

(2) Generate a new mutually conjugate direction ƺConju.

(3) Update the search direction set {ƺ} ?

No
Yes

(4) Find the minimum of Fobj along the conjugate direction ƺ Conju.

(5)Stopping rule satisfied?

No
Yes
(6) Output the results:
Fobj and SOP operation S

(7) End

Fig. 3. Flow chart of the proposed PDS method.

)
ðw1  2w2 þ w3 Þðw1  w2  Dmax Þ2 < 0:5  Dmax ðw1  w3 Þ2
w3 < w1
ð23Þ




where w1 ¼ F obj S ðkÞ
0
ðkÞ
, w2 ¼ F obj S 3n
ðkÞ ðkÞ
, w3 ¼ F obj 2  S 3n  S 0 . Thus,
n o
ðkÞ ðkÞ ðkÞ ðkÞ ðkÞ
fngðkþ1Þ ¼ n1 ; . . . ; nm1 ; nconju: ; nmþ1 ; . . . ; n3n ð24Þ

Otherwise fngðkÞ is not updated in this iteration, let

ðkþ1Þ ðkÞ
S0 ¼ S 3n and jump to Step 5.

An example shown in Fig. 4 illustrates the searching of the opti-

mization process. As the red1 solid line shows, the optimization
ð1Þ
starts from an initial approximation S 0 and then reaches the mini-
ð1Þ ð1Þ ð1Þ
mum S1 ; S2 ; S3 along each of the coordinate directions in fngð1Þ . Fig. 4. Example of one SOP optimization by PDS method.
ð1Þ
The first conjugate direction nconju: is hence obtained, as shown by
the blue dotted arrow. After searching along the conjugate direction, (ii) Ignoring all the quadratic terms (line losses) from the power
ð2Þ flow equations in (1). The branch power P k and Q k are
a new approximation S 0 is obtained for the second iteration.
obtained by summing up the downstream power loads:

2.2.2.3. Determine the initial approximated SOP operation. The PDS X

n

method is able to start from any initial point A good approxi-

ð1Þ
S0 .
Pk  PL;i ð27Þ
i¼kþ1
mation of SOP operation is obtained by running the PDS method
once using the objective function of (19) with simplified power
flow equations given by [5]: X
n
Qk  Q L;i ð28Þ
(i) All nodal voltage magnitudes are assumed to be 1 p.u., Eqs. i¼kþ1

(11) and (13) are reduced to The constraints of nodal voltage and branch capacity are consid-
ered by:
X
nl

PLoss  r k  P2k þ Q 2k þ PSOP;Loss ð25Þ
k¼1
X
i
jV i j2  jV 0 j2  2 ðr k Pk þ xk Q k Þ 6 V 2max ð29Þ
k¼1
X
nl
P 2k þ Q 2k
LIB  ð26Þ
k¼1 I2k;rate

Sk  P 2k þ Q 2k 6 S2k;max ð30Þ

Using the simplified equations, as shown in (25)–(30), the initial

1
For interpretation of color in Fig. 4, the reader is referred to the web version of approximated SOP operation is obtained directly without using the
this article. accurate load flow calculations.
 W. Cao et al. / Applied Energy 165 (2016) 36–47 41

2.3. Network reconfiguration considering SOPs total real and reactive power loads are 3715 kW and 2300 kVar. Four
normally-open switches, i.e., the switches between buses 25 and 29,
Reconfiguring distribution networks is used to achieve better 33 and 18, 8 and 21, 12 and 22, are chosen as candidate places for SOP
network operation including power loss minimization, feeder load installation. The capacity limit of each SOP unit is 3 MVA.
balancing and supply restoration. To investigate the performance Four cases were defined and used for quantifying the benefits of
of distribution network reconfiguration when SOPs are installed SOPs for improvement of network performance:
to replace some of the normally-open points, the proposed PDS
method for optimal SOP operation was combined with the network Case I: Improve network performance using SOPs.
reconfiguration method introduced in [20]. Case II: Improve network performance considering both SOPs
The procedure of the combined algorithm is given in Fig. 5. and network reconfiguration.
According to [20], a shortest-path algorithm is used to find the Case III: Impact of DG connections.
optimal electricity supply path for each load busbar. A genetic Case IV: Impact of power losses of SOP devices.
algorithm (GA) with the selection, crossover and mutation opera-
tors is used to optimize the sequence of load busbars searching 3.1. Improve network performances using SOPs
for the supply paths because the sequence affects the obtained net-
work configurations. The improved PDS method for optimal SOP The impact of different number of SOPs installed in the network
operation is integrated after the shortest path algorithm. The fit- on both power loss minimization and feeder load balancing was
ness functions utilized are formulated in Section 2.2. investigated. The device power losses were ignored for this case.

