Structure and Stability of the Indian Power

Transmission Network
arXiv:2206.10366v2 [[Link]-ph] 16 May 2023

Vedang Tamhane1 , G Ambika1,2

1
Indian Institute of Science Education and Research (IISER) Tirupati,
Tirupati-517507, India
2
Indian Institute of Science Education and Research (IISER) TVM,
Thiruvananthapuram-695551, India
E-mail: [Link]@[Link]

Abstract. We present the study on the Indian power transmission network using
the framework of a complex network and quantify its structural properties. For this,
we build the network structure underlying the Indian power grid, using two of its most
prevalent power lines. We construct an equivalent model of an exponential network and
study its structural changes with changes in two parameters related to redundancy and
dead-ends. Then we analyze its stability against cascading failures by varying these two
parameters using the link failure model. This helps to gain insight into the relation
of network topology to its stability, and indicates how the optimum choice of these
parameters can result in a power grid structure with minimum failed links. We apply
the same model to study the robustness of the Indian power grid against such failures.
In this case, we find that when a link connected to a generator fails, it results in a
cascade that spreads in the grid until it is split into two separate stable clusters of
generators and consumers, with over one-third of its nodes nonfunctional.

1. Introduction

Power grids are one of the largest engineering systems among the real-world complex
networks. It has two types of nodes on it, power generators and consumers, connected
with multiple types of connections, which makes it a complex network[1]. The efficient
functioning of power grids requires synchronization in phase and frequency among all
the nodes, which is studied using phase oscillators as nodal dynamics[2]. However, in
spite of the efforts of power engineers to maintain synchrony in such power networks,
failures do occur, causing undesirable disruptions to power transmission. The failure of
a few nodes or links can disturb the synchrony, leading to disruption of the distribution
of power and sometimes overloading. Consequently, the system is disturbed and cannot
fulfill the power requirement of the loads depending on it[3]. In such cases, the power is
supplied to demanding nodes via other paths, which causes overload in the remote nodes
as well, resulting in sequential tripping, known as cascading failure[4]. This cascading
failure is the process that leads to undesirable blackouts that can spread in the entire
Structure and Stability of the Indian Power Transmission Network 2

power grid system[5]. The blackouts of US-Canada in 2003, China in 2005[6], Indonesia
in 2005[7] and India in 2012[8], etc. are all realistic occurrences of the same. Hence,
studies on strategies to contain the failures locally to avoid blackouts are highly relevant.
Due to the dynamic nature of its complex connections, control of power systems
to optimize their performance is a challenge, requiring detailed knowledge of structural
and dynamic stability. Several blackouts around the world in recent history have been
investigated for their causes, and models to analyze cascading failure were proposed to
arrive at emergency measures for prevention[6]. Recent studies indicate that the local
patterns in the network topology influence the ability of the power grid to withstand
cascading failures [9]. In this, using the concept of basin stability, it is shown that
dead-ends diminish the stability in the artificially generated power grid model. The
reported research on cascading failures in [5] explains how networks can evolve towards
optimal topologies that attenuate the cascade. The study by Arianos et al. [10] analyzes
malicious and accidental blackouts in the power grids and their tolerant nature using
complex network theory. They introduce a new parameter, net ability, to quantify the
performance of power grids. In [11], a forecasting method is proposed to help identify
critical transmission lines and components in the European power grid network. In [12]
and [13], the harmonic closeness approach is used to define load on the link, and they
obtain values of a tunable parameter that can be used in distributing load among links
of the network. A study on cascading failures is conducted on weighted networks in
which a local weighted flow redistribution rule is implemented [14].
The Indian power grid is managed by Power System Operation Corporation
Ltd. (POSOCO)([Link] It is the single large grid that keeps
the entire nation under electrification. It is composed of nearly 1575 important
grid elements, which include power generation plants, distribution substations and
consumers, connected via around 2200 connections. These connections are added in
the cost-minimizing and simplest way of the tree-like scheme but carry different levels
of electric power. The Indian power grid underwent one of its major blackouts in recent
history in 2012, affecting three of its five regional grids. During maintenance, the outage
in the Bina-Gwalior 400kV line led the entire grid to risk by overloading one of the North
region grid stations beyond its safe limits. This event caused the cascade to spread in
the East and the North-East regional grids.
In spite of being one of the most extensive grids in the world, the Indian power grid
remains understudied. A complex network approach reported to investigate the blackout
of 2012 in the Indian power grid[8] uses a practical model considering the active and
reactive power loads with the locally preferential load redistribution rule. In [16], the
structural vulnerability of the Indian power grid and random attacks on the same are
discussed. The other studies reported include the topology and network measures of the
power transmission network of Odisha[17] and West Bengal power transmission system
by H Das et al. [18]. While the work of H Das et al. covers only a particular region,
that by Zhang et al.[8] dates back to the last decade. So, we find more detailed studies
of its present structure, are required to understand the subtle features underlying the
Structure and Stability of the Indian Power Transmission Network 3

whole Indian power grid.

