Professional Documents
Culture Documents
Applied Energy
journal homepage: www.elsevier.com/locate/apenergy
h i g h l i g h t s
A model to optimal siting and sizing of SOPs was proposed considering DG impacts.
The problem was converted to an MISOCP model with improved accuracy and convergence.
Economic benefits and algorithm validation were given for different scale systems.
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 to replace normally open points in active
Received 13 September 2016 electrical distribution systems. SOPs can provide active/reactive power flow control and voltage regula-
Received in revised form 12 December 2016 tion under normal operating conditions as well as fast fault isolation and supply restoration under abnor-
Accepted 14 December 2016
mal conditions. The application of SOPs can improve the controllability of distribution systems, thus
further enhances the economy, flexibility and reliability of the grid. In this paper, considering the
long-term operation characteristics of distributed generation, a mixed integer non-linear optimization
Keywords:
problem is formulated to optimally determine the siting and sizing of SOPs based on the typical operation
Active electrical distribution network
Soft open points (SOPs)
scenarios generated by Wasserstein distance. It is then transformed to and solved as a mixed integer
Optimal siting and sizing second-order programming model. Finally, case studies on the IEEE 33-node system and Taiwan Power
Operation scenario generation Company distribution system are used to verify the effectiveness and efficiency of the proposed method.
Mixed integer second-order cone Ó 2016 Elsevier Ltd. All rights reserved.
programming (MISOCP)
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
2. Optimization model to site and size SOPs in an active electrical distribution network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
2.1. Typical operation scenario generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
2.2. Modelling of soft open points. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
2.3. Optimization model to site and size SOPs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
3. MISOCP model conversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
4. Case study. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
4.1. IEEE 33-node system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
4.2. Taiwan power Company distribution system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
4.3. Algorithm validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
⇑ Corresponding author.
E-mail address: lip@tju.edu.cn (P. Li).
http://dx.doi.org/10.1016/j.apenergy.2016.12.075
0306-2619/Ó 2016 Elsevier Ltd. All rights reserved.
302 C. Wang et al. / Applied Energy 189 (2017) 301–309
Furthermore, the number of decision variables and constraints is The probability ps of a scenario is calculated as follows:
increasing rapidly with the system scale, making the problem solv- 8 Z zs þzsþ1
ing difficult and time-consuming. Heuristic methods such as >
> 2
>
> p ¼ f ðxÞdx; s ¼ 1; 2; . . . ; S
genetic algorithm, simulated annealing and particle swarm algo- >
> s zs1 þzs
>
>
rithm are usually used to solve such problem. Ref. [17] used a >
>
2
Z z1 þz2
< 2
genetic algorithm to determine the active distribution network p1 ¼ z þz f ðxÞdx ð2Þ
expansion planning, in which the rewiring, network reconfigura- >
>
>
>
0 1
2
>
> Z zS þzSþ1
tion, new protection devices and DG installation were considered. >
>
>
>
2
Recently, some mathematical programming methods are also : pS ¼ z þz f ðxÞdx
S1 S
adopted to solve the problem. In [18], a convex formulation of 2
C. Wang et al. / Applied Energy 189 (2017) 301–309 303
Here, z0 and zSþ1 are the lower and upper limits of x, which are min f ¼ C CAP þ C OPE þ C LOSS ð6Þ
1 and þ1 if there is no additional information. The Wasserstein
distance is the integral of two PDF gaps [21]. The annual expense of the overall distribution system is com-
Weibull distribution [22] and Beta distribution [23] are usually posed of the following three parts.