3. Case study 3.1.1. Power loss minimization

Fig. 7a shows the results of minimized power losses with 1–4
A 33-bus distribution network, as shown in Fig. 6, was used for SOPs installed in the network. There was a significant reduction
case study [5]. This network has 32 normally-closed switches, 5 (by 42%) of the system power loss with one SOP installation. With
normally-open switches and the nominal voltage is 12.66 kV. The more SOPs installed in the system, the total power losses were

Start

Determine the sequence of load busbars to search supply paths

Find the optimal network configuration using a shortest path algorithm

GA algorithm with
selection,
Determine optimal SNOP operation using the PDS method crossover and
mutation

Calculate the combined results as fitness values

No
Stopping criteria is satisfied?
Yes

Output optimization results

End

Fig. 5. Flowchart of proposed network reconfiguration considering SOPs.

Fig. 6. 33-bus distribution network.

42 W. Cao et al. / Applied Energy 165 (2016) 36–47

Effect of Number of SOP Installation on Power Loss

(a) 240
No. SOP Locations (Branches)
1 25-29

Total Power Losses (kW)

180 2 25-29 18-33
3 25-29 18-33 12-22
4 25-29 18-33 12-22 8-21
120

60

0
0 1 2 3 4
Number of SOP

Impact of Number of SOP Installation on System Voltage Profile

(b) 1

0.98
Voltage (p.u.)

0.96

0.94

0.92

0.9
0 SOP 1 SOP 2 SOP 3 SOP 4 SOP
0.88
1 4 7 10 13 16 19 22 25 28 31
Bus Number

Fig. 7. Impact of different number of SOP installation on power loss minimization and voltage profile improvement.

Impact of Number of SOP Installation on Load Balancing

(a) 1.2
0 SOP 1 SOP 2 SOP
1 3 SOP 4 SOP
Branch Loading

0.8

0.6

0.4

0.2

0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
Branch Number

Impact of Number of SOP Installation on System Voltage Profile

(b) 1.00
0.98
0.96
Voltage (p.u.)

0.94
0.92
0.90
0.88 0 SOP 1 SOP 2 SOP
3 SOP 4 SOP
0.86
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33
Bus Number

Fig. 8. Impact of different number of SOP installation on load balancing and voltage profile improvement.
 W. Cao et al. / Applied Energy 165 (2016) 36–47 43

Table 1 as shown in Fig. 8b. The minimum bus voltage (at bus 32) was 0.88
System load balancing index with different number of SOP installation. p.u. due to overloading, and it was improved by 9.09% after the sys-
Number of SOP installed 0 1 2 3 4 tem loading was balanced using SOPs.
SOP locations (branches) – 25–29 25–29 25–29 25–29 The results also show that beyond two SOPs, the benefits of loss
18–33 18–33 18–33 minimization, load balancing and voltage regulation are not
12–22 12–22 increased significantly. Therefore, the proposed method can be
8–21 used to derive the optimum amount of SOPs for a given network.
System load balancing index 6.156 3.218 2.566 2.428 2.389
% LBI reduction – 47.726 58.317 60.056 61.192
3.1.3. Performance of the improved PDS method for optimal SOP
operation
The effectiveness of the proposed method with a good initial
further reduced (by 57%). The voltage profiles of the system were approximation was evaluated in this case study. Table 2 lists the
also improved with the SOP installations, as shown in Fig. 7b. computation time required for the calculation of above case studies
considering only one SOP. The total time required by using the con-
3.1.2. Feeder load balancing ventional PDS method (starting from an arbitrary point) and the pro-
One feeder of the network from branch L25 to branch L32 as posed method with a good initial approximation were compared. It
shown in Fig. 8 was assumed to be heavily loaded, i.e., 1.6 times can be seen that for both power loss minimization and feeder load
higher than the loading under the normal condition. Table 1 shows balancing there were significant reductions in computation time
the impact of different number of SOPs on feeder load balancing. It by using the improved PDS method. Especially for solving the feeder
is observed that the system load balancing index LBI was reduced load balancing, the total CPU time required was reduced by 68%
by 47.73% with one SOP. Although the LBI reduction was further (from 8.295 s to 2.614 s) after using the improved PDS method.
improved with more SOPs installed, the rate of improvement was
diminishing. Fig. 8a illustrates the branch loading profile of the 3.2. Improve network performance considering both SOP and network
network. By using SOPs, the loading of those heavily loaded reconfiguration
branches (e.g. branches L1 to L6 and L25 to L30) was reduced dra-
matically by transferring loads to the lightly loaded branches via The benefits of combining SOP and network reconfiguration for
SOP control. As a consequence, the loading levels from branches power loss minimization and feeder load balancing were evalu-
L12 to L24 were increased. The voltage profile was also improved, ated. In this case, the SOP device was assumed to be located

Table 2
Results of computing time for optimal SOP operation.