In this context, we present a detailed study of the Indian power grid using a complex
network approach to understand its topology and compute its network measures. We
first construct the Indian power grid network using the data of July 2021 made publicly
available by POSOCO([Link] establish the exponential nature of its
degree distribution. We compare its network measures with those of other well-studied
power grids and standard complex networks. Using a synthetically generated network,
equivalent to the Indian Power Grid, we study how the presence of dead-ends and
redundancy affect the structure and topology of the network. The study of cascading
failures is then carried out using the link failure model that helps to understand the
dependence of its severity on the network’s topology. Specifically, we study the roles
of redundancy and dead-ends in the spread of cascading failures considering two types
of initial failures. Then we apply the same method to the actual Indian power grid
structure and analyze the extent of failed links and spread of cascades for the 400kV
and the 200kV grids.
The paper is organized as follows. We present the construction of the Indian power
grid from data and its structural characteristics in Section 2. Next, we provide details on
the construction of a synthetic exponential power grid and its dependence on features
like redundancy and dead ends(Section 3). In Section 4, we present the link-failure
model for the spread of cascade on the constructed network. This is then applied to the
Indian Power grid in Section 5, and we end with the summary and conclusion in Section
6.

2. Structure of the Indian power grid

The Indian power grid system is divided into five major regional grids, namely western,
eastern, southern, northern and north-eastern regional grids. They are interconnected to
form a single large grid with power lines of 66kV, 100kV, 110kV, 132kV, 220kV, 400kV,
765kV and HVDC(high voltage direct current) for transmitting the electric power among
the different stations. The 220kV and 400kV lines are the most abundant transmission
lines in the power grid of India that connect important grid elements.
To construct the Indian power transmission network, we refer to the lists of
important grid elements that are made publicly accessible by the Regional Load
Dispatch Centers of POSOCO([Link] Since the data of the Northeastern
power grid is not available, we use the maps provided by the Central Electricity
Authority(CEA)([Link] that administer the working of the grid elements
in this region. The connections are then added to the edge list using the breadth-first
search algorithm[19] to get the adjacency matrix of the network.
The diversity of the connections in the Indian power grid is such that a particular
kV line does not connect all the nodes. We observe that there are small clusters in the
network that are either connected by 132kV lines and lower and 765kV lines or groups of
small-scale power generation plants and consumers. But 220kV and 400kV transmission
Structure and Stability of the Indian Power Transmission Network 4

10

Figure 1: The Indian power transmission network with 400kV lines as links. The size
of the node is proportional to its degree, while its color indicates its value.

lines form the largest clusters covering the entire network. The structures of the networks
thus constructed with 400kV lines and 220kV lines as links(excluding a few isolated small
clusters) are presented in Figures 1 and 2 respectively, while the network including both
the lines is given in Figure 3. We use the standard methods using adjacency matrix
[20] to compute the relevant measures like degree distribution, average degree, Link
Density(LD), Clustering Coefficient(CC), Average Shortest Path Length(ASP L) and
General Efficiency(GE) for a complete characterization of its topology.
The degree distributions P (k) of the Indian Power Transmission network, including
both 220kV and 400kV transmission lines and that of the separate networks, are then
computed. We plot Log(P (k)) for these networks in Figure 4. Its linearly decreasing
nature indicates the exponential nature of the network. Therefore, we can write,
k
P (k) ≈ e− γ (1)
where k is the degree and γ is the scaling index. The plots are linearly fitted as shown
in Figure 4 with dotted lines which gives γ = 1.78 for the full network. When estimated
separately, γ comes out as 2.3 for 220 kV and 2.5 for 400 kV networks.
Structure and Stability of the Indian Power Transmission Network 5

10

Figure 2: The Indian power transmission network with 220kV lines as links. The size
of each node is proportional to its degree, while its color indicates its value.

