The Journal of CPRI,

Vol. 10, No. 1, March 2014 pp. 21-26

Node Ordering Scheme of Large Scale Power Systems

Using Sparse Matrix Techniques

Prabhu S*, Chandrasekar S** and Kaliappan P***

Power flow is the basic tool for power system analysis which reveals the system operation in a steady-
state mode for evaluation of the power system planning and operations. The accuracy, simulation
time, computer storage size and convergence of any model used depend largely on the size of the
bus admittance matrix of the system under study. This paper, therefore, presents the study of the bus
admittance matrix of the different systems with sparse techniques. And also analyze the number of
zeros and non-zeros element in the different systems with minimum ordering schemes. The proposed
method is validated using a 5-bus, 30-bus, 118-bus and 300-bus systems. The results are presented in
graphical form and discussed. The sparse Matrix techniques show that as the system is increasing in
size, the percentage of stored bus admittance elements decreases and changing the order of the nodes
gives more impact on the size of the system. Thus, an appreciable reduction in the computer memory
required to store the bus admittance matrix and in turn reduces the overall simulation time.

Keywords: Sparse Matrix Techniques, Non-zero element and LU Factorization.

1.0 INTRODUCTION been carried out by different researchers using

different methods. The most commonly used
Power flow analysis is an important tool during methods for power flow study are Gauss-Seidel,
planning stages of new power system or addition Newton-Raphson and fast decoupled.
to existing ones like adding new generator sites,
meeting increase load demand and locating new Among all these methods, Newton-Raphson
transmission sites [1]. In power flow analysis, is found to be the most widely used model in
some power system parameters are important power systems applications [3]. However, these
in determining the system performance. These conventional power flow algorithms are not
parameters include voltage magnitude and angle efficient due to the challenges they encountered,
at every bus of the system, real and reactive which in most cases, results to divergence of the
power injections at all the buses and power flows algorithms. These challenges include singularity
through interconnecting power channels, reactive of the Jacobian matrix most especially in Newton-
power flows along the transmission lines, real Raphson method, large memory capacity for
and reactive power losses along the transmission storing the elements of the Jacobian and the bus
lines and total losses [2]. In the study of electric admittance matrix, convergence error or problem,
power systems, several different researches have high or large simulation time etc.

* Senior Research Fellow, Power System Division, Central Power Research Institute, Bangalore – 560 080, Karnataka, India.
Email: [email protected], Mob: +91 9611950529
**Principal/Professor, Department of EEE, Gnanamani College of Technology, Namakkal-637018. Tamilnadu, India.
E-mail:[email protected], Mob: +91 9944492300
*** Engg. Officer. Gr.-4, Power System Division, Central Power Research Institute, Bangalore – 560 080, Karnataka. India.
E-mail:[email protected], Mob: +91 9449149924
22 The Journal of CPRI, Vol. 10, No. 1, March 2014

To combat the problems inherent in divergence reduction in the storage handling of the computer
algorithm, various modifications have been used which indicates a substantial improvement
by different researchers in the past [5]. This paper compared to the work reported in literature.
provides an idea of sparse Matrix techniques on This sparsity feature of Ybusmatrix also extends
IEEE bench-mark systems to gives excellent to Jacobian matrix. According to reference [6],
method for computation of the power systems. All “Sparsity can be simply defined to indicate the
practical power systems have majority or larger absence of certain problem interconnections”.
percentage of their buses not connected through Mathematically, the sparsity of an n×n matrix is
transmission lines. This characteristic is explored given by reference [5] as
in this paper such that only non-zero elements
are stored and thereby resulting to savings in
computer memory (the CPU time per iteration ....(1)
is made to be relatively small). The savings in
computer memory is very important when dealing
In a large power system such as the ones
with large practical power systems to reduce the
considered in this work, sparsity may be as
computation time [4] by reducing large memory
high as 97%. Though Ybus is sparse, Zbus is
required for the storage of the elements of bus
full. This sparsity is employed in this work
admittance matrix.
to ensure that only the non-zero elements are
stored and the full characteristic of the original
The nodal admittance matrix of the power network matrix is not lost.
is typically very sparse. However, the equivalent
model after network reduction is usually denser
(less sparse) because the connections between 2.1 Triangular Factorization
retained buses have to be preserved after
reduction. All nodes that were connected to the To solve the generalized jacobian matrix equation
deleted node (boundary nodes) are mutually represented here as
connected, whether they were originally directly
connected or not. In order to get a reduced model
....(2)
that is easier to analyze than the original one it is
necessary to preserve its sparsity. High number
of new branches can decrease the effect of nodes For increments in voltage, the direct method is
reduction on the sparsity of the model and the to find the inverse of [J] and solve for [∆E] from
impedance of new branches can sometimes be
very high leading to numerical problems. The
....(3)
sparse Matrix techniques show that as the system
is increasing in size, the percentage of stored bus
admittance elements decreases and changing the In power systems [J] is usually sparse but [J]-1 is a
order of the nodes gives more impact on the size full matrix. The method of triangular factorization
of the system. solves for the vector [∆E] by eliminating [J]
to an upper triangular matrix with a leading
2.0 SPARSITY SOLUTION TECHNIQUES diagonal and then back-substituting for [∆E] i.e.,
eliminate[1] to
In large power systems, each bus is connected to
only a small number of other buses. Therefore, ....(4)
bus admittance matrix of a large power system is
very sparse. This means that the bus admittance and back-substitute
matrix will contain larger percentage of zeros
as compared to the non-zero elements. This
....(5)
characteristic feature shows a considerable
The Journal of CPRI, Vol. 10, No. 1, March 2014 23

Where of the 5 and 300 bus systems. A Figure 2 and

Figure 5 are the LU factorization yields matrices
[U] = Upper Triangular Factorization of the 5 and 300 bus system. Figure 2 and
[J] = Jacobian matrix Figure 5 notice that two sub diagonals have created
a large number of fills and extend between them
[∆E] = Nodal Voltages and the main diagonal. This factorization gives
more impact on memories and computation time.