Optimization technique Power loss minimization Feeder load balancing

PDS Improved PDS PDS Improved PDS
CPU time (s) Light Loading 4.565 1.998 Overload condition 8.295 2.614
Normal Loading 5.364 3.343
Heavy Loading 5.774 3.578

Table 3
Results of different methods for power loss minimization.

Case studies Load Level

Light (50%) Normal (100%) Heavy (160%)
Base case Switches with open status 8–21 12–22 25–29 9–15 18–33
No reconfiguration Power loss (kW) 47.118 202.876 575.966
No SOP installed Minimum voltage (p.u.) 0.958 0.913 0.853
Only network reconfiguration Switches opened 7–8 9–10 14–15 25–29 32–33
Power loss (kW) 33.312 137.946 381.418
% Loss reduction 29.301 32.005 33.778
Minimum voltage (p.u.) 0.970 0.938 0.897
Only SOP Switches with open status 8–21 12–22 9–15 18–33
SOP operation P 1S (MW) 0.230 0.605 0.998
inj

Q 1S (MVar) 0.225 0.471 0.784

inj

Q 2S inj (MVar) 0.610 1.239 2.017

Power loss (kW) 29.774 124.456 337.525
% Loss reduction 36.810 38.653 41.398
Minimum voltage (p.u.) 0.967 0.933 0.890
Combined network reconfiguration with SOP Switches with open status 7–8 9–10 14–15 18–33
SOP operation P 1S (MW) 0.183 0.374 0.607
inj

Q 1S inj (MVar) 0.215 0.424 0.697

Q 2S inj (MVar) 0.516 1.045 1.686
Power loss (kW) 22.758 93.915 250.179
% Loss reduction 51.700 53.708 56.564
Minimum voltage (p.u.) 0.978 0.955 0.925
44 W. Cao et al. / Applied Energy 165 (2016) 36–47

Volatage Profile of 33-bus Distribution Netwo at Normal Load

1

0.98

Voltage (p.u.)
0.96

0.94

0.92

0.9 Base Case Only Reconfiguration

Only SOP Combined Method
0.88
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33
Bus Number

Fig. 9. Voltage profiles of the network under normal loading condition.

Table 4
Results of different methods for load balancing.

Case studies Base case Only network reconfiguration Only SOP Combined reconfiguration with SOP
System load balancing index 6.156 4.139 3.218 2.594
% LBI reduction – 32.765 47.726 68.337
Minimum voltage (p.u.) 0.880 0.904 0.928 0.943
SOP operation P 1S inj (MW) – – 0.916 1.142
Q 1S (MVar) – – 0.569 0.577
inj

Q 2S inj (MVar) – – 1.791 2.417

between buses 25 and 29 and its power losses were ignored. Three and heavy (160%). The simulation results are listed in Table 3. The
case studies were also carried out for comparisons, which are base percentages of total power loss reduction implies that using only
case study with neither reconfiguration nor SOP; case study con- one SOP achieved a similar power loss reduction to that of network
sidering only network reconfiguration; and case study considering reconfiguration under three loading conditions. The most signifi-
only one SOP, which is located between buses 25 and 29. cant power loss reductions and voltage improvement under all
three loading conditions were obtained using the combined
method. The SOP operation required to achieve power loss mini-
3.2.1. Power loss minimization
mization indicates that the combined method contributed more
The network performance on power loss minimization was sim-
to power loss reduction while requiring smaller SOP sizes. Fig. 9
ulated under three loading conditions: light (50%), normal (100%),

Branch Loading Using Different Methods for Load Balancing

(a) 1.2 Base Case Only Reconfiguration
Only SOP Combined Method
0.9
Branch Loading

0.6

0.3

0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
Branch Number

Volatage Profile of 33-bus Distribution Network

(b)
0.99
0.97
Voltage (p.u.)

