energies

Article
Dynamic Power Flow Cascading Failure Analysis of
Wind Power Integration with Complex
Network Theory
Yushu Sun 1 ID
, Xisheng Tang 1, *, Guowei Zhang 1 , Fufeng Miao 2 and Ping Wang 3
1 Institute of Electrical Engineering, Chinese Academy of Sciences, Haidian District, Beijing 100190, China;
yushusun@mail.iee.ac.cn (Y.S.); zhangguowei@mail.iee.ac.cn (G.Z.)
2 State Grid Henan Power Company Economic and Technological Research Institute, Zhengzhou 450052,
China; miaoff@vip.126.com
3 Jiangsu Province Electrochemical Energy Storage Technology Key Laboratory, Taizhou 225500, China;
wangping@shuangdeng.com.cn
* Correspondence: tang@mail.iee.ac.cn; Tel.: +86-10-82547108; Fax: +86-10-82547103

Received: 30 October 2017; Accepted: 6 December 2017; Published: 29 December 2017

Abstract: The impact of the rapid development of large-scale centralized wind power farms on
the power system is drawing more and more attention. Some topics about grid-connected wind
power are discussed from the view of complex network theory in this paper. Firstly, a complex
network cascading failure model is established, combined with dynamic AC power flow (DACPF).
Then, the IEEE 30 bus system is used to analyze its validity using the simulations of nodes removal,
wind power integration, as well as the change of current and voltage boundaries. Furthermore,
the influences of wind power before and after smoothing are investigated. Also, different wind
power coupling locations are studied. Finally, some significant conclusions are obtained to provide
references for large-scale wind power integration.

Keywords: wind power; complex network theory; cascading failure; dynamic AC power flow

1. Introduction
With the rapid development of the larger-scale interconnected power systems, the complexity
of sizes and operations in power systems is increasing. This means that the stability, security and
reliability of power system are facing an enormous challenge. Large blackouts have attracted more
and more attention due to power systems suffering from declining stability, and even cascading failure
events. The 2003 “8·14” blackout in North America, a load loss up to 61.8 GW, led to a huge economic
loss of 30 billion dollars [1]. In the same year, the “8·28” blackout in London brought about a load
loss of 724 MW and resulted in 500,000 passengers being trapped in the subway [2]. These large
blackouts are triggered by the successive malfunction of plenty of elements, that is, cascading failure,
always induced by an initial disturbance or event, for example, component failure, ground fault,
protection malfunctioning, or a hostile environment.
Recently, large-scale and long-distance grid-connected wind power have brought a more
prominent impact on the stable operation of power system owing to its intermittent and randomness.
The probability of cascading failure is then increased. As we know, the power system is a typical
complex network, which has plenty of nodes and lines, as well as complicated topology structure
and operating characteristics. Therefore, cascading failure in grid-connected wind power system is
analyzed based on complex network theory in this paper.
So far, complex network theory has been widely applied in power system stability analysis.
Chowdhury et al. [3] use topological indices to analyze power system vulnerability with renewable

Energies 2018, 11, 63; doi:10.3390/en11010063 www.mdpi.com/journal/energies

Energies 2018, 11, 63 2 of 15

energy sources. Wang et al. [4] present complex network theory to identify the vulnerable lines of
power system with a vulnerability index. Luo et al. [5] can identify transmission sections from three
levels of transmission lines, key transmission links and partition sections automatically. Chu et al. [6]
illustrate how complex network theory can be applied to modern smart grids in structural vulnerability
assessment, cascading blackouts, network reconfigurations, and so on. Correa et al. [7] obtain a similar
conclusion from the graph theory on electric network vulnerability by comparisons of physical power
flow models and scale-free graph statistic indexes. Fan et al. [8] propose a cascading failure model
based on complex network theory by combining the node overload failures and hidden failures of
transmission lines in blackouts together. The simulation results are very useful for cascading failure
propagation mechanism analysis and the protection strategy development against cascading failures
in power grids.
A lot of research on cascading failures in power system has been done, and the most important
relates to how to define the load in the established model. Yin et al. [9] define a function of node degree
β
kj , Fj = k j , j = 1, 2, . . . , N, as initial load of each node j in the power grid. Here, the cascading failure
of complex power networks is analyzed with the local preferential redistribution rule of the load of
failure node. Zhang et al. [10] adopt Li = k i ∑m∈Γi k m as an initial load of node i, where ki is the


 α
degree of node i and Γi denotes the neighborhood of node i. Liu et al. [11] use L j = k j ∑m∈Γ j k m as
the load of a node j, where α is a tunable parameter. Zhang et al. [12] establish a similar model with
complex network theory, where the initial load of node i is described as Li = αk i + (1 − α)∑ j∈Γi k j .
Wang et al. [13] apply a cascading model using a local load redistribution rule. The initial load of edge ij