We compare γ and other topological measures like CC, LD, k̄ and ASP L of the
Indian power grid with those of two other well-studied power grids, the Italian power
grid and Western US power grid in Table 1. In addition, we include the Global Efficiency
value of the largest connected component of the Indian power grids. We also present the
measures of other standard network topologies like scale-free, small-world and random
networks of the same sizes(N = 1575) for comparison. For this, the standard networks
are generated using, (i) Barabási-Albert model with the initial number of nodes m =
2 for a scale-free network, (ii) Watts-Strogatz model starting with 2 nearest neighbors
and probability of rewiring p = 0.2 for a small-world and (iii) Erdös-Rényi graph with
the probability of connection p = 0.005 for a random network[20]. We note, compared
to other standard networks with similar link density, the Indian power grid has higher
CC and ASP L.
Structure and Stability of the Indian Power Transmission Network 6

16

14

12

10

Figure 3: The Indian power transmission network with both 220kV and 400kV lines
included as links. The links in blue represent 220kV lines, those in red represent 400kV
lines, and the yellow ones represent connections with both 220kV and 400kV lines. The
size and color of the node relate to their degree as in earlier figures.

3. Construction and Characteristics of Equivalent Exponential Network

We construct a synthetic exponential network that is topologically equivalent to the

Indian power grid using the method described in [22]. This requires defining the
parameters, total number of nodes N , number of initial nodes N0 , probability of
connections p and q, a measure of redundancy r and measure of dead-ends s.
Following the method of construction in [22], we start with a minimum spanning
tree using Prim’s algorithm. Thus the network is initially coarse and tree-like that covers
the spatial extents with minimal costs. During the growth, links are added to optimize
the heuristic function f (i, k, G), estimating the trade-off between cost and redundancy.
(dG (i, j) + 1)r
f (i, j, G) = (2)
dspatial (i, j)
Structure and Stability of the Indian Power Transmission Network 7

220kV
0.5 400kV
220kV and 400kV
1.0
Log(P(k))

1.5

2.0

2.5

3.0
2 4 6 8 10 12 14 16
Degree(k)

Figure 4: Semi-logarithmic plot of degree distributions of the Indian Power Transmission

networks including both 220kV and 400kV transmission lines and separate networks.
The linear fits indicating their exponential nature are shown using dotted lines in the
same color.

Network N CC GE LD k̄ ASP L γ
Indian Power Grid
220kV 645 0.10 0.101 0.0022 2.33 13.00 2.3
400kV 665 0.11 0.102 0.0038 2.66 14.45 2.5
220kV+400KV 1575 0.12 0.109 0.0018 2.76 11.33 1.78
Italian power grid[21] 341 0.06 - - 2.46 16.18 1.8
Western US PG[22] 6594 0.08 - - 2.67 18.99 2.12

Scale free 1575 0.0267 - 0.0025 3.99 4.22 NA

Small world 1575 0.0005 - 0.0015 2.36 13.43 NA
Random 1575 0.0045 - 0.0025 7.79 4.22 NA

Table 1: Characteristic measures of The Indian power grid networks compared with
other power grids and standard networks.

where i and j are any two nodes of the network, dG (i, j) is the shortest path length
between the nodes i, and j. dspatial (i, j) is the spatial distance between the nodes i and
j.
Based on a random number S ∈ [0, 1] generated, we construct the network as
Structure and Stability of the Indian Power Transmission Network 8