0.95
0.93
0.91
0.89 Base Case Only reconfiguration
0.87 Only SOP Combined Method
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33
Bus Number

Fig. 10. Results of load balancing capability and relevant voltage profile improvement under different methods.
 W. Cao et al. / Applied Energy 165 (2016) 36–47 45

Table 5
Results of different methods for power loss minimization with DG connections.

Case studies Base case without DGs Base case with DGs Only network reconfiguration Only SOP Combined reconfiguration with SOP
Total power losses (kW) 202.875 345.502 120.783 92.334 63.221
% Loss reduction – – 65.044 73.275 81.702
SOP operation P 1S inj (MW) – – – 1.661 1.606
Q 1S (MVar) – – – 0.376 0.381
inj

Q 2S (MVar) – – – 0.899 0.904

inj

Volatage Profile of 33-bus Distribution Network with DG connections

1.10
1.08
1.06
Voltage (p.u.)

1.04
1.02
1.00
0.98
0.96 Base Case with DG connection Only reconfiguration
0.94 Only SOP Combined Method
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33
Bus Number

Fig. 11. Voltage profiles of the network under different cases with DG connections.

illustrates the voltage profiles of all case studies under the normal case study using only network reconfiguration; case study using
loading condition. The shapes of the voltage profile under the other only one SOP; and case study using the combined network recon-
two loading condition were the same except minor change in mag- figuration and SOP.
nitude, and hence are not illustrated in the paper.

3.3.1. Power loss minimization

3.2.2. Feeder load balancing
Table 5 shows that network power losses for the five case stud-
The same overloading condition described in the previous case
ies under normal loading conditions. It can be seen that the net-
study was simulated. Table 4 presents the load balancing index for
work power losses were increased by 70% due to the DG
the four case studies. The branch loading profiles and voltage pro-
connections. A significant power loss reduction (72%) was achieved
files are shown in Fig. 10. From these results, it is seen that by using
by using only one SOP, which gave better performance than by
only one SOP to control network power flows achieved better per-
using only network reconfiguration. The most significant power
formance on load balancing than that of using network reconfigu-
loss reduction was obtained by using the combined method while
ration to change the open/close status of a sequence of switches.
requiring smaller SOP size. Fig. 11 shows the voltage profile of the
The highest LBI reduction, i.e., the most well-balanced network,
network. There was a notable voltage rise due to the DG connec-
and the highest improvement on minimum voltage were achieved
tions. By using the SOP (either with or without network reconfigu-
using the combined method.
ration) achieved much flattered network voltage profile.

3.3. Impact of DG connections

3.3.2. Feeder load balancing
A large capacity of intermittent renewable energy in the distri- To evaluate the benefits of the SOP for feeder load balancing
bution network can increase feeder loads unbalance and network with DG connections, especially when one feeder has high DG
power losses. To evaluate the benefits of using the SOP for power power generation while the other one is under a heavy loading
loss minimization and feeder load balancing with DG connections, condition, the same overloading condition adopted previously
three DGs were assumed to be connected to buses 16, 17 and 18. was used again. Table 6 shows the load balancing index for the five
Each DG had a power production of 1 MW with a unity power fac- case studies. It is observed that the network was more unbalanced
tor and was modeled as a negative load in this study. An SOP was (higher LBI) after connecting DGs. A significant reduction on LBI
assumed to be located between buses 18 and 33 and its power (77.51%) was achieved by transferring loads from the heavily
losses were ignored. Five case studies were carried out for compar- loaded feeder to that with DG integration via the SOP, which gave
isons, which are base case study with and without DG connections; much better performance than that by changing the open/close

Table 6
Results of different methods for load balancing with DG connections.

Case studies Base case without DGs Base case with DGs Only network reconfiguration Only SOP Combined reconfiguration with SOP
System load balancing index 6.156 7.238 2.995 1.628 1.267
% LBI reduction – – 58.621 77.507 82.495
SOP operation P 1S inj (MW) – – – 1.692 1.480
Q 1S inj (MVar) – – – 0.457 0.523
Q 2S (MVar) – – – 1.252 1.211
inj
46 W. Cao et al. / Applied Energy 165 (2016) 36–47

Branch Loading Using Different Methods for Load Balancing with DG connections
(a) 1
Base Case with DGs Only Reconfiguration
0.8 Only SOP Combined Method

Branch Loading
0.6

0.4

0.2

0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
Branch Number

Volatage Profile of 33-bus Distribution Network with DG connections

(b) 1.10
1.08
1.06
Voltage (p.u.)