α
is Li = k i k j , where ki and kj are the degrees of the nodes connected by the edge. Wang et al. [14] use
h
 iα
Lij = k i ∑ a∈Γi k a k j ∑b∈Γ j k b as the initial load of edge ij, and analyze the cascading phenomenon
of uncorrelated scale-free networks with two different attack strategies on edges.
In power system, some nodes or edges have smaller degree, but play an important role in
power transmission. Therefore, the concept of betweenness is introduced to reveal the propagation
mechanism of network failure. Peng et al. [15] establish a network cascading failure model, where the
number of shortest paths passing through a node (i.e., node betweenness) are defined as its load.
Nicholson [16] uses betweenness as edge load to sort the critical edges for the implementation of
preparedness options.
In the analysis of betweenness, power transmission between two nodes is supposed to pass
the shortest path, which is far from Kirchhoff’s law. Power transmission from a generator node
to a load node will be undertaken by all possible lines. Bompard et al. [17] define an extended
betweenness centrality considering the complex characteristics of power system. The local importance
of the components in power system can also be measured. Bompard [18] and Yan [19] et al.
propose a joint method of extended betweenness to define electric power load considering network
structure and electrical characteristics. By using this power transfer distribution factor-based model
(i.e., extended betweenness), the most critical lines and buses in an electrical power grid can be
evaluated. Lin et al. [20] propose a self-healing transmission network reconfiguration algorithm based
on the complex network theory to analyze the capacities of generators and the amounts of high priority
loads, as well as the distribution and importance of each transmission line.
However, the extended or electrical betweenness identifies the critical nodes or lines from the
perspective of topological structure without considering the operational states of power system.
Yan et al. [21] discuss the validity of a typical DC power flow-based cascading failure simulator in
cascading failure analysis using the critical moment. Qi et al. [22] use the probabilistic power flow
combined with DC power flow model to analyze single line failure probability. Chen et al. [23] embed a
DC power flow model with hidden failures into the traditional error and attack tolerance methodology
to develop a new approach for power system vulnerability modeling and assessment.
DC power flow model is modeled with linear equations, so it does not accurately reflect the
nonlinear characteristics of the power system. Dwivedi et al. [24] explore a new centrality index with
Energies 2018, 11, 63 3 of 15

the maximum AC power flow to analyze the 118 bus systems. The capacity of the connecting links is
defined, and the maximum flow from the source to the sink is calculated. Dey et al [25] investigate the
relationship between the topological parameter variation and failure propagation rate in the cascading
mechanism through various test beds based on AC power flow equations. Sun et al. [26] propose
a cascading failure AC power flow model to discuss the impact of wind power on power systems,
and some interesting conclusions are obtained.
AC power flow (ACPF) is employed to analyze power system performance characteristics,
where the unbalanced power occurred in the system is completely compensated by a slack node.
However, the unbalanced power is usually processed by coordinated dispatching and control of
multiple adjustable generating sets in the actual operation. Therefore, a dynamic AC power flow
(DACPF) model combined with complex network theory is proposed to discuss power system
cascading failure, and then the impact of wind power integration on the power grid is studied
more accurately.
This paper is organized as follows. The introduction is displayed in Section 1. Section 2 describes
complex network theory in details. Section 3 establishes the DACPF model with complex network
theory. Different simulation scenarios are developed in Section 4. Section 5 gives the conclusions.

2. Complex Network Theory

According to graph theory, complex network theory investigates the target system from structure,
physical and fault propagation characteristics regardless of the dimensionality of the system overall.
It can reveal the operation law of power system, explain the process mechanism of cascading failures,
and provide effective measures to reduce and prevent the occurrence of cascading failures or blackouts.
The power system can be equivalent to a weighted, directed map on account of complex network
theory, where lines are power transmission lines, and nodes are buses.

2.1. Typical Networks

Nowadays, many kinds of complex network models have been proposed, such as the regular
network, random network, small-world network, and scale-free network.
Regular network is a symmetric ring, the number of neighboring nodes of each vertex in
the network are identical [27]. Then, Erdös and Rényi propose the ER (Erdös and Rényi) andom
network model [28], where the edge is generated randomly. When the network size is large enough,
degree distribution follows the binomial or Poisson distribution, which means the majority of the
nodes have similar degree. In addition, Watts and Strogatz construct the WS (Watts and Strogatz)
mall-world network model, which is between a regular network and random network. Due to
a smaller shortest path length as well as a larger clustering coefficient, the speed and the extent
of fault propagation of small-world network are higher than those of regular network and random
network [29,30]. Studies have shown that the Iranian power grid has obvious small-world property [31],
and the medium- and low-voltage power grids of Northern Netherlands are small-world network [32].
Finally, Barabási and Albert establish the BA (Barabási and Albert) cale-free network model [33],
which possesses a power-law degree distribution characteristic. It has a handful of core nodes, so that
it is vulnerable to deliberate attack but robust against random attacks. The North American power
grid has been proved to have prominent scale-free features [34].

