You are on page 1of 6

Performance Analysis of Static Network Reduction

Methods Commonly Used in Power Systems


Syed Mohammad Ashraf, Bhavna Rathore, and S. Chakrabarti, Senior Member, IEEE
Department of Electrical Engineering, Indian Institute of Technology Kanpur, India
smashraf@iitk.ac.in, bhavnar@iitk.ac.in, saikatc@iitk.ac.in

Abstract— Network reduction techniques are useful means area for the internal subsystem under consideration. For
of analyzing large, interconnected power systems. The most running power system analysis functions in the internal area,
desirable property of a reduced network is that it should the system is typically modelled in detail. The external areas
represent the original network as accurately as possible. This are usually represented by simple models, referred to as the
paper explores some of the commonly used static network
external equivalent system.
reduction techniques, such as Ward reduction, Kron
reduction, Dimo’s method, and Zhukov’s method. The At certain intervals, the interconnected utilities exchange
performances of the reduced networks are evaluated in terms detailed information regarding the network topology and
of their ability to follow the bus voltages of the original operating conditions, based on which the external network
network, with changes in operating conditions. The IEEE 14- equivalents are determined. However, the amount of such
bus and 118-bus systems are taken as the test beds. data exchange for a large interconnected system can be
prohibitively large, requiring extensive investment in the
Keywords— Dimo’s method, Kron reduction, Ward communication infrastructure. From the computational point
reduction, Zhukov’s method. of view also, it requires significant resources to work with
detailed models of all the external areas, for various power
I. INTRODUCTION
system studies. The usual practice is that the external
Modern power systems have become highly inter- network equivalents are updated for small changes in the
connected to maintain acceptable levels of reliability and operating conditions in the external network, usually by
quality of the supply. These networks are now being following some network parameter updation methodology.
operated under heavily loaded conditions in order to meet the For major changes in the operating condition or topology in
ever-growing demand for electricity, with the limited
the external areas, the equivalents need to be computed
generation, transmission, and distribution resources, and to
afresh.
maximize the economy. During the past few decades, the
Network reduction techniques can be broadly classified
overall characteristics of the power system generations and
as static and dynamic, based on the representation of the
loads have changed significantly due to increasing
model and its intended use [1-2]. Static reduction techniques
penetration of renewable energy sources, large-scale use of
are used for static analysis only, such as power flow
power-electronic converter interfaced sources and loads, and
calculations and system operation and planning studies.
competitive electricity markets. Special attention, therefore,
Whereas, for the analysis of system dynamics the reduced
has to be given on the modelling of the power system for
dynamic equivalents of the external areas are required. In
various types of analyses.
this paper, the term ‘network reduction’ implies only the
It is a common practice for many of the power system
static reduction.
analysis problems to use reduced model of the system in
This paper compares the performance of some of the
place of the complete detailed model. The reasons for using
commonly used network reduction methods, such as Ward
the reduced model can be many, some of which are as
reduction, Kron reduction, Dimo’s method, and Zhukov’s
follows: (i) It may be required to monitor the system by
method.
using only a limited number of measurements from
measuring instruments, such as phasor measurement units II. REDUCTION TECHNIQUES
(PMUs); (ii) an interconnected power system typically
consists of a number of areas owned by various utilities, each A. Ward Reduction
which are usually reluctant to share complete system The construction of this equivalents starts from the
information with others; (iii) practical limitations on the solved model of the entire interconnected system. The
computational resources; and (iv) the fact that, as the injected current i(i) at each bus i is obtained from the bus’s
electrical distance from the point of interest increases, the
requirement for detailed modelling of the remote location known complex power s(i) and voltage v(i ) .
also reduces.
Modern power systems are mostly interconnected, with i (i ) = s* (i ) v* (i ) (1)
multiple areas connected with each other by tie lines. Reduction of the network and the current vector is performed
Usually, a utility’s own system called the internal subsystem by using the Gaussian elimination technique. Let a power
or internal area. The rest of the system is called the external system be described by the following set of nodal equations,
978-1-4799-5141-3/14/$31.00 ©2014 IEEE
Ybus v = i, (2) voltages need not be calculated explicitly and thus can be
eliminated. These equivalents can be constructed using
where Ybus is the n × n bus admittance matrix, v is the n × 1 forecast or assumed bus power and voltage magnitudes [5-
vector of complex voltages at all nodes, and i is the 6].
n × 1 vector of complex currents injected at all nodes.
C. Dimo’s Reduction
After elimination of the kth node, Ybus is modified as,
The idea behind Dimo equivalent is to aggregate the
Yik Ykj power and current injection of a group of buses to be
Y 'ij = Yij − ; ∀i, j = 1,..., n; i, j ≠ k (3) eliminated to a fictitious radial equivalent independent
Ykk (REI) node. First the buses to be eliminated are aggregated
The current vector i is also modified as, into fictitious equivalent nodes through zero power balance
network (ZPBN), as shown in Fig. 2 [7-8]. Then Gaussian
Yik elimination is performed to obtain the equivalent. The steps
i '(i ) = i (i ) − i (k); ∀i = 1,..., n; i ≠ k (4) in the process are:
Ykk
1. Define the set of nodes to be retained and to be
The element Y 'ij distinguishes the elements of the new aggregated.
(n − 1) × (n − 1) Ybus from the original n × n Ybus , and i ' is 2. Analyze the nodal injections at the non-essential
buses and define the aggregating pattern for
the modified current injection vector. If the network is
reduced to r nodes, a new bus admittance matrix of generations and load.
dimension r × r , and a new current injection vector of 3. Assemble ZPBN.
dimension r ×1 are obtained. After reduction, the modified 4. Assemble bus admittance matrix for extended
current vector is converted back to complex power, for use in model.
load-flow solutions [3-5]. This reduced network carries full 5. Gaussian elimination of non-essential buses.
information of the original power system at the base case.
Fig. 1(a) is used to represent the original network and Fig.
1(b) is representing the reduced network. i