follows. If S > s, a node i is chosen from the set of geometric coordinates, and a link
is added to the node j in the existing network for which dspatial (i, j) is minimum. Then
with probability p, a node k is identified such that f (i, k, G) is maximum and a link
i − k is added. With probability q, another pair of unconnected nodes i0 and j 0 is chosen
for which f (i0 , j 0 , G) is maximum and link i0 − j 0 is added. Hence topology changes from
cost-effective to redundant depending on the value of r.
If S < s, an existing link i − j is removed. A new node h is added at the midpoint
of i and j, with connections, i − h and j − h. This node gets degree k=2 while other
nodes retain their degrees. Moreover, this can form ring or chain-like structures in the
network, which results in less number of dead ends in the network. We repeat the above
steps till we reach the total number of nodes in the networks as the chosen N .
For the Indian power grid, we estimate the parameters r = 0.1 and s = 0.3 for
the 400kV grid of size 665 and r = 0.15 and s = 0.38 for the 220kV grid of size 645.
So, to construct a topologically equivalent network, we choose N = 655, N0 = 100,
r = 0.1 and s = 0.3 and optimize probabilities as p = 0.3 and q = 0.2 to get
γ = 2.4, close to the values of these power grids. We then construct networks with
(N, N0 , p, q) = (655, 100, 0.3, 0.2) and vary the parameters r ∈ [0, 1.2] and s ∈ [0, 0.5] to
study how they influence the topological features of the network. (For values of r > 1.2
and s > 0.5, we observe that the networks lose their characteristic exponential degree
distribution).
In Figure 5, we show the changes in CC, GE and k̄ with redundancy parameter r,
keeping s = 0.3. The dotted lines indicate their trends. We observe that as r increases,
CC decreases while GE shows an increasing trend. By the method of construction,
a new node is always linked to two more, giving equal weight to spatial and network
distance. Also, in every step, two further nodes become directly linked. This produces
more long-range connections in the structure. Increasing the parameter r increases such
distant redundant connections, and the local clustering coefficient, and therefore CC,
decreases with r. At the same time, the average shortest path between two distant
nodes decreases. This results in GE increasing with r, while the average degree remains
more or less unaffected.
Figure 6 shows the changes in CC, GE and k̄ with dead-ends parameter s, keeping
r = 0.1. In this case, we find that these measures show decreasing trends. Since
increasing s can form ring or chain-like structures in the network, it results in a larger
average shortest path of the network with less number of triangular connections. So, CC
and GE decreases with increase in s. As explained in [22], the average degree depends
on s in such a way that it will decrease as s increases.

4. Model for Cascading Failure

The most relevant issue in the context of a power grid is its stability against local
or global perturbations and natural or man-made failures. Most often, the cascading
failure, that can lead to blackouts can be local or system-wide, depending on their
Structure and Stability of the Indian Power Transmission Network 9

0.148
0.08
0.146
0.07
0.144
0.06
0.142

GE
CC

0.05 0.140
0.04 0.138
0.03 0.136
0.02 0.134
0.0 0.2 0.4 0.6 0.8 1.0 1.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2
r r

(a) (b)

2.490

2.485

2.480
k

2.475

2.470

0.0 0.2 0.4 0.6 0.8 1.0 1.2

r

(c)

Figure 5: Variation of (a)Clustering Coefficient(CC) (b)Global Efficiency (GE)

(c)average degree(k̄) as redundancy parameter r is varied keeping dead-ends parameter
at s = 0.3. The average values from 30 realizations of the constructed exponential
networks are shown in each case, with their standard deviations as error bars(in red).

origin and the topology of the network. If a line connecting a small degree node is
overloaded, it results in an outage in a particular locality only. Hence the blackout
is localized and does not affect the remote connections. On the other hand, if higher
degree nodes, which are most often generators, get disconnected, then redistribution of
power flow causes overload in remote connections, also leading to sequential failures.
Such cascading failures have various possible sources. A few of them are listed below:
• The redistribution of the load on a node or a link which leads to overload[23] or
under-load
• The existence of direct dependency, where if a node becomes dysfunctional, all
nodes that are dependent on it also become dysfunctional [24]
• The number of functioning adjacent nodes is above or below threshold[25].
There are several models proposed for the study of cascading failure using
Structure and Stability of the Indian Power Transmission Network 10

0.165
0.11 0.160
0.10 0.155
0.09 0.150
0.08 0.145

GE
CC

0.07 0.140
0.06 0.135
0.05 0.130
0.04 0.125
0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3 0.4 0.5
s s

(a) (b)

2.70
2.65
2.60
2.55
k

2.50
2.45
2.40
2.35
0.0 0.1 0.2 0.3 0.4 0.5
s

(c)

Figure 6: Variation of (a)Clustering Coefficient(CC) (b)Global Efficiency (GE)

(c)average degree(k̄)as a function of s for r = 0.1. The average values are plotted
from 30 realizations of the constructed exponential networks. The error bars in this
case are not visible as they are very small.

simulations in AC power grids, like link overload-based models as well as node failure-
based models. In the node failure model reported [26], the tolerance parameter T is
calculated for each node, such that if it goes beyond a critical threshold TC , the node
is considered to be dysfunctional. The link overload-based model reported recently[23],
assumes that the power is exchanged along the shortest path between any two nodes
of the given network. Since the power flow from one node to another one does not
always take place along the shortest paths, in [27], an improved version of this model is
introduced, which is the method followed in the present study. We briefly discuss below
the steps involved in analyzing cascading failure using this model and apply them to
the synthetic network and the Indian power grid.
In this model, the input power at each node of the grid is taken using the following
Structure and Stability of the Indian Power Transmission Network 11

equation for active power Pi [27].