2.2. Basic Characteristics

There are four basic characteristics including node degree, clustering coefficient, electrical distance,
and electrical betweenness used to interpret complex network theory [26]. Node degree can measure
the significance of nodes in the power system. Clustering coefficient describes the tightness and
clustering of the whole network. Electrical distance reflects power transmission difficulty and loss in
the power grid. Electrical betweenness characterizes the contribution of nodes and edges to power
transmission, which is important for the identification of the critical nodes and lines in power system.
Energies 2018, 11, 63 4 of 15

3. Complex Network Model

A complex network cascading failure model combined with DACPF is proposed in this section.
When failure occurs in the system, current and voltage are used as the boundary conditions to prevent
fault propagation. Then, evaluation indicators are adopted to evaluate power grid damage.

3.1. DACPF
There is only a slack node in the calculation of ACPF, that is, all the unbalanced power of the
system is compensated by the slack node. For online applications, power imbalance and frequency
change in the system are usually caused by line outage, generator failure, load change, and so on.
A part of the unbalanced power is adjusted by the load, and most of those are undertaken by the
generators. This process is implemented by the coordinated control of multiple generators, not a
balancing machine. Therefore, the DACPF is adopted to meet practical needs.
For a given system, the active power flow equation is calculated as follows.

PGi − PDi − Pi ( x ) = 0, i = 1, 2, . . . , n (1)

where PGi , PDi , Pi (x) are generation, load and power loss of node i, respectively.
If the unbalanced power occurs in power system

N N
∆P∑ = ∑ PGi − ∑ PDi − Ploss (x) (2)
i =1 i =1

where Ploss is the total transmission loss of power system, which is compensated by all generators.
Then, Equation (1) can be displayed as follows.

PGi − αi ∆P∑ − PDi − Pi ( x ) = 0, i = 1, 2, . . . , n (3)

 
n
where αi ∑ αi = 1, αi ≥ 0 is the proportion of the unbalanced power compensated by generator i.
i =1
If the node i is not a generator, or its output power cannot be adjusted, αi is equal to 0. For the actual
system, αi can be determined by the frequency response coefficient of each generator, certain economic
criteria, and so on.
The modification of Jacobian matrix in the ACPF equation is ∆J due to the introduction of system
loss Ploss ( x ). In order to improve calculation efficiency of DACPF, the impact of Ploss ( x ) on the Jacobian
matrix is ignored.
In order to correct the active power mismatch in step i of the DACPF iterative calculation,
the variation of system loss in step i − 1 is given:

∆Ploss ( xi−1 ) = Ploss ( xi−1 ) − Ploss ( x0 ) (4)

Then, the active equation mismatch of the generator in step i can be obtained.

0
∆PGi = ∆PGi − αi ∆Ploss ( xi−1 ) (5)

Finally, DACPF can be obtained exactly through the above calculations.

3.2. Boundary Conditions

The voltage and current are used as the boundary constraints in the power grid based on DACPF.
According to the Motter and Lai model [35], the voltage and current limits of the power network are
proportional to their rated voltage and current.

Uc = (1 + βU ) · Ur (6)
Energies 2018, 11, 63 5 of 15

Ic = (1 + β I ) · Ir (7)

where Ur , Ir are rated voltage and current; Uc , Ic are the boundaries of voltage and current; β is the
tolerance factor, which reveals the ability of the power network to compensate the extra load; 1 + βU is
the voltage setting coefficient, and 1 + β I is the current setting coefficient.

3.3. Evaluation Indicators

Evaluation indicators are used to determine the damaged condition of power system when the
network failure occurs.
A. Connectivity
Connectivity is the rate of the node number in maximum power supply radius before and after
failure. It reveals the split degree of power system after malfunction.

N0
g= (8)
N

where g expresses connectivity; N and N 0 indicate the number of nodes in maximum power supply
radius before and after failure.
B. Global effective performance
Global effective performance describes the power transmission efficiency.

1 1
NG NL ∑ i∈G,j∈ L dij
E= (9)

where E expresses global effective performance; subscript G and L represent the sets of generation and
load nodes; NG and NL display the total number of generation and load nodes; dij present the shortest
electrical distance from generation node i to load node j.

3.4. Simulation Process

(1) Initial DACPF, including active power, reactive power, voltage and current are calculated based on
the given conditions. Then, determine the boundaries of network voltage and current according
to Equations (6) and (7).
(2) Wind power integration.
(3) Recalculate DACPF of power system, and remove the nodes or edges, whose voltage or current
exceed the limit values. Repeat this process until no failure occurs.
(4) Estimate the impact of wind power on power grid with evaluation index.

4. Case Study
According to topology principles and methods of complex network [26], the IEEE 30 bus system
is modeled as a directed, weighted and sparse connected map including 30 nodes and 41 edges,
as demonstrated in Figure 1. The advantage of DACPF in cascading failure analysis is discussed in the
following applications.
Energies 2018, 11, 63 6 of 15
Energies 2018, 11, 63 6 of 15
Energies 2018, 11, 63 6 of 15