used to describe PDFs of wind speed and light intensity for a long
time scale, which are used frequently in the study of electric power
(1) C CAP : capital cost of SOP
system planning problems. With the PDFs illustrated in the refer-
ences, the typical operation scenarios considered in siting and siz-
dð1 þ dÞ X N Xy
ing SOPs can be generated based on the Wasserstein distance. C CAP ¼ y cSOP SSOP
ij ð7Þ
ð1 þ dÞ 1 i¼1 j2XðiÞ
2.2. Modelling of soft open points
where d is the discount rate, and y is the SOP lifetime. N is the num-
An SOP is able to control the active power flow between the ber of nodes, XðiÞ is the set of all adjacent nodes of node i, and cSOP is
connected feeders and supply or absorb the reactive power at its the capital cost per unit capacity.
interface terminals under normal operation conditions. Fig. 1
shows a typical integration of an SOP. (2) C OPE : annual operational cost of SOP
This paper uses B2B VSCs in the optimization model for siting X
N X
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
(2) System power flow constraints Xn
K ¼ fx 2 Rn : x1 P j¼2 j
x2 ; x1 P 0g ð24Þ
X X
aji ðPs;ji r ji I2s;ji Þ þ Ps;i ¼ aik Ps;ik ð14Þ X
n
j2UðiÞ k2WðiÞ K ¼ fx 2 Rn : 2x1 x2 P x2j ; x1 ; x2 P 0g ð25Þ
X X j¼3
aji ðQ s;ji xji I2s;ji Þ þ Q s;i ¼ aik Q s;ik ð15Þ
j2UðiÞ k2WðiÞ As shown above, the second-order cone programming has nota-
U 2s;i U 2s;j 2ðrij Ps;ij þ xij Q s;ij Þ þ ðr 2ij þ x2ij ÞI2s;ij ¼ 0 ð16Þ bly strict demands on the mathematical formulation. The objective
function needs to be a linear function of the decision variable x,
I2s;ij U 2s;i ¼ P2s;ij þ Q 2s;ij ð17Þ and its feasible region is composed of linear equality constraints
Ps;i ¼ P DG SOP LOAD
s;i þ P s;i P s;i ð18Þ and convex cone constraints, where Ax ¼ b represents the linear
constraints and x 2 K represents the convex cone constraints.
Q s;i ¼ Q DG
s;i þ Q SOP
s;i Q LOAD
s;i ð19Þ
Therefore, the optimization model to site and size SOPs has to be
Eqs. (14) and (15) represent the law of power conservation at changed in advance before applying the second-order cone
each node i in the sth scenario, where P s;i and Q s;i denote the real programming.
and reactive power injection at node i. UðiÞ # N denotes the set of The aforementioned optimization model is described with
all parents of node i, and WðiÞ # N denotes the set of all children numerous nonlinear functions, such as the square representation
of node i. The Ohm’s law over branch ij at time t is expressed as of the voltage amplitude and current amplitude. First, additional
Eq. (16). The current magnitude of each line can be determined variables is introduced for each node and line to linearize nonlinear
by Eq. (17). Eqs. (18) and (19) indicate the power injection at each functions using variable substitution.
node i, which is the generation minus the load on node i in the sth e s;i denote the square of the magnitude of
For nodes i; j 2 N, let U
scenario. P DG DG
s;i and Q s;i are the real and reactive power, generated by its complex voltage, and let eI s;ij denote the square of the magnitude
the DG that is connected to node i via inverters. P LOAD and Q LOAD are of the complex current. Then, substituting the new optimization
s;i s;i
e s;i and eI s;ij into the proposed model, the objective func-
variables U
the real and reactive power consumption in the sth scenario. P SOP
s;i
tion and nonlinear constraints become linear except for the equal-
and Q SOP
s;i are the real and reactive power delivered by the SOP. ity constraints described in (17). To cast them as second-order
cone constraints, these nonlinear equality constraints are relaxed
(3) System security constraints to inequality constraints [26]:
ðU min
2
Þ 6 U 2s;i 6 ðU max Þ
2
ð20Þ I2s;ij U 2s;i P P 2s;ij þ Q 2s;ij ð26Þ
i i
2
I2s;ij 6 ðImax Þ ð21Þ Clearly, the original feasible set is enlarged, although previous
ij
studies [16] have shown that the relaxations are exact if moderate
where U max
i and U min
i are the upper and lower limits of the voltage conditions are satisfied, which implies that inequalities (26) can be
amplitude at node i. Imax
ij is the upper limit of the current amplitude tight in the optimal solution. More mathematical proofs can be
of line ij. found in [27,28].