X |Vi ||Vj |
Pi = aij sin θij (3)
j xij

Here, aij is the element of the adjacency matrix, θij is the angular difference between the
voltages of two nodes Vi and Vj . xij = x0 .dspatial (i, j) is reactance with x0 being specific
reactance and dspatial (i, j), the length of the transmission line. We consider the value of
the specific reactance to be x0 = 0.265Ω/km[28]. Since we are considering high-voltage
AC grids, resistance in transmission lines can be conveniently neglected.
Following this, we calculate the power transmitting through each link of the grid at
every time step by solving the nonlinear equation (3) to calculate the θij . We use it to
calculate the power flow in the transmission line connecting nodes i and j as Fij given
by
|Vi ||Vj |
Fij = sin θij (4)
xij
Ideally, total power in the power grid should equate to zero for conservation, i.e.,
P
i Pi = 0 [27]. It is required for the optimum loading on each of the nodes and its
stable function. In cases of disturbance, this sum may not be zero at some nodes, which
leads to an overload in transmission lines. Up to a certain extent, transmission lines can
tolerate the excess power flow without failing. The critical power flow that the link can
tolerate is given by Cij

Cij = (1 + α)Fij (t0 ) (5)

We set the tolerance parameter α = 0.3 for every edge in the network in equation (5).
This would mean that every link can bear the overload up to 1.3 times the safe power
limits at initial time t0 .
We apply this method first to the synthetic exponential network constructed, as
mentioned in section III. From available data, we estimate the number of generators or
power distribution stations in the Indian power grid to be approximately 1/5 of that
of the total number of nodes. So, we take N/5 high-degree nodes of the network as
generators, with the input power positive and the rest as consumers with negative input
power. Also, in the initial step, we take Pi = 1 for generators and Pi = −1/4 for
consumers so that the net power in the network is zero.
We study the cascading failures in the synthetic power grid network with two types
of initial failing links. In the first case, the link connecting two high-degree nodes,
specifically a generator to a generator, is failed. In the other case, we choose a link that
fails as the one connecting a dead-end node to the power grid. At every step, to make
the net power equal to zero, we regulate the power at each node using the equation
1 X
Pi0 = Pi − Pj (6)
N (ζ) j∈ζ
Structure and Stability of the Indian Power Transmission Network 12

0.0500
0.0498
0.0496
Lf/Lt

0.0494
0.0492
0.0490
0.0488
0.0 0.2 0.4 0.6 0.8 1.0 1.2
r

Figure 7: Variation in the extent of cascading failure in exponential networks as a

function of redundancy parameter r. Here the mean values of the ratio of failed links,
Lf /Lt , are shown for 30 realizations of the constructed network with s = 0.3, with
standard deviations as error bars(in red).

where N (ζ) is number of nodes in component ζ. Then we calculate the flow in each link
and the link for which Fij > Cij is removed. We repeat the same procedure until there
is no link satisfying this condition.
We choose the parameters of synthetically constructed exponential networks as
(N, N0 , p, q) = (655, 100, 0.3, 0.2), equal to the estimated ones for the Indian power grid.
The distances dspatial (i, j) are taken to be of the same order of magnitude as the Indian
power grid and the load |Vi | = |Vj | = 400kV . We vary r ∈ [0, 1.2] and s ∈ [0, 0.5] and
study the dependence of the spread of failure on the topology of the network. From 30
realizations, we compute the ratio of failed links to total links, Lf /Lt , that will give a
measure of the extent of the spread of failure over time.
When a link connecting a generator to a generator is chosen to initiate the failure,
the values of (Lf /Lt ) with change in parameter r, keeping s = 0.3 are shown in Figure
7. We observe that this ratio first decreases up to a certain value and then increases,
indicating a minimum value at r = 0.8. This is because when r is small, with less
number of redundant connections, the excessive power cannot be redistributed among
its links, which results in the failure of many links. On the other hand, on networks
with large r and having too many redundant connections, the failure spreads across the
network. In between, there can be a value of r for the given network at which Lf /Lt
can be small, indicating less damage to the network.
In the same manner, we study the effect of parameter s on the failure ratio of links
Structure and Stability of the Indian Power Transmission Network 13