The triangulation of the Jacobian is best done by

rows. Those rows below the one being operated
on need not be entered until required. This means
that the maximum storage is that of the resultant
upper triangle and diagonal. The lower triangle
can then be used to record operations. The number
of multiplications & additions to triangulate a
full matrix is 1/3N3, compared to N3 to find the
inverse with sparsity programming the number of
operations varies as a factor of N. If rows are
normalized N further operations are saved.

2.2 Node Ordering Schemes

Node ordering schemes are important in

minimizing the number of multiplications and
divisions required for both L & U triangularization
and forward/backward substitution. A good
ordering will result in the addition of few fills to Fig. 1 Jacobian matrix of 5-bus system
the triangular factors during the LU factorization
process. A fill is a non-zero element in the L or
U matrix that was zero in the original A matrix.

If A is a full matrix, multiplications

& divisions are required for the LU Factorization

process and β=n2 multiplications and divisions are
required for the forward / backward substitution
process. The number of multiplications and
divisions required can be substantially reduces in
sparse matrix solutions if a proper node ordering
is used.

3.0 IMPLEMENTATION

The triangular factorization and Minimum order

scheme is implemented on the IEEE standard bus
system. The below Figure 1 to Figure 3 is for
5-bus system and Figure 4 to Figure 6 is for 300-
bus sytem. The details of the plots as follows. Fig. 2 LU Factorization matrix of 5-bus
system
The Figure 1 and Figure 4 is the jacobian matrix
24 The Journal of CPRI, Vol. 10, No. 1, March 2014

4.0 RESULTS AND DISCUSSION

The results of the simulation, using the proposed

techniques, validated using the standard bus
systems are presented in Table 1 and 2. In the
Table 1, Shows that number of non-zero element
present in the system is compared with the total
element in the full matrix. In the Table 1 shows
that, 5-bus system has 76% ratio compare with
the 300-bus system has only 1.24% variation.

Fig. 3 Reordered LU factors of 5-bus system

Fig. 5 LU Factorization matrix of 300-bus

system

Fig. 4 Jacobian matrix of 300-bus system

The Sparse column minimum degree permutation

schemes are used to minimize the number of fills
in the matrix. The Figure 3 and Figure 6 shows
that having the 49 & 9889 non -zero element in
the 5-bus and 300-bus system. This Minimum
ordering scheme reduces 70% of the non-zero
Fig. 6 Reordered LU factors of 300-bus
element in the system. system
The Journal of CPRI, Vol. 10, No. 1, March 2014 25

TABLE 1 upon the size of the system. If the system size is

Percentage Ratio of non-zero increasing, the no. of zero elements also increases.
element in the Matrix Same as the proper ordering also give more
No. of impact on the size of the system. By using the
No. of Bus ordering scheme, we can minimize the number of
element No. of
in the % Ratio non-zero elements and decrease the computation
in Full non- Zero
system
Matrix element
(Sparsity) time and memory space of the system. So it
(N)
(N2) is concluded that the non-zero element in the
5 25 19 76.00% system is decreasing when the size of the system
30 900 112 12.44% is increasing.
118 13924 476 3.42%
300 90000 1118 1.24%
REFERENCES
In the Table 1 summarized that memory of the [1] Mariesa Crow, Sparse Matrix Solution
system also depends upon the size of the system. Techniques, in Handbook of Computational
Normally, the single element takes 8 bytes in the Methods for Electric Power Systems, CRC
memory of the computer. In the above Table 2 Press, p 81 (2003).
shows that, the 5-bus system has 93.18% ratio
compare with the 300-bus system has only [2] A B Ogundare and A S Alayande, Load
11.82% using minimum ordering scheme. Flow Analysis and simulation using
Newton-Raphson iterative method, Global
journal of Engg. & Tech ., Vol. 2, No.4,
TABLE 2
pp.561-568 (2009)
Percentage Ratio of Memory space
of the LU factors & Minimum [3] T Udomsak, P Padej and K Thanatchai,
ordering Application of load Transfer Technique for
LU Fac- Minimum Distribution Power Flow, World Academy
No. of
torization ordering of Science, Engineering and Technology, p
Bus in the % Ratio
Memory scheme- 860 (2008).
system
space memory
[4] K S Saha and B C Roy ,Sparse Matrix
5 704 Bytes 656 Bytes 93.1%
Solutions in optimal D.C Load-Flow
20352
30 6924 Bytes 34.0% by Crowt Method, International Journal
Bytes
of Recent trends in Engineering,Vol.2,
158084 23300
118 14.7% No.7,pp.66-69( 2009).
Bytes Bytes
1082424 127944 [5] William F Tinney, Mark K, W Tinney and S
300 11.8%
Bytes Bytes Meyer,Solution of Large Sparse Systems by
Ordered Triangular Factorization, I.E.E.E.
From the Table 1 & 2 summarized that the size of Transactions on Automatic Control, Vol.
the system and the reordering of the nodes gives Ac-18, No 4, pp 333-346 (1973).
more impact on the computation of the simulation. [6] A B Ogundare, Application of sparsity
characteristics of power systems to AC
5.0 CONCLUSION Power-Flow Modelling and Simulation,
International Journal of Engineering
Node Ordering Scheme shows that the number Research & Technology, Vol.2,Issue 2
of the zero elements in the system is depends (2013).
26 The Journal of CPRI, Vol. 10, No. 1, March 2014