The relationship between the location and capacity of an SOP is The corresponding second-order cone programming formula-
represented as follows: tion is
SSOP ¼ mij smodule ð1 aij Þ ð22Þ 2Ps;ij
ij
2Q
s;ij 6 eI s;ij þ U
e s;i ð27Þ
where smodule is the minimum optimum capacity, which represents
eI U e
the capacity of the basic power electronic module that comprises s;ij s;i 2
the SOP. mij is the quantity of the modules. Eq. (27) ensures that the decision variables are of the quadratic
The optimization variables in the model include the installation cone form and limits the search space within the convex cone
sites and capacities, switch states, and the transmitted active/reac- range.
tive power of the SOPs in each scenario. As a consequence, Eqs. (3)– The capacity constraints of an SOP are transformed as follows
(22) form the optimization model to determine the installation according to the SOCP requirements:
sites and capacities of SOPs in an active distribution system.
2 SOP 2 SSOP
ij SSOP
ij
3. MISOCP model conversion ðPSOP
s;i Þ þ ðQ s;i Þ 6 2 pffiffiffi pffiffiffi ð28Þ
2 2
Considering typical operation scenarios, the optimization model SOP 2 SOP 2 SSOP
ij SSOP
ij
ðPs;j Þ þ ðQ s;j Þ 6 2 pffiffiffi pffiffiffi ð29Þ
to determine the siting and sizing of SOPs becomes complicated, 2 2
making it more difficult to be solved. Second-order cone program-
ming (SOCP) is a mathematically convex programming approach Eqs. (28) and (29) are equivalent to the original constraints (4)
that can be considered as the generalisation of both linear and non- and (5) while satisfying the requirement of the rotated quadratic
linear programming. SOCP can solve minimum linear objective cone form.
functions based on a convex cone in a linear space. It has excellent For the absolute terms jP SOP SOP
s;i j and jP s;j j in Eqs. (3) and (10), aux-
performance in terms of global optimality and computation effi-
iliary variables M SOP SOP
s;i and M s;j are introduced to represent and lin-
ciency. The SOCP standard form is written as follows [25]:
earize them as follows:
minfc T xjAx ¼ b; x 2 Kg ð23Þ
MSOP SOP
s;i P 0; M s;j P 0 ð30Þ
where x is the dimensional decision variables. c, b and A are the con-
stant vectors and matrix. K denotes the Cartesian product of a lim- MSOP
s;i P PSOP SOP
s;i ; M s;i P PSOP
s;i ð31Þ
ited nonempty-pointed convex cone, which is generally expressed MSOP SOP SOP SOP
s;j P P s;j ; M s;j P P s;j ð32Þ
using a quadratic cone (24) or a rotated quadratic cone (25):
C. Wang et al. / Applied Energy 189 (2017) 301–309 305
PV WT WT
23 24 25 26 27 28 29 30 31 32 33
PV WT PV WT WT
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
19 20 21 22
Table 1
Installation capabilities of DGs.
Parameters WT PV
Location 10 16 17 30 32 7 13 27
Capacity, kVA 500.0 200.0 150.0 200.0 300.0 500.0 300.0 400.0
25
20
Wind speed, m/s
15
10
0
0 2000 4000 6000 8000 8760
Time, h
1200
Light intensity, W/m2
900
600
300
0
0 2000 4000 6000 8000 8760
Time, h
After the conic conversion, the MISOCP model to determine the 4.1. IEEE 33-node system
installation sites and capacities of SOPs is formulated.