0.051

0.050
Lf/Lt

0.049

0.048

0.0 0.1 0.2 0.3 0.4 0.5

s

Figure 8: Variation in the extent of cascading failure in exponential networks as a

function of dead-end parameter s keeping r = 0.1. The mean values of the ratio of
failed links, Lf /Lt from 30 realizations of the constructed network are plotted with
standard deviations as error bars(in red).

keeping the value of r fixed to 0.38. As shown in Figure 8, it has an increasing trend
with parameter s. Since large s corresponds to less number of dead-ends, it is clear that
the failure cannot be contained, and the number of failed links increases.
However, when a link connecting dead-end to the main network is set to fail initially,
a maximum of two links and a few nodes and links that are directly dependent on it
are isolated from the main network, leaving the remaining network unaffected. We note
that the values of r and s that give a minimum (Lf /Lt ) may also depend on the size of
the network, chosen values of N0 , tolerance parameter α and the initial failing link. Our
results indicate that for a given power grid, an optimum redundancy and dead-ends can
be found to minimize the extent of cascading failure.

5. Cascading failure on the Indian power grid

Once the stability of the synthetic exponential networks is studied and quantified, we
apply the same link failure model to the Indian power grid to analyze its robustness. We
use the networks of 220kV lines and 400kV lines and apply the same method described
in the previous section. Since the data on lengths of transmission lines is not available
for the North-Eastern regional power grid, that is not included here.
We choose the tolerance parameter α = 0.3 and randomly choose N/5 nodes with
higher degrees to be generators. We take |Vi | = |Vj | = 220kV and |Vi | = |Vj | = 400kV
Structure and Stability of the Indian Power Transmission Network 14

for 220kV and 400kV networks respectivly. We simulate cascading failures as described
in the previous section, and the extent of link failures are shown in Figures 9 and 10 for
the 220kV and 400kV networks. Here failed links are shown in red dotted lines on the
network with active links in green. For the 220kV grid, we find that Lf /Lt = 0.045. For
the 400kV grid, we calculate this ratio to be 0.038, making it slightly more robust than
the 220kV network. The spread of the cascade on the network in time is compared for
both these networks in Figure 11 where the changes in Lf /Lt are plotted against time
steps. The ratio saturates in time as the spread of overloading stops. We note the ratio
is higher but saturates earlier for the 220kV grid compared to the 400kV grid. Thus, the
400kV grid fails slowly but to a lesser extent compared to the 220kV grid. In both these
cases, we note that evenly distributed generators help these networks to separate into
two stable and functioning clusters, each having generators and consumers. However, in
this process, over 200 nodes and associated links are isolated from the two main clusters,
and the cascade terminates.
We carry out a detailed analysis to identify links that can cause an extensive spread
of cascades. For this, we repeat the same analysis by considering each link in each
network as the starting link to fail and compute the corresponding Lf /Lt values. We
find the links connecting to dead ends, with one end being a node of degree ki = 1, are
non-critical. The Lf /Lt value is minimum for each of them and the cascade does not
spread in the network. In 220kV network, we find 15% links are of this type, which leads
to the value of Lf /Lt = 0.002. In the 400kV network, we find 21% links connecting dead
ends that give Lf /Lt = 0.001. On the other hand, links connecting high-degree nodes,
with degrees more than or equal to 5 on either or both sides, are critical and they most
often connect generator to generator. In the 220kV network, there are 5% links that
give Lf /Lt values in the range of 0.039 to 0.055, and 2% links in 400kV with Lf /Lt in
the range 0.025 to 0.035. We also find links with tree-like structures on either or both
sides that can be identified by the local clustering coefficients of their connecting nodes.
In 220kV network we have 63% of such links, with Lf /Lt in (0.013, 0.035) and in 400kV
network, there are 50% of such links with Lf /Lt in (0.006, 0.025). However, there are
also links that connect nodes, with degrees equal to 2 or 3 on either side but result in
higher values of Lf /Lt . We find they lie on the shortest paths between two generators.
The remaining links mostly have clustered components on both sides and, depending
on the extent of clustering, can result in a range of Lf /Lt values, as 0.01–0.039(220kV )
and 0.004 − 0.025(400kV )

6. Conclusion

From the data available with POSOCO(as on July 2021), the power transmission
network of India is constructed with 220kV and 400kV transmission lines and its
structural characteristics are computed. We find that the Indian power grid network
is exponential in nature with scaling index γ = 1.78, similar to the Italian power
grid(γ = 1.8) and Western US grid (γ = 2.21).
Structure and Stability of the Indian Power Transmission Network 15