Figure 1.
Figure 1. IEEE
IEEE30
30bus
bussystem.
system.
Figure 1. IEEE 30 bus system.
4.1.
4.1. ImpactofofRemoving
Impact RemovingNodesNodeson on Power
Power System
System
4.1. Impact of Removing Nodes on Power System
The impacts of ACPF and DACPF model on power system are analyzed by nodes removal
The impacts of ACPF and DACPF model on power system are analyzed by nodes removal
The impacts
randomly under of theACPF and DACPF
same boundary modelThe
condition. on simulation
power system are
results of analyzed
removing by nodes
node removal
1, 3, 10, 13,
randomly under the same boundary condition. The simulation results of removing node 1, 3, 10, 13,
randomly under
28 are shown in the
Figuresame boundary condition.
2. Connectivity, The
that is, g, of simulation
ACPF and DACPF results of removing
model node
are identical when1, 3,node
10, 13,
28 are shown in Figure 2. Connectivity, that is, g, of ACPF and DACPF model are identical when node
2813
areis shown
removed,in Figure
while the 2. Connectivity,
former is less that theg,latter
than is, of ACPF and DACPF
by removing othermodel
nodes.are identical
Global when
effective
13 is removed, while the former is less than the latter by removing other nodes. Global effective
node 13 is removed,
performance, while
that is, E, ofthe former
ACPF andisDACPF
less thanmodel
the latter by removing
are both other nodes.
0 by removing node 1Global
and nodeeffective
3.
performance,
However, global that is, E, of ACPF and DACPF model are both 0 by removing node 1 and node 3.
performance, that effective
is, E, of performance
ACPF and DACPF of DACPF model
model areisboth
larger than
0 by that of ACPF
removing nodemodel
1 andwhennode 3.
However,
other nodesglobal effective performance
are removed. of DACPF model isDACPF
larger than that of ACPF model when
However, global effective Therefore,
performanceit is of
more reasonable
DACPF model foris larger model to analyze
than that power
of ACPF modelsystem
when
other nodes
cascading are removed. Therefore, it is more reasonable for DACPF model to analyze power system
other nodes failure triggered
are removed. by nodesitremoval.
Therefore, is more reasonable for DACPF model to analyze power system
cascading failure triggered by nodes removal.
cascading failure triggered by nodes removal.
ACPF model
DACPF
ACPF model
model
DACPF model
0.6

0.6
0.5

0.5
0.4
g

0.4
0.3
g

0.3
0.2

0.2
0.1

0.1
0.0
1 3 10 13 28
0.0 node number
1 3 10 13 28
node
(a)number

(a)

Figure 2. Cont.
Energies 2018, 11, 63 7 of 15
Energies 2018, 11, 63 7 of 15
Energies 2018, 11, 63 7 of 15

ACPF model
ACPF model
DACPF model
DACPF model
1.1
1.1
1.0
1.0
0.9
0.9
0.8
0.8
0.7
0.7
0.6
EE

0.6
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0.0
0.0
1 3 10 13 28
1 3 node10number 13 28
node number

(b)
(b)
Figure 2. Contrastive analysis of the impacts of removing nodes, (a) Connectivity; (b) Global effective
Figure 2. Contrastive
2. Contrastive analysis
analysis of the
of the impacts
impacts of removing
of removing nodes,nodes, (a) Connectivity;
(a) Connectivity; (b)effective
(b) Global Global
performance.
effective performance.
performance.

4.2. Impact of Wind Power Integration on Power System

4.2. Impact of Wind Power Integration on on Power
Power System
System
From Figure 3, the installed capacity of the wind farm is 100 MW and the sampling interval is
From Figure 3, the the installed
installed capacity
capacity of the wind
wind farm
farm is
is 100
100 MW
MW and
and the
the sampling
sampling interval
interval is
minute level. Its output power varies greatly. In order to improve the ability of power systems to
minute level. Its output power varies greatly. In order to improve the ability of power power systems
systems to
accept wind power, wind farm power output fluctuation is mitigated by the energy storage system
accept wind power,
power, wind
wind farm
farm power
power output
output fluctuation
fluctuation is
is mitigated
mitigated by
by the
the energy
energy storage
storage system
system
with a flexible first-order low-pass filter control strategy [36]. Wind power after smoothing is flat,
with aa flexible
with flexible first-order
first-order low-pass
low-pass filter
filter control
control strategy
strategy [36].
[36]. Wind power after smoothing is flat,
which is shown in Figure 4.
which is
which is shown
shown in in Figure
Figure 4.
4.

0.30
0.30

0.25
0.25
(p.u.)
power(p.u.)

0.20
0.20
Windpower

0.15
0.15
Wind

0.10
0.10

0.05
0.05

0.00
0.00
0 4 8 12 16 20 24
0 4 8 12 16 20 24
t (h)
t (h)

Figure 3. Wind
Figure 3. Wind power
powerover
overthe
thecourse
courseof
ofaaday.
day.
Figure 3. Wind power over the course of a day.
 Energies 2018, 11, 63 8 of 15

before smoothing
after smoothing
Energies 2018, 11, 63 8 of 15
0.30
Energies 2018, 11, 63 8 of 15

0.25 before smoothing

Wind power (p.u.)

after smoothing
0.30
0.20

0.25
0.15
Wind power (p.u.)