{R}
{R} {E}

Original Reduced Load Currents


I_R Network G(i)+jB(i)
I_R Network

Generators and
Synchronous
(a) (b) Condensers Fictitious Ground
Other Injection
Fig. 1 Network Elimination: (a) network before elimination; (b) I_gf
network after elimination. {E} is set of eliminated nodes; {R} (DC ties, AC ties, etc)
is set of retained node Fictitious Load
V_f Center
S_gf
B. Kron Reduction
Fig. 2 Dimo’s Reduction: Transition from the meshed power
Gaussian elimination avoids the need of matrix inversion system network to the radial scheme of short-circuit
while solving the nodal equation of large power systems. It admittances, also known as the REI net
also leads to reduced order network equivalents. This is
used to analyze power system with special focus on voltages The ZPBN is used for aggregating the system generator
at some selected buses. For this purpose selective and loads into a fictitious single-load center while
numbering of system buses is required so as to eliminate preserving the properties and the power balance of the base
buses which are not of importance. The current injections at case. The procedure followed to get the ZPBN is: All the
the retained nodes are taken care of by using (4). non-essential nodes are replaced by linearizing impedances,
Current injection is always zero at the buses which have and are then connected to the fictitious ground. The
no external loads or generators connected with them. Non- impedances that are used for linearization are computed as
B ( i ) = −Q ( i ) V ( i )
2
essential buses having current injection are converted to (6)
shunt admittance ykk prior to reduction,
G (i ) = P ( i ) V ( i )
2
Y kk = − ( PLk − jQLk ) / Vk2 (5) (7)