1.04
1.02
1.00
0.98
0.96
0.94
0.92 Base Case with DGs Only reconfiguration
Only SOP Combined Method
0.90
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33
Bus Number

Fig. 12. Results of load balancing capability and relevant voltage profile improvement under different cases with DG connections.

switch status via network reconfiguration. The most well-balanced SOP. The combined method achieved the lowest peak branch load-
network was achieved by using the combined method with smaller ing. The voltage profile was also much flatter by using the SOP and
SOP size required.Fig. 12 shows the branch loading profiles and the combined method, as shown in Fig. 12b.
voltage profiles. As shown in Fig. 12a, the peak branch loading of This study considered only balanced 3-phase feeders. However
the network was reduced from 72% to 50% by using only one the proposed method is also applicable for unbalanced 3-phase
SOP. It shows that the increase in peak currents in the feeders distribution networks where 3-phase unbalanced power flow anal-
and branch loading was reduced effectively by only using one ysis needs to be employed.

(a) SOP Power Losses Feeder Power Losses

200
Total Power Losses (kW)

175
150
125
100
75
50
25
0
Base Case 0 0.01 0.015 0.02 0.025 0.065
Device Loss Coefficient k I /k J

(b) 600 Power Losses in Base Case

Total Power Lolsses (kW)

No Device Losses
500 99% VSC Efficiency at Rated Power
98% VSC Efficiency at Rated Power
400

300

200

100
0 10 20 30 40 50 60
Percentage of Load Increase (%)

Fig. 13. Impacts of the SOP device losses on total network power losses.
 W. Cao et al. / Applied Energy 165 (2016) 36–47 47