0.20
0.10

0.15
0.05

0.10
0.00
0 4 8 12 16 20 24
0.05
t (min)

0.00 Figure 4. Wind power before and after smoothing.

0 4 8 12 16 20 24

In addition, Figure 5 indicates that the tmax (min)30-min wind power fluctuation rate before
smoothing is 25.79%, while the one after smoothing is less than 7%. Therefore, wind power has been
Figure 4.Wind
Windpower
power before
before and
andafter
aftersmoothing.
smoothing.
smoothed effectively. WindFigure
power4.fluctuation rate is calculated as follows.
max(the k ) − min( P1,2 k )
P1,2max
In addition, Figure 5 indicates that 30-min wind power fluctuation rate before
In addition,
smoothing is Figure
25.79%,5while
indicates that
the one =thesmoothing
fafter max 30-min wind
is less power
than fluctuation
7%. Therefore, rate
wind before
power (10)
hassmoothing
been
Prated
is 25.79%, while the one after smoothing is less than 7%. Therefore,
smoothed effectively. Wind power fluctuation rate is calculated as follows. wind power has been smoothed
where P1,2Wind
effectively. power
k is wind fluctuation
power at timerate
k; Pis calculated
r a tmax(
ed
as follows.
isPthe installed
) − min( P
capacity
)
of wind power.
1,2 k 1,2 k
f = (10) 30
In the following analysis, wind power is used to represent the generator node 13 in the IEEE
max( P1,2···k P − min( P1,2···k )
)rated
bus system. Then, the rationality and f =validity of DACPF model are analyzed compared with ACPF(10)
where
model from P1,2section
k is wind
A topower
C. In at time k; some
addition, Pr a t e d is thePrated
installed
significant capacity
analysis of wind power.based on DACPF
is implemented
modelP inInsection
where theis following
D and
wind analysis,
E. at time
power wind
k; power
P isis the
used to represent
installed the generator
capacity of windnode 13 in the IEEE 30
power.
1,2···k rated
bus system. Then, the rationality and validity of DACPF model are analyzed compared with ACPF
model from section A to C. In addition, some significant analysis is implemented based on DACPF
before smoothing
model in section D and E.
after smoothing
0.30
before smoothing
after smoothing
0.25
Wind power (p.u.)

0.30

0.20
0.25
Wind power (p.u.)

0.15
0.20

0.10
0.15

0.05
0.10

0.00
0.05
0 4 8 12 16 20 24

0.00
t (min)
0 4 8 12 16 20 24
Figure 5. Wind power fluctuations before and after smoothing.
t (min)before and after smoothing.
Figure 5. Wind power fluctuations

Figure 5. Wind power fluctuations before and after smoothing.

In the following analysis, wind power is used to represent the generator node 13 in the IEEE 30
bus system. Then, the rationality and validity of DACPF model are analyzed compared with ACPF
model from section A to C. In addition, some significant analysis is implemented based on DACPF
model in section D and E.
Energies 2018, 11,
Energies 2018, 11, 63
63 99 of
of15
15

A. Contrastive analysis of the impact of wind power

A. Contrastive analysis of the impact of wind power
The power grid can accept more wind power after smoothing with smaller power fluctuation.
The power grid can accept more wind power after smoothing with smaller power fluctuation.
Therefore, the generator node 13 is replaced by wind power after smoothing with the same installed
Therefore, the generator node 13 is replaced by wind power after smoothing with the same installed
capacity. As can be obtained from Figure 6, the connectivity and global effective performance of the
capacity. As can be obtained from Figure 6, the connectivity and global effective performance of the
DACPF and ACPF models vary with the change of wind power. Additionally, the former
DACPF and ACPF models vary with the change of wind power. Additionally, the former connectivity
connectivity and global effective performance are larger than the latter ones, that is, the influence of
and global effective performance are larger than the latter ones, that is, the influence of wind power on
wind power on power grid can be better reflected by the DACPF model.
power grid can be better reflected by the DACPF model.

ACPF model
DACPF model
1.0

0.8
g

0.6

0.4

0.2

0 4 8 12 16 20 24

t (h)
(a)

ACPF model
DACPF model
2.0

1.8

1.6

1.4

1.2
E

1.0

0.8

0.6

0.4

0.2

0.0
0 4 8 12 16 20 24

t (h)
(b)

Figure 6. Comparative analysis of the impacts of wind power, (a) Connectivity; (b) Global effective
Figure 6. Comparative analysis of the impacts of wind power, (a) Connectivity; (b) Global
performance.
effective performance.

B. Contrastive analysis of the impact of grid current tolerance capability

The impact of wind power (after smoothing) on power system is displayed in Figure 7, which is
obtained by adjusting the current setting coefficient with a fixed voltage limit. Connectivity and
global effective performance of DACPF model are larger than those of ACPF model when the current
Energies 2018, 11, 63 10 of 15

B. Contrastive analysis of the impact of grid current tolerance capability

The impact of wind power (after smoothing) on power system is displayed in Figure 7, which is
obtained by 11,
Energies 2018, adjusting
63 the current setting coefficient with a fixed voltage limit. Connectivity and global 10 of 15
effective performance of DACPF model are larger than those of ACPF model when the current setting
setting coefficient
coefficient is less
is less than 1.4.than 1.4.current
If the If the current
settingsetting coefficient
coefficient is overis1.4,
over 1.4,are
they they are
the the same.
same. Hence,Hence,
it is
it is more
more meaningful
meaningful for DACPF
for DACPF model model to discuss
to discuss the impact
the impact of wind
of wind power power
on a on a power
power grid grid within
within the
the effective current boundary
effective current boundary condition. condition.