The modified IEEE 33-node system is presented in Fig. 2. It has
32 normally closed branches and 5 normally open branches. The
4. Case study IEEE 33-node system has been widely used to verify and demon-
strate various types of methods and algorithms developed for the
In this section, the modified IEEE 33-node system and TPC dis- planning and operation of medium voltage distribution networks.
tribution system are used to demonstrate the effectiveness and More detailed parameters are shown in [30].
efficiency of the proposed method. The proposed method in this In order to consider the impact of high penetration of DGs, five
paper was implemented in the YALMIP optimization toolbox [29] wind turbines and two photovoltaic cells are integrated into the
with MATLAB R2013a, and solved by IBM ILOG CPLEX 12.6. The networks, respectively. And the basic configuration parameters
computation is performed on a PC with an Intel Xeon CPU E5- are shown in Table 1. All the DGs are operated with a power factor
1620 @3.70 GHz processor and 32 GB RAM. 1.0.
306 C. Wang et al. / Applied Energy 189 (2017) 301–309
The annual wind speed and light intensity curves used are According to the wind speed and light intensity from Fig. 3, the
shown in Fig. 3. The parameters of the studied case are shown in Weibull distribution parameters can be obtained: the scale param-
Table 2. eter c is 2, and the shape parameter k is 10. The Beta distribution
parameters k1 and k2 are 0.95. Considering the operation charac-
Table 2 teristics of WT and PV, the typical operation scenarios are obtained
Parameters of the studied case. based on the Wasserstein distance from their PDFs.
According to the Weibull distribution parameters, the WT
Parameters Value
parameters and the rated output power (PWT;N ¼ 1:0), the WT
Discount rate 0.08
PDF is obtained as follows:
SOP economical service life, year 20.00
8
< 0:088 dðPWT 1Þ
> ðPWT ¼ 0Þ
SOP minimum optimum capacity, kVA 100.00
SOP unit capital cost, $/kVA 308.80
f ðPWT Þ ¼ 1:5P0:333
0:667
SOP loss coefficient [31] 0.02 e2:25PWT ð0 < PWT < 1Þ ð33Þ
>
:
WT
Coefficient of annual operational cost 0.01 0:103 dðPWT 1Þ ðPWT ¼ 1Þ
Cut-in wind speed, m/s 3.00
Rated wind speed, m/s 15.00 According to the Beta distribution parameters, the PV parame-
Cut-out wind speed, m/s 25.00
ters and the rated output power (PPV;N ¼ 1:0), the PV PDF is as
Electricity Price, $/kWh 0.08
follows:
Table 3
Probability distribution of wind and photovoltaic power.
P WT P PV
Scenario 1 Scenario 2 Scenario 3 Scenario 4 Scenario 5
0.0966 0.2971 0.5000 0.7029 0.9034
Scenario 1 0.0000 0.0180 0.0174 0.0173 0.0174 0.0180
Scenario 2 0.0795 0.0916 0.0885 0.0882 0.0885 0.0915
Scenario 3 0.3163 0.0468 0.0452 0.0451 0.0452 0.0468
Scenario 4 0.7094 0.0268 0.0259 0.0258 0.0259 0.0268
Scenario 5 1.0000 0.0210 0.0203 0.0203 0.0203 0.0210
PV WT WT
23 24 25 26 27 28 29 30 31 32 33
PV WT PV WT WT
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
19 20 21 22
(a)
PV WT WT
23 24 25 26 27 28 29 30 31 32 33
PV WT PV WT WT
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
19 20 21 22
(b)
Fig. 4. (a) Location of SOPs in Scheme I and (b) Location of SOPs in Scheme II.