3.4. Impact of the SOP device losses However, a greater positive impact was obtained when system
loading increased. The requirements on SOP device efficiency for
A significant reduction on feeder power losses using SOPs has total network loss reduction were reduced when the system load-
been demonstrated in the previous study, where the SOP device ing increases.
losses were not considered. Although the SOP device losses have Although the focus of this paper is to quantify the technical
minor impact on its power flow control, it may lower the economic benefits of SOPs, the proposed algorithm is also able to be used
benefits obtained from reducing the total network power losses. for further economic analysis and planning studies of the applica-
The impact of SOP device losses on the total network power loss tion of SOPs, where life cycle costs need to the quantified and the
reduction was investigated based on sensitivity analysis. Different optimal amount, locations and sizes of SOPs need to be determined
VSC efficiencies at the rated power, i.e. the values of loss coeffi- considering the capital cost of SOPs.
cients kI and kJ in (9), were considered. Here for the VSCs in the
back-to-back converters, it was assumed that kI is equal to kJ. Acknowledgements
The constant power loss of each SOP device was set as 0.2% of
the rated power. This work was supported in part by the UK–China NSFC/EPSRC
Fig. 13a illustrates the total network power loss minimization OPEN Project (Grant No. EP/K006274/1 and 51261130473), the UK/
(feeder losses and SOP device losses) under the normal loading India HEAPD Project (Grant No. EP/K036211/1), and EU Horizon
conditions using one SOP which is located between buses 25 and 2020 P2P-SmarTest Project.
29 (without DG connections). Different device efficiencies of each
VSC were considered from 93.5% to 100%. It is observed that the References
total network losses were able to be reduced by using the SOP.
However, the benefits were lowered when the device efficiency [1] Hung DQ, Mithulananthan N, Bansal RC. Integration of PV and BES units in
commercial distribution systems considering energy loss and voltage stability.
decreased. The SOP lost its capability in reducing network power Appl Energy 2014;113:1162–70.
losses when the device loss coefficients fell to 0.065, i.e., 93.5% [2] Mu Y, Wu J, Jenkins N, Jia H, Wang C. A spatial-temporal model for grid impact
VSC efficiency at the rated power. Fig. 13b shows the performance analysis of plug-in electric vehicles. Appl Energy 2014;114:456–65.
[3] Järventausta P, Repo S, Rautiainen A, Partanen J. Smart grid power system
on total network power losses for each percentage of system load- control in distributed generation environment. Annu Rev Control
ing increase from 0% to 60%. The figure shows that SOP had a 2010;34:277–86.
greater positive impact on the network power losses under higher [4] Enacheanu B, Raison B, Caire R, Devaux O, Bienia W, HadjSaid N. Improving
voltage profile of residential distribution systems using rooftop PVs and
system loading conditions. The requirements on SOP device effi- battery energy storage systems. Appl Energy 2014;134:290–300.
ciency for total network loss reduction were reduced when the sys- [5] Baran ME, Wu FF. Network reconfiguration in distribution systems for loss
tem loading increases. reduction and load balancing. IEEE Trans Power Deliv 1989;4:1401–7.
[6] Kechroud A, Myrzik JMA, Kling W. Taking the experience from flexible AC
transmission systems to flexible AC distribution systems. In: 42nd
4. Conclusions International universities power engineering conference; 2007. p. 687–92.
[7] Mutale J, Strbac G. Transmission network reinforcement versus FACTS: an
economic assessment. In: IEEE power industry computer applications
The benefits of using SOPs for medium voltage distribution net- conference; 1999. p. 961–67.
works were investigated focusing on power loss reduction, feeder [8] Bloemink JM, Green TC. Increasing distributed generation penetration using
load balancing and voltage profile improvement. A generic power soft normally-open points. In: IEEE power and energy society general meeting;
2010. p. 1–8.
injection model of SOP that is suitable for steady state analysis [9] Bloemink JM, Green TC. Benefits of distribution-level power electronics for
was developed, taking into account both physical limits and inter- supporting distributed generation growth. IEEE Trans Power Deliv
nal power losses of a typical SOP device: back to back voltage 2013;28:911–9.
[10] Sayed MA, Takeshita T. All nodes voltage regulation and line loss minimization
source converters. The optimal SOP operation is obtained using in loop distribution systems using UPFC. IEEE Trans Power Electron
improved Powell’s Direction Set method and the combined method 2011;26:1694–703.
considering both SOP and network reconfiguration was proposed [11] Okada N, Takasaki M, Sakai H, Katoh S. Development of a 6.6 kV–1 MVA
transformerless loop balance controller. In: IEEE power electronics specialists
to demonstrate the superiority of using SOPs. Different quantity
conference; 2007. p. 1087–91.
of SOPs were considered and showed that SOPs contributed to sig- [12] Baran ME, Wu FF. Optimal sizing of capacitors placed on a radial distribution
nificant power loss reduction, feeder load balancing and voltage system. IEEE Trans Power Deliv 1989;4:735–43.
profile improvement. By comparing with network reconfiguration, [13] Flourentzou N, Agelidis VG, Demetriades GD. VSC-based HVDC power
transmission systems: an overview. IEEE Trans Power Electron
using only one SOP achieved similar improvement on network 2009;24:592–602.
power loss reduction and feeder load balancing. The greatest [14] Daelemans G, Srivastava K, Reza M, Cole S, Belmans R. Minimization of steady-
improvements were obtained when combining SOP and network state losses in meshed networks using VSC HVDC. In: IEEE power and energy
society general meeting; 2009. p. 1–5.
reconfiguration where smaller SOP sizes were required. High pen- [15] Powell MJD. An efficient method for finding the minimum of a function of
etration of DGs in the distribution network increases the needs for several variables without calculating derivatives. Comput J 1964;7:155–62.
loss minimization and feeder load balancing. Using only one SOP [16] Cheng-Ling L, Tai-Ning C, Yuan-Yao L, Pin H, Lee RK. Powell’s method for
designing optical multilayer thin-film filters. In: Optoelectronics and
achieved better performance than using network reconfiguration. communications conference; 2010. p. 388–89.
Combining SOP and network reconfiguration contributed to the [17] Lazarou S, Vita V, Ekonomou L. Application of Powell’s optimisation method
greatest improvements. In addition, SOPs are able to significantly for the optimal number of wind turbines in a wind farm. IET Sci Measur
Technol 2011;5:77–80.
reduce the peak currents in feeders and alleviate undesirable volt- [18] Cho H, Smith AD, Mago P. Combined cooling, heating and power: a review of
age excursions induced by the connection of DG and demand. performance improvement and optimization. Appl Energy 2014;136:168–85.
Therefore SOPs can be used as an alternative to infrastructure [19] Kiefer J. Sequential minimax search for a maximum. Proc Am Math Soc
1953;4:502–6.
upgrades in accommodating distributed energy resources. The
[20] Yu Y, Wu J. Loads combination method based core schema genetic shortest-
impact of SOP device losses was illustrated which shows that the path algorithm for distribution network reconfiguration. In: Proceedings of
economic benefit of SOP obtained from reducing total network power system technology conference, vol. 3; 2002. p. 1729–33.
power losses was lowered with the decrease of device efficiency.