Figure 9: Cascading failure in the 220kV transmission network of the Indian power grid,
where 0.045 of the total links are failed (shown in red dotted line)

To understand the efficient and stable functioning of the power grids, it is

important to calculate the dependence on the features of the network, like redundant
connections and dead-ends. For this, we construct a synthetic exponential network
that is topologically equivalent to the Indian power grid and report how the network
measures change with redundancy and dead-ends. With increasing redundancy, CC
of the network decreases while GE shows an increasing trend. With an increase in
s or a decrease in the number of dead-ends in the network, CC, GE and k̄ exhibit
decreasing trends. These trends in the network measures are important to understand
the dependence of the cascading failures in the structure of power grid networks.
We study the spread of cascading failure due to the failure of a link connecting
one generator to the other in the synthetic exponential network and its robustness as
network features are changed. We find that for a given size of the network, it is possible
to obtain optimum values for parameters related to redundancy and dead-ends that
Structure and Stability of the Indian Power Transmission Network 16

Figure 10: Cascading failure in the 400kV transmission network of the Indian power
grid. Here the failed links(shown in red dotted line) are 0.038 of the total links.

can minimize the failure ratio (Lf /Lt ) and hence provide maximum resilience against
cascades. On the contrary, if a link connecting a dead-end with the main cluster fails, it
leads to a local blackout only without affecting the stability of the remaining network.
We apply the link failure model to study cascading failures in the Indian power grid.
We find the structure of the Indian power grid is such that 4.5% of links in the 220kV
network and 3.8% of links in the 400kV network are seen to fail when N/5 nodes are
taken as generators. Moreover, the networks form two stable and functioning clusters.
We find the spread of the cascade is maximum when the link that fails first is the
one connecting generator to a generator in the network. The spread is minimum when
it starts with links that connect to dead ends. The extent of spread also depends on
whether the starting links are the ones that connect to tree like structures on either or
both ends or are connected to clustered components.
Our study on the synthetically generated exponential network and the Indian
Structure and Stability of the Indian Power Transmission Network 17

0.05
220KV
400kV
0.04

0.03
Lf/LT

0.02

0.01

0.00
0 200 400 600 800 1000 1200 1400
t

Figure 11: Variation of the link failure ratio, Lf /Lt with time in 220kV network and
400kV network. At any time step, the extent of failure in the 400kV network is less
compared to the other one, but it takes more time to reach saturation.

Power grid brings out how the spread of failures depends on the network structure and
topological features. This can guide engineers to choose a proper tolerance parameter
to avoid the probability of small-scale and large-scale cascading failures and also to
arrive at the most efficient way of adding nodes or extending connections for optimum
performance.

7. Data Availability

The data used is obtained from POSOCO’s individual regional load dispatch centers
(RLDCs)([Link] as on July 2021, for Eastern, Northern, Western and
Southern regions. For North Eastern region, the data was collected from the Central
Electricity Authority, Ministry of Power, Government of India([Link]

Acknowledgements
One of the authors(VT) acknowledges the Department of Science and Technology(DST)
India for INSPIRE scholarship and IISER Tirupati for facilities during MS thesis work.

References
[1] G. A. Pagani and M. Aiello, “The power grid as a complex network: a survey,” Physica A: Statistical
Mechanics and its Applications 392, 2688–2700 (2013).
[2] J. Grzybowski, E. Macau, and T. Yoneyama, “On synchronization in power-grids modelled as
networks of second-order Kuramoto oscillators,” Chaos: An Interdisciplinary Journal of Nonlinear
Science 26, 113113 (2016).
Structure and Stability of the Indian Power Transmission Network 18