where PLk and QLk are injected active and reactive loads at where G ( i ) and B ( i ) are conductance and susceptance
the kth bus. The equivalent then becomes a passive between ith non-essential node and fictitious ground,
network with all zero current injections. At such buses, respectively. P ( i ) , Q ( i ) and V ( i ) are active power,
reactive power and complex voltage at the ith non-essential S a = ∑ i∈{ A} s ( i ) (13)
bus, respectively. The current, i Ag (i ) from these non-
essential nodes ( A) to the fictitious ground (g) is The transformation of network is described as
computed as ⎡i R ⎤ ⎡ YRR YRA ⎤ ⎡ v R ⎤
⎢i ⎥ = ⎢ Y ⎥⎢ ⎥ (14)
⎣ A ⎦ ⎣ AR YAA ⎦ ⎣ v A ⎦
i Ag (i) = ( s A (i) v A (i) ) ; ∀i ∈ { A} ,
*
(8)
Equation (12) can be rewritten as,
and the current between fictitious ground and the fictitious
⎡ i R ⎤ ⎡ YRR y Ra ⎤ ⎡ v R ⎤
load center ( f ) is the sum of current flowing from non- ⎢ ⎥=⎢ ⎥⎢ ⎥, (15)
essential buses into the fictitious ground, ⎢⎣ I a ⎥⎦ ⎢⎣ y aR Y aa ⎥⎦ ⎢⎣V a ⎥⎦
I gf = ∑ i∈{ A} i Ag (i ) (9) where I a and V a represent the injected current and voltage
of aggregated node. For current and voltages at the retained
Further application of power conservation theorem at nodes to remain unchanged,
fictitious ground gives
YRA v A = y Ra V a (16)
S gf = ∑ i∈{ A} s Ag (i ), (10)
or,
The knowledge of complex power and current that circulate y Ra = YRA γ (17)
between the fictitious load center and fictitious ground is
used to calculate the line parameters ( R and X ) , between where,
−1
the nodes γ = V a vA (18)
2
R + jX = S gf I gf (11) Equation (11) can be represented as,
*
And the complex voltage at the fictitious load bus is V a I a = vTA i*a (19)
calculated as
* Putting values of (12) and (13) in (17) gives,
V f = S gf I gf (12) * *
V a y*aR v*R + V a Y aa V a = vTA YAR
*
v*R + vTA YAA
*
v*A (20)
Replacing the loads in the original system with the ZPBN
This condition holds true only if,
produces an extended network that has exactly the same
power balance as the original network. y aR = γ *T YAR (21)
The REI net is obtained by applying Gaussian
elimination of all the passive nodes in the extended network. Y aa = γ *T YAR γ (22)

D. Zhukov’s Reduction Eq. (21) and (22) relates the parameters of reduced network
with the parameters of original network by transformation
Zhukov’s method consists of aggregating the selected
nodes through ideal transformer. In this method, a set of ratio ( γ ). This implies that the reduced network obtained for
nodes { A} is aggregated by a single equivalent node ‘a’, as a given initial conditions is also valid for other conditions if
the transformation ratios are the same for all these states.
shown in Fig. 3 [9-10]. The aggregation is done such that the
current ( i R ) and the voltages ( v R ) at the retained nodes III. CASE STUDY
{ R} remained unchanged and the complex power injection
The equivalents described in the last section were used on
at the equivalent nodes is same as the sum of injections at the IEEE 14-bus test system and IEEE 118-bus test system. All
aggregated nodes the tests were based on the initial load-flow solution of the
complete system. All the external buses were eliminated in
Ward and Kron reduction techniques. In Dimo and
{R}
Zhukov’s techniques, the aggregated bus is retained along
{R} {A} a
with the internal buses. Power flow analysis is carried out
Reduced
Original
i_R Network I_ a on full and reduced system with a set percentage change in
i_R Network i_A
V_ a
load and corresponding generation rescheduling.