ACPF model
DACPF model
0.30

0.25

0.20
g

0.15

0.10

0.05

0.00

1.0 1.5 2.0 2.5 3.0 3.5 4.0

Current setting coeffic ient

(a)

ACPF model
0.7 DACPF model

0.6

0.5

0.4
E

0.3

0.2

0.1

0.0

-0.1
1.0 1.5 2.0 2.5 3.0 3.5 4.0

Current setting coeffic ient

(b)

Figure 7. Comparative analysis of the impact of grid current tolerance capability, (a) Connectivity; (b)
Figure 7. Comparative analysis of the impact of grid current tolerance capability, (a) Connectivity;
Global effective performance.
(b) Global effective performance.

C. Contrastive analysis of the impact of grid voltage tolerance capability

The influence of wind power (after smoothing) on power grid is studied by adjusting the voltage
setting coefficient at a fixed current limit. As can be seen from Figure 8, connectivity of DACPF and
ACPF model are identical when the voltage setting coefficient is less than 1.056. The former is
superior to the latter if the voltage setting coefficient is over 1.056. Furthermore, global effective
Energies 2018, 11, 63 11 of 15

C. Contrastive analysis of the impact of grid voltage tolerance capability

The influence of wind power (after smoothing) on power grid is studied by adjusting the voltage
setting coefficient at a fixed current limit. As can be seen from Figure 8, connectivity of DACPF and
ACPF model
Energies 2018,are
11, identical
63 when the voltage setting coefficient is less than 1.056. The former is superior
11 of 15
to the latter if the voltage setting coefficient is over 1.056. Furthermore, global effective performance
Therefore,
of the DACPFthe DACPF
model has amodel
largerhas good
value simulation
with the changeresults
of thewithin
voltagethesetting
effective voltage boundary
coefficient. Therefore,
condition.
the DACPF model has good simulation results within the effective voltage boundary condition.

ACPF model
DACPF model

1.2

0.8
g

0.4

0.0
1.05 1.10 1.15 1.20

Voltage setting coefficient

(a)

ACPF model
DACPF model

2.0

1.5
E

1.0

0.5

0.0
1.05 1.10 1.15 1.20

Voltage setting coefficient

(b)

Figure 8. Comparative analysis of the impact of grid voltage tolerance capability, (a) Connectivity; (b)
Figure 8. Comparative
Global analysis of the impact of grid voltage tolerance capability, (a) Connectivity;
effective performance.
(b) Global effective performance.
D. Contrastive analysis of the impact of wind power before and after smoothing
Based on the DACPF model, the generator node 13 is represented by wind power before and
after smoothing in Figure 4, respectively. The connectivity and global effective performance of wind
power after smoothing are greater than the ones before smoothing. In other words, wind power after
smoothing has less effect on power system, which can be obtained from Figure 9. So it is necessary to
smooth the wind power fluctuation from the view of complex network theory.
Energies 2018, 11, 63 12 of 15

D. Contrastive analysis of the impact of wind power before and after smoothing
Based on the DACPF model, the generator node 13 is represented by wind power before and
after smoothing in Figure 4, respectively. The connectivity and global effective performance of wind
power after smoothing are greater than the ones before smoothing. In other words, wind power after
smoothing
Energies 2018, has
11, 63less effect on power system, which can be obtained from Figure 9. So it is necessary
12 of to
15
smooth the wind power fluctuation from the view of complex network theory.

before smoothing
after smoothing
1.0

0.9

0.8
g

0.7

0.6

0.5
0 4 8 12 16 20 24
t (h)
(a)

before smoothing
after smoothing
2.0

1.8

1.6

1.4
E

1.2

1.0

0.8

0.6

0 4 8 12 16 20 24
t (h)
(b)

Figure 9. Comparative analysis of the impacts of wind power before and after smoothing, (a)
Figure 9. Comparative analysis of the impacts of wind power before and after smoothing,
Connectivity; (b) Global effective performance.
(a) Connectivity; (b) Global effective performance.