C. Wang et al. / Applied Energy 189 (2017) 301–309 307
Table 4
Location and capacity of the SOPs without/with network reconfiguration.
f ðP PV Þ ¼ 0:904P0:05
PV ð1 PPV Þ0:05 ð34Þ The optimal installation sites and capacities of the SOPs on IEEE
33-node system are shown in Fig. 4 and Table 4, and the objective
Based on the Wasserstein distance, five wind power scenarios, function values are shown in Table 5.
five photovoltaic scenarios and their corresponding probabilities Table 5 shows that Scheme II has better economic benefits. The
are simultaneously obtained. Then, the combined 25 scenarios annual total cost of Scheme II is $25,210.7 less than that of without
and corresponding probabilities are also derived and presented in SOP, i.e., it decreased by 29.7%. Among the expenses, the annual
Table 3. energy loss cost of the distribution system is reduced by $40251.0
The following two schemes are used to determine the optimal (reduction of 47.4%). Thus, the operational economy of the power
installation sites and capacities of the SOPs on IEEE 33-node distribution system is significantly improved. Based on the dynamic
system. payback period, it takes 9.20 years to recover the SOP capital cost in
Scheme I. However, only 3.55 years are needed to recover the SOP
Scheme I: Without the consideration of network topology capital cost by considering network reconfiguration.
changing, only normally open points are considered as the can-
didate locations of SOPs.
4.2. Taiwan power Company distribution system
Scheme II: With the consideration of network reconfiguration,
both normally open points and normally closed points are con-
Taiwan Power Company (TPC) distribution system is a typical
sidered as the candidate locations of SOPs.
multi-sectioned electrical distribution network, which has 11 feed-
ers, 83 normally closed branches and 13 normally open branches.
Compared with IEEE 33-node system, this case is more challenging
in system scale and complexity, which is used to further verify the
Table 5 scalability of the proposed method. The detailed parameters of the
Annual cost of different schemes. TPC distribution system are provided in [32].
C CAP , $ C OPE , $ C LOSS , $ C, $ Considering multiple operation scenarios of DG, six wind tur-
bines and twelve photovoltaic cells are integrated into the system.
Without SOP – – 84,926.2 84,926.2
The capacities of wind turbines and photovoltaic arrays are
Scheme I 25,145.3 4,935.3 44,458.1 74,538.7
Scheme II 12,572.7 2,467.6 44,675.2 59,715.5 800 kVA and 600 kVA, respectively. The optimal installation sites
and capacities of the SOPs are shown in Fig. 5 and Table 6. The
PV WT 8 9 PV WT PV
A 10 G
1 2 3 4 5 6 7 55 54 53 52 51 50 49 48 47
WT PV PV WT PV
B H
12 13 64 63 62 61
11 60 59 58 57 56
14
PV WT PV
22 PV
C I
15 16 17 18 19 20 21 23 24 72 71 70 69 68 67 66 65
D J
25 26 27 28 29 76 75 74 73
PV WT PV 39 40 PV
E K
30 31 32 33 34 35 36 37 38 41 42 83 82 81 80 79 78 77
F
43 44 45 46
Table 6
Location and capacity of the SOPs in the TPC distribution system.
Table 7
Comparison of the performance of the MISOCP model and original model.
annual total cost, including capital cost and annual operational cost distribution devices such as SOPs will be integrated into the active
of the SOP, and annual energy loss cost of the distribution system, electrical distribution networks. The work conducted by this paper
is $272,108.1. It is less than that of without SOPs, which is can give a better siting and sizing scheme for the balance of capital
$292,013.5. Among the costs, the annual energy loss cost of the dis- and profit in the engineering practice. Meanwhile, considering sig-
tribution system is reduced by $68,804.3 (reduction of 23.6%). nificant spatial and temporal uncertainties, large-scale mixed inte-
ger nonlinear optimization problems in distribution systems
4.3. Algorithm validation become more difficult to solve, and the advantages of the proposed
method will become more obvious.