(a) (b)
A. Results for the IEEE 14 Bus System
Fig. 3 Node aggregation using Zhukov’s method: (a) network
before aggregation; (b) network after aggregation Fig. 4 shows the demarcation of internal and external
areas assumed for the IEEE 14-bus test system for
demonstration purpose. The area consisting of the buses 1,
2, 3, 4 and 5 is treated as the internal area. The remaining
buses constitute the external area as shown in Table I. To
test the effectiveness of these network reduction
methodologies, the load is varied by ±10%, ±20%, and
±30%, and the generations are rescheduled accordingly. In
Fig. 5 voltage magnitudes are plotted and compared for
variation in load of entire power system to variation in only
the internal system. Full Newton-Raphson load flow
technique is used in each case. Normal load flow 5. (a) Base Case
convergence was obtained for each equivalent case. It is
observed that with the increase in load, the per unit (p.u.)
error increases at the boundary buses. When the load is
decreased, the p.u. error in voltage at the boundary buses
does not follow any particular pattern, with the highest error
being at bus number 4 at 30% increase in load. For load less
than base case load, the p.u. error is generally less. The
maximum error observed is 0.0361 p.u. in case of Kron 5. (b) Increment of 10% in load
reduction. The maximum error for Dimo’s and Zhukov’s
reduction technique is 0.0305 p.u. and 0.0310 p.u.
respectively. The least error is obtained using Ward
reduction, i.e. 0.0106 p.u.

TABLE I.

PARTITIONING OF THE IEEE 14-BUS SYSTEM 5. (c) Increment of 20% in load


Internal System Buses External System Buses
1,2,3,4,5 6, 7, 8, 9, 10, 11, 12, 13, 14

B. Results for the IEEE 118 Bus System


The proposed methodologies are also applied to the
IEEE 118 bus system. The division of the buses in internal
and external areas is shown in Table II. The buses in the 5. (d) Increment of 30% in load
external area are eliminated by using the network reduction
methodologies described earlier.

Bus 13 Bus 14
Bus 12

EXTERNAL
AREA
5. (e) Decrement of -10% in load
Bus 10

Bus 6 Bus 9
Bus 11

Bus 7 Bus 8

Bus 5 5. (f) Decrement of -20% in load


Bus 4 INTERNAL
AREA

Bus 3
Bus 1

Bus 2
5. (g) Decrement of -30% in load

Fig. 4 Partitioned IEEE 14-bus system


Legend Description
Dimo’s reduction
Original system

Ward reduction
6. (b) Increment of 5% in load
Kron reduction

Zhukov’s reduction

Fig. 5 Variation of internal bus voltage magnitude with


change in load for IEEE 14 bus system.
TABLE II.
6. (c) Increment of 10% in load
PARTITIONING OF THE IEEE 14-BUS SYSTEM

Internal System Buses External System Buses


24,45-77,116,118 1-23,25-44,78-115,117

Load in the full and reduced system are varied in steps


of 5% and the voltage magnitudes at the retained buses are 6. (d) Increment of 15% in load
observed. Fig. 6 shows the voltage magnitude plot of those
five buses which have maximum deviation from the full
power system voltage magnitudes.
It is observed that the same trend is followed in the IEEE
118 bus system case also, i.e. with increasse in load, the p.u.
voltage magnitude error also increases. The error at
boundary buses are more as compared to the error at the 6. (e) Decrement of 5% in load
internal non-boundary buses. The highest error for Ward
method is 0.0239 p.u. at bus number 76, while for Kron
method maximum error is 0.0028 p.u. at bus number 45
with 115% load, while that for Dimos’s method is 0.0316
p.u. at bus 76 with 115% load. Zhukov’s method does not
show any convergence for the base case. This may be
6. (f) Decrement of 10% in load
because of the higher resistance to reactance ratio in this
method. Maximum and mean errors in p.u. for the different
scenarios studied are represented in Table III

TABLE III.