E. Contrastive analysis of the impacts of different wind power coupling locations

Based on the DACPF model, the impact of different wind power (after smoothing) coupling
locations is analyzed. From Figure 10, node 6 with lager degree and electric betweenness, and node
29 with smaller those are selected as the points of wind power integration, respectively. The
connectivity of node 6 is larger than that of node 29 from 0 h to 2.4 h and from 10.4 h to11.2 h.
Moreover, the global effective performance of node 6 is always higher than that of node 29 in a day.
In a word, the selection of greater degree and electric betweenness node for wind power coupling
Energies 2018, 11, 63 13 of 15

E. Contrastive analysis of the impacts of different wind power coupling locations

Based on the DACPF model, the impact of different wind power (after smoothing) coupling
locations is analyzed. From Figure 10, node 6 with lager degree and electric betweenness,
and node 29 with smaller those are selected as the points of wind power integration, respectively.
The connectivity of node 6 is larger than that of node 29 from 0 h to 2.4 h and from 10.4 h to 11.2 h.
Moreover, the global effective performance of node 6 is always higher than that of node 29 in a day.
In a word, the selection of greater degree and electric betweenness node for wind power coupling
point has
Energies 2018,less impact on power grid performance.
11, 63 13 of 15

node 6
node 29
1.00

0.95

0.90
g

0.85

0.80

0.75

0.70

0 4 8 12 16 20 24
t (h)
(a)

node 6
node 29
1.8

1.7

1.6

1.5
E

1.4

1.3

1.2

1.1

0 4 8 12 16 20 24

t (h)
(b)

Figure 10. Comparative analysis of the impacts of different wind power coupling locations, (a)
Figure 10. Comparative analysis of the impacts of different wind power coupling locations,
Connectivity; (b) Global effective performance.
(a) Connectivity; (b) Global effective performance.

5. Conclusions
The impact of wind power on power grid is discussed with complex network theory in this
paper. Firstly, a DACPF cascading failure model is established on the basis of the IEEE 30 bus system.
Then, comparative analysis of the ACPF and DACPF models is implemented by nodes removal, wind
Energies 2018, 11, 63 14 of 15

5. Conclusions
The impact of wind power on power grid is discussed with complex network theory in this
paper. Firstly, a DACPF cascading failure model is established on the basis of the IEEE 30 bus system.
Then, comparative analysis of the ACPF and DACPF models is implemented by nodes removal,
wind power integration and the change of current and voltage boundaries. The validity of the
DACPF model based complex network theory is verified. In addition, wind power after smoothing
has less impact on power system compared with wind power before smoothing. The reason why
grid-connected wind power should be mitigated is interpreted based on complex network theory.
Finally, selecting larger degree and electric betweenness node as wind power coupling point can
reduce the impact on power systems.

Acknowledgments: This work was financially supported by a grant from the National Key Research and
Development Program of China (No. 2016YFB0900400) and Foundation of Director of Institute of Electrical
Engineering, Chinese Academy of Sciences (No. Y760141CSA) and Jiangsu Province 2016 Innovation Ability
Construction Special Funds (No. BM2016027).
Author Contributions: Yushu Sun and Xisheng Tang conceived and designed the simulation and experiment.
Guowei Zhang provided the simulation and experimental platform. Fufeng Miao worked on the simulation and
the writing of paper. Ping Wang analyzed the data.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. US-Canada Power System Outage Task Force. Final Report of the August 14th Blackout in US and Canada;
United States Department of Energy and National Resources Canada: Washington, DC, USA; Ottawa, ON,
Canada, 2004.
2. National Grid Company. Investigation Report into the Loss of Supply Incident Affecting Parts of South London at
18:20 on Thursday; Technical Report; National Grid UK: London, UK, 2003.
3. Chowdhury, T.; Chakrabarti, A.; Chanda, C.K. Analysis of vulnerability indices of power grid integrated DG
units based on complex network theory. In Proceedings of the IEEE India Conference, New Delhi, India,
17–20 December 2015.
4. Wang, Z.Y.; Chen, G.; Hill, D.J.; Dong, Z.Y. A power flow based model for the analysis of vulnerability in
power networks. Physica A 2017, 460, 105–115. [CrossRef]
5. Luo, G.; Shi, D.G.; Chen, J.F.; Duan, X.Z. Automatic identification of transmission sections based on complex
network theory. IET Gener. Transm. Distrib. 2014, 8, 1203–1210.
6. Chu, C.C.; Iu, H.H.C. Complex networks theory for modern smart grid applications: A survey. IEEE J. Emerg.
Sel. Top. Circuits Syst. 2017, 7, 177–191.
7. Correa, G.J.; Yusta, J.M. Grid vulnerability analysis based on scale-free graphs versus power flow models.
Electr. Power Syst. Res. 2013, 101, 71–79.
8. Fan, W.L.; Liu, Z.G.; Hu, P.; Mei, S.W. Cascading failure model in power grids using the complex network
theory. IET Gener. Transm. Distrib. 2016, 10, 3940–3949.
9. Yin, R.R.; Liu, B.; Liu, H.R.; Li, Y.Q. Research on invulnerability of the random scale-free network against
cascading failure. Physica A 2016, 444, 458–465.
10. Zhang, W.; Pei, W.; Guo, T.D. An efficient method of robustness analysis for power grid under cascading
failure. Saf. Sci. 2014, 64, 121–126.
11. Liu, J.; Xiong, Q.Y.; Shi, X.; Wang, K.; Shi, W. Robustness of complex networks with an improved breakdown
probability against cascading failures. Physica A 2016, 456, 302–309.
12. Zhang, J.H.; Xu, X.M.; Hong, L.; Wang, S.L.; Fei, Q. Attack vulnerability of self-organizing networks. Saf. Sci.
2012, 50, 443–447.
13. Wang, J.W.; Rong, L.L. Robustness of the western United States power grid under edge attack strategies due
to cascading failures. Saf. Sci. 2011, 49, 807–812.
14. Wang, J.W.; Rong, L.L. Vulnerability of effective attack on edges in scale-free networks due to cascading
failures. Int. J. Mod. Phys. C 2009, 20, 1291–1298. [CrossRef]
Energies 2018, 11, 63 15 of 15