To verify the effectiveness and efficiency of the model and test
the correctness of the conic conversion, KNITRO [33] in GAMS Acknowledgements
package is used as a reference. KNITRO is an optimization package
based on KKT algorithm and interior point method, which can This work was supported by the National Natural Science
obtain the global optimum and has been widely applied in solving Foundation of China (51261130473, 51307116) and conducted in
the MINLP problem. In Scheme I of IEEE 33-node system, all of the coopetation of APPLIED ENERGY UNiLAB-DEM: Distributed Energy
algorithms converge and obtain the same installation sites and & Microgrid. UNiLAB is an international virtual lab of collective
capacities of the SOPs with 5 integer variables. However, KNITRO intelligence in Applied Energy.
does not converge with more variables and constraints. Table 7
shows the solving time and results for the different algorithms.
References
Table 7 shows that in the case of fewer integer variables and
constraints such as the Scheme I of the IEEE 33-node system, all [1] Hung DQ, Mithulananthan N, Bansal RC. Integration of PV and BES units in
algorithms converge to the identical result, but the MISOCP model commercial distribution systems considering energy loss and voltage stability.
is more efficient. When the quantity of integer variables and con- Appl Energy 2014;113:1162–70.
[2] Lopes JAP, Hatziargyriou N, Mutale J. Integrating distributed generation into
straints increases such as the Scheme II of the IEEE 33-node system electric power systems: a review of drivers, challenges and opportunities.
and the TPC distribution system, the rapid growth in problem scale Electric Power Syst Res 2007;77:1189–203.
causes the divergence of KNITRO. Because of the linearization and [3] Conti S, Nicolosi R, Rizzo SA, et al. Optimal dispatching of distributed
generators and storage systems for MV islanded microgrids. IEEE Trans
convex relaxation of the original model, the MISOCP model still
Power Delivery 2012;27(3):1243–51.
shows good accuracy and convergence by reducing the complexity [4] Rueda-Medina A, Padilha-Feltrin A. Distributed generators as providers of
of the problem. Therefore, the MISOCP model in this paper is more reactive power support-a market approach. IEEE Trans Power Syst 2013;28
effective than the original MINLP model. (1):490–502.
[5] Kuroda K, Magori H, Ichimura T, et al. A hybrid multi-objective optimization
method considering optimization problems in power distribution systems. J
5. Conclusion Modern Power Syst Clean Energy 2015;3(1):41–50.
[6] Verzijlbergh R, Vries LD, Lukszo Z. Renewable energy sources and responsive
demand. Do we need congestion management in the distribution grid. IEEE
In this paper, a large-scale MINLP model is proposed to opti- Trans Power Syst 2014;29(5):2170–8.
mally determine the optimal installation sites and capacities of [7] Morvaj B, Evins R, Carmeliet J. Optimization framework for distributed energy
systems with integrated electrical grid constraints. Appl Energy
SOPs, considering the characteristics of DG and network topology 2016;171:296–313.
changes. Based on the typical scenarios generation and conic con- [8] Huang S, Wu Q, Oren S, et al. Distribution locational marginal pricing through
version, an MISOCP model is adopted to solve the problem. The quadratic programming for congestion management in distribution networks.
IEEE Trans Power Syst 2015;30(4):2119–28.
effectiveness and efficiency of the proposed method are verified [9] Long C, Wu J, Thomas L, et al. Optimal operation of soft open points in medium
in case studies. Results showed that the application of SOPs can voltage electrical distribution networks with distributed generation. Appl
significantly improve the operational economy of the distribution Energy 2016;184:427–37.
[10] Bloemink JM, Green TC. Increasing distributed generation penetration using
systems. With the integration of renewable energy resource, the
soft normally-open points. 2010 IEEE PES General Meeting, vol. 1. p. 1–8.
flexible control mode of SOPs will also increase the DG hosting [11] Bloemink JM, Green TC. Benefits of distribution-level power electronics for
capacity of distribution networks, improve the voltage profile supporting distributed generation growth. IEEE Trans Power Delivery 2013;28
and enhance the reliability of power supply. (2):911–9.
[12] Cao W, Wu J, Jenkins N. Feeder load balancing in MV distribution networks
With the development of power electronic technologies and using soft normally-open points. Proceedings of the 2014 innovative smart
increasingly complex distribution systems, more flexible grid technologies conference Europe, vol. 1. p. 1–6.