PARTITIONING OF THE IEEE 14-BUS SYSTEM 6. (g) Decrement of 15% in load

IEEE 14-Bus IEEE 118-Bus Legend Description


Reduction Dimo’s reduction
Mean Mean
Techniques Maximum Maximum
Error (p.u.)
Error
Error (p.u.)
Error Original system
(p.u.) (p.u.)
Ward 0.0106 0.0023 0.0239 0.0014 Ward reduction
Kron 0.0361 0.0081 0.0028 0.0007
Dimo 0.0305 0.0065 0.0316 0.0017 Kron reduction
Zhukov 0.0310 0.0065 - -

Fig. 6 Five internal buses with large variation in voltage magnitude


for IEEE 118 bus system

IV. CONCLUSION
Network reduction methods are found to be very useful
in analysis of very large and complex power systems. The
changes in the internal system parameters can be used to
(a) Base Case approximate the changes in the external system and hence,
can be used for stability and security analysis of very large [4] Ward, J. B., "Equivalent Circuits for Power-Flow Studies",
Transactions of the American Institute of Electrical Engineers,
power systems. IEEE 14 and IEEE 118 bus test cases are vol.68, no.1, pp.373,382, July 1949.
analyzed and it is observed that Ward and Kron methods are [5] Arthur R. Bergen, Power Systems Analysis. Pearson Education India,
applicable at both the systems, but the accuracy of Kron 2009.
method is lower as compared to other methods. Wards [6] S. Deckmann, A. Pizzolante, A. Monticelli, B. Stott, O. Alsac,
reduction technique gives the best result for both the cases. "Numerical Testing of Power System Load Flow Equivalents," IEEE
These methods are less complex than the methods used for Trans. on Power Apparatus and Systems, vol.PAS-99, no.6,
pp.2292,2300, Nov. 1980.
aggregation of nodes. Zhukov’s method works fine with
[7] S. C. Savulescu, "Equivalents for Security Analysis of Power
smaller IEEE 14 bus test system with greater accuracy, but Systems," IEEE Trans. on Power Apparatus and Systems, vol.PAS-
is not found useful for IEEE 118 bus test system. This may 100, no.5, pp.2672,2682, May 1981.
be due to the large r/x ratio of lines in the equivalent system. [8] S. C. Savulescu. Real-time Stability Assessment in Modern Power
Dimo’s method gives better results than Zhukov’s for both System Control Centers. Vol. 42. John Wiley & Sons, 2009.
the test systems. However it is helpful only with small [9] J. Machowski, A. Cichy, F. Gubina, P. Omahen, "External Subsystem
changes in loads, for large changes in loads, the reduced Equivalent Model for Steady-state and Dynamic Security
Asessment," IEEE Trans. Power Systems, vol.3, no.4, pp.1456,1463,
system does not follow the original system. Nov 1988.
[10] J. P. Yang, G. H. Cheng, Z. Xu, "Dynamic Reduction of Large Power
REFERENCES System in PSS/E," IEEE/PES Transmission and Distribution
Conference and Exhibition: Asia and Pacific, 2005 , vol., no., pp.1,4,
[1] Jan Machowski, Janusz Bialek, and Jim Bumby. Power System 2005.
Dynamics: Stability And Control. John Wiley & Sons, 2011. [11] M. Vardikar, S. Chakrabarti, E. Kyriakides, "Transformation of
[2] J. F. Dopazo, G. Irisarri and A. M. Sasson, “Real-time External measurements for using external network equivalents in state
System Equivalent for On-line Contigency Analysis”, IEEE Trans. on estimation," IEEE Power and Energy Society General Meeting (PES),
Power Apparatus and Systems, vol. PAS-98, no. 6, Nov./Dec. 1979. vol, pp.1-5, 21-25 July 2013.
[3] John J. Grainger, and William D. Stevenson. Power System Analysis. [12] K. Dongare, S. Chakrabarti, and E. Kyriakides. "Power System State
New York: McGraw-Hill, 1994. Estimation Considering Real-time Equivalents of the External
Networks." IEEE Innovative Smart Grid Technologies-Asia (ISGT
Asia), pp. 1-6, 10-13 Nov.2013.

You might also like