[3] J.F. Zheng, Z.Y. Gao, and X.M. Zhao, “Clustering and congestion effects on cascading failures of
scale-free networks,” Europhys. Lett., Vol.79, pp.58002-1–58002-5, 2007.
[4] A. Sciré, I. Tuval, and V.M. Egúluz, “Dynamic modeling of the electric transportation network,”
Europhys. Lett., Vol.71, No.2, pp.318–324, 2005.
[5] L. D. Valdez, L. Shekhtman, C. E. La Rocca, X. Zhang, S. V. Buldyrev, P. A. Trunfio, L. A.
Braunstein, and S. Havlin, “Cascading failures in complex networks,” Journal of Complex Networks
8, cnaa013 (2020)
[6] J. Wang, “Power grid cascading failure blackouts analysis,” in AIP Conference Proceedings, Vol.
2066 (AIP Publishing LLC, 2019) p. 020046.
[7] Y. Nurdin, “Understanding the cascading failures in Indonesian power grids with complex network
theory,” in 2019 IEEE International Conference on Cybernetics and Computational Intelligence
(CyberneticsCom) (IEEE,2019) pp. 50–55.
[8] G. Zhang, Z. Li, B. Zhang, and W. A. Halang, “Understanding the cascading failures in Indian
power grids with complex networks theory,” Physica A: Statistical Mechanics and its Applications
392, 3273–3280 (2013).
[9] P. J. Menck, J. Heitzig, J. Kurths, and H. Joachim Schellnhuber, “How dead ends undermine power
grid stability,” Nature communications 5, 1–8 (2014).
[10] S. Arianos, E. Bompard, A. Carbone, and F. Xue, “Power grid vulnerability: A complex network
approach,” Chaos: An Interdisciplinary Journal of Nonlinear Science 19, 013119 (2009).
[11] Schäfer, B., Witthaut, D., Timme, M. et al. Dynamically induced cascading failures in power grids.
Nat Commun 9, 1975 (2018). [Link]
[12] Yucheng Hao, Yanhui Wang, Limin Jia, Zhichao He. Cascading failures in networks with the
harmonic closeness under edge attack strategies. Chaos, Solitons & Fractals,135, 2020, 109772.
[13] Hao Y, Jia L, Wang Y: Robustness of weighted networks with the harmonic closeness against
cascading failures. Physica A: Statistical Mechanics and its Applications 541:123373, 2020.
[14] Chen G, Wang W: Universal robustness characteristic of weighted networks against cascading
failure. Physical Review E 77:26101, 2008.
[15] [Link]
[16] P. Panigrahi and S. Maity, “Vulnerability Analysis of Weighted Indian Power Grid Network
Based on Complex Network Theory,” 2017 14th IEEE India Council International Conference
(INDICON), Roorkee, India, 2017, pp. 1-6, doi:10.1109/INDICON.2017.8487727.
[17] H. Das, S. K. Mishra, and D. S. Roy, “The topological structure of the Odisha power grid: a
complex network analysis,” IJMCA 1, 012–016 (2013).
[18] H. Das, G. S. Panda, B. Muduli, and P. K. Rath, “The complex network analysis of power grid:
a case study of the West Bengal power network,” in Intelligent Computing, Networking, and
Informatics (Springer, 2014) pp. 17–29.
[19] J. R. Patel, T. R. Shah, V. P. Shingadiya, and V. B. Patel, “Comparison between breadth-first
search and nearest neighbor algorithm for waveguide path planning,” Int. J. Research and Scientific
Innovation 2, 19–21 (2015).
[20] M. E. Newman, “The structure and function of complex networks,” SIAM review 45, 167–256
(2003).
[21] P. Crucitti, V. Latora, and M. Marchiori, “A topological analysis of the Italian electric power
grid,” Physica A: Statistical mechanics and its applications 338, 92–97 (2004)
[22] P. Schultz, J. Heitzig, and J. Kurths, “A random growth model for power grids and other spatially
embedded infrastructure networks,” The European Physical Journal Special Topics 223, 2593–2610
(2014).
[23] A. E. Motter and Y.-C. Lai, “Cascade-based attacks on complex networks,” Physical Review E
66, 065102 (2002).
[24] R. Parshani, S. V. Buldyrev, and S. Havlin, “Critical effect of dependency groups on the function
of networks,” Proceedings of the National Academy of Sciences 108, 1007–1010 (2011).
[25] G. J. Baxter, S. N. Dorogovtsev, A. V. Goltsev, and J. F. Mendes, “Bootstrap percolation on
Structure and Stability of the Indian Power Transmission Network 19

complex networks,” Physical Review E 82, 011103 (2010).

[26] J.-W. Wang and L.-L. Rong, “Cascade-based attack vulnerability on the US power grid,” Safety
science 47, 1332–1336 (2009).
[27] A. Plietzsch, P. Schultz, J. Heitzig, and J. Kurths, “Local vs. global redundancy–trade-offs between
resilience against cascading failures and frequency stability,” The European Physical Journal
Special Topics 225, 551–568 (2016).
[28] Machowski, Z. Lubosny, J. W. Bialek, and J. R. Bumby, Power system dynamics: stability and
control (John Wiley & Sons, 2020).