C. Wang et al. / Applied Energy 189 (2017) 301–309 309
[13] Bloemink JM, Green TC. Increasing photovoltaic penetration with local energy [23] Karaki SH, Chedid RB, Ramadan R. Probabilistic performance assessment of
storage and soft normally-open points. 2011 IEEE PES general meeting, vol. 1. autonomous solar-wind energy conversion systems. IEEE Trans Energy
p. 1–8. Convers 1999;14(3):766–72.
[14] Cao W, Wu J, Jenkins N, et al. Operating principle of Soft Open Points for [24] Flourentzou N, Agelidis VG, Demetriades GD. VSC-based HVDC power
electrical distribution network operation. Appl Energy 2016;164:245–57. transmission systems: an overview. IEEE Trans Power Electron
[15] Cao W, Wu J, Jenkins N, et al. Benefits analysis of Soft Open Points for electrical 2009;24:592–602.
distribution network operation. Appl Energy 2016;165:36–7. [25] Huang C, Li F, Ding T, et al. Second-order cone programming-based optimal
[16] Huang X, Yang Y, Taylor GA. Service restoration of distribution systems control strategy for wind energy conversion systems over complete operating
under distributed generation scenarios. CSEE J Power Energy Syst 2016;2 regions. IEEE Trans Sust Energy 2015;6(1):263–71.
(3):43–50. [26] Low SH. Convex relaxation of optimal power flow-Part I: formulations and
[17] Martins VF, Borges CL. Active distribution network integrated planning relaxations. IEEE Trans Control Network Syst 2014;1:15–27.
incorporating distributed generation and load response uncertainties. IEEE [27] Farivar M, Low SH. Branch flow model: relaxations and convexification–parts I.
Trans Power Syst 2011;26:2164–72. IEEE Trans Power Syst 2013;28:2554–64.
[18] Nick M, Cherkaoui R, Paolone M. Optimal allocation of dispersed energy [28] Farivar M, Low SH. Branch flow model: relaxations and convexification–parts
storage systems in active distribution networks for energy balance and grid II. IEEE Trans Power Syst 2013;28:2565–72.
support. IEEE Trans Power Syst 2014;29:2300–10. [29] Lofberg J. Yalmip: a toolbox for modeling and optimization in MATLAB. IEEE
[19] Hochreiter R, Pflug GC. Financial scenario generation for stochastic multi-stage Int Sympos Comput Aided Control Syst Design 2004;1:284–9.
decision processes as facility location problems. Ann Oper Res 2007;152 [30] Baran ME, Wu F. Network reconfiguration in distribution systems for loss
(1):257–72. reduction and load balancing. IEEE Trans Power Delivery 1989;4(2):1401–7.
[20] Zou K, Agalgaonkar AP, Muttaqi KM, et al. Distribution system planning with [31] Mostafa N, Rachid C. P Mario. optimal allocation of dispersed energy storage
incorporating dg reactive capability and system uncertainties. IEEE Trans Sust systems in active distribution networks for energy balance and grid support.
Energy 2012;3(1):112–23. IEEE Trans Power Syst 2014;29(5):2300–10.
[21] Piccoli B, Rossi F. Generalized Wasserstein distance and its application to [32] Su C, Lee C. Network reconfiguration of distribution systems using improved
transport equations with source. Arch Ration Mech Anal 2014;211: mixed-integer hybrid differential evolution. IEEE Trans Power Delivery
1335–58. 2003;18:1022–7.
[22] Shu Z, Li Q, Chan P. Investigation of offshore wind energy potential in Hong [33] Bonami P, Biegler LT, Conn AR. An algorithmic framework for convex mixed
Kong based on Weibull distribution function. Appl Energy 2015;156:362–73. integer nonlinear programs. Discrete Optim 2008;5:186–204.