15. Peng, X.Z.; Yao, H.; Du, J.; Wang, Z.; Ding, C. Invulnerability of scale-free network against critical node
failures based on a renewed cascading failure model. Physica A 2015, 421, 69–77. [CrossRef]
16. Nicholson, C.D.; Barker, K.; Ramirez-Marquez, J.E. Flow-based vulnerability measures for network
component importance: Experimentation with preparedness planning. Reliab. Eng. Syst. Saf. 2016,
145, 62–73. [CrossRef]
17. Bompard, E.; Pons, E.; Wu, D. Extended topological metrics for the analysis of power grid vulnerability.
IEEE Syst. J. 2012, 6, 481–487. [CrossRef]
18. Bompard, E.; Wu, D.; Xue, F. Structural vulnerability of power systems: A topological approach. Electr. Power
Syst. Res. 2011, 81, 1334–1340. [CrossRef]
19. Yan, J.; He, H.B.; Sun, Y. Integrated security analysis on cascading failure in complex networks. IEEE Trans.
Inf. Forensics Secur. 2014, 9, 451–463. [CrossRef]
20. Lin, Z.Z.; Wen, F.S.; Xue, Y.S. A restorative self-healing algorithm for transmission systems based on complex
network theory. IEEE Trans. Smart Grid 2016, 7, 2154–2162. [CrossRef]
21. Yan, J.; Tang, Y.F.; He, H.B.; Sun, Y. Cascading failure analysis with DC power flow model and transient
stability analysis. IEEE Trans. Power Syst. 2015, 30, 285–297. [CrossRef]
22. Qi, H.B.; Shi, L.B.; Ni, Y.X.; Yao, L.Z.; Masoud, B. Study on power system vulnerability assessment based on
cascading failure model. In Proceedings of the 2014 IEEE PES General Meeting Conference & Exposition,
National Harbor, MD, USA, 27–31 July 2014; pp. 1–7.
23. Chen, G.; Dong, Z.Y.; Hill, D.J.; Zhang, G.H.; Hua, K.Q. Attack structural vulnerability of power grids:
A hybrid approach based on complex networks. Physica A 2010, 389, 595–603. [CrossRef]
24. Dwivedi, A.; Yu, X.H. A maximum-flow-based complex network approach for power system vulnerability
analysis. IEEE Trans. Ind. Inform. 2013, 9, 81–88. [CrossRef]
25. Dey, P.; Mehra, R.; Kazi, F.; Wagh, S.; Singh, N.M. Impact of topology on the propagation of cascading failure
in power grid. IEEE Trans. Smart Grid 2016, 7, 1970–1978. [CrossRef]
26. Sun, Y.S.; Tang, X.S. Cascading failure analysis of power flow on wind power based on complex network
theory. J. Mod. Power Syst. Clean Energy 2014, 2, 411–421. [CrossRef]
27. Sole, R.V.; Valverde, S. Information transfer and phase transitions in a model of internet traffic. Physica A
2001, 289, 595–605. [CrossRef]
28. Erdös, P.; Rényi, A. On the evolution of random graphs. Publ. Math. Inst. Hung. Acad. Sci. 1960, 5, 17–61.
29. Watts, D.J.; Strogatz, S.H. Collective dynamics of “small—World” networks. Nature 1998, 393, 440–442.
[CrossRef] [PubMed]
30. Newman, M.E.J.; Watts, D.J. Renormalization group analysis of the small-world network model. Phys. Lett. A
1999, 263, 341–346. [CrossRef]
31. Monfared, M.A.S.; Jalili, M.; Alipour, Z. Topology and vulnerability of the Iranian power grid. Physica A
2014, 406, 24–33. [CrossRef]
32. Pagani, G.A.; Aiello, M. Towards decentralization: A topological investigation of the medium and low
voltage grids. IEEE Trans. Smart Grid 2011, 2, 538–547. [CrossRef]
33. Barabási, A.L.; Albert, R. Emergence of scaling in random networks. Science 1999, 286, 509–512. [PubMed]
34. Albert, R.; Albert, I.; Nakarado, G.L. Structural vulnerability of the North American power grid. Phys. Rev. E
2004, 69, 025103(R). [CrossRef] [PubMed]
35. Lai, Y.C.; Motter, A.E.; Nisllikawa, T. Attacks and cascades in complex networks. Lect. Notes Phys. 2004, 650,
299–310.
36. Jiang, Q.Y.; Gong, Y.Z.; Wang, H.J. A battery energy storage system dual-layer control strategy for mitigating
wind farm fluctuations. IEEE Trans. Power Syst. 2013, 28, 3263–3273. [CrossRef]

© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).