Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Comparative Study of Model Order

Reduction Techniques
Manoj Lavanshi Dr. D. K. Sambariya
Department of Electrical Engineering Department of Electrical Engineering
Rajasthan Technical University Rajasthan Technical University
Kota, Rajasthan (324010) Kota, Rajasthan (324010)

Abstract:- The higher order system consumes more space  The Problem Associated with Model Order Reduction as
because to higher order matrix. The higher order follow:
interconnected power systems stability analysis is time The Error of Approximation should be Minimum.
consuming and large system performance of the system
cannot be understanding easily. The analysis of the higher
 The original system's characteristics must be kept
system is drudging and inordinate. The analysis and
protected.
control of such system presents a great challenge for
 The computation of the reduction of system should be
system engineer. The lower order system drives a comfort
well-organized or sequential.
exploratory and optimization which results in favorable
similarity to the system. This paper's objectives are to  The reducing procedure need to be automated[5-7, 9-
examine the reduced order model of large scale LTI 15].
system using balanced truncation method. In this
technique the reduction of nominator and denominator For linear time invariant (LTI) dynamic system various
polynomial using balanced truncation (BT) method, reduction methods are proposed for model order reduction of
which results more accurate result. This approach higher order system. The technique which are more
preserves the original system's stability and steady state frequently used for matching the time moments of original
value in the lower order model. and reduced system is pade approximation. But this technique
has a drawback that it has potential to give unstable lower
Keywords:- Reduction Techniques, Stability Equation Method order reduced order model for stable higher order system.
(SEM), Differentiation Equation Method (DEM), Balanced Routh stability reduction technique is commonly used for
Truncation Method, Reduced Model. the reduction of the complexity and matching the transient
response of higher order system with lower order system.
I. INTRODUCTION Unfortunately, it also has limitation to defend the dominant
poles in the reduced order model for a non-minimum phase
system it lags.
Modelling is the process of brief explanation of a
system using mathematical equations and matrix. The
In this paper, for linear time-invariant system, a mixed
mathematical models have a capability of better
method approach is proposed for model order reduction. In
understanding of a system and how it can be control[1-4].
this paper the Stability Equation Method, Differential
The mathematical modelling is used in different fields like,
Equation method and Balanced Truncation Method is used to
power system, control theory and sociology and physiology
reduce a higher order LTI system to a reduced order system.
etc. Study of a mathematical model which has many states
Now to obtain desired reduced order model again reduce the
variables behaviour [5]. The construction of a system or
intermediate lower system.
model which is nearly similar to real phenomenon. A large-
scale order system is therefore challenging to analyse and
II. STATEMENT OF PROBLEM
simulate in terms of synthesis and control. This also affects
the computational time due to large number of system
 Considering a higher order (𝑛𝑡ℎ ) system and a reduced
variables [6].
order (𝑟 𝑡ℎ ) system may be represented as following:
The method of getting an approximation reduced order
∑𝑛−1
𝑖=0 𝑚𝑖 𝑠
𝑖
of a big system with the same properties as the real original G(s)= (1)
∑𝑟−1
𝑖=0 𝑙𝑠
𝑖
system is called "model order reduction". The development of
optimal model order reduction and its generalisation to ∑𝑟−1 𝑞𝑖 𝑠 𝑖
discrete time non-linear systems [7]. There are various model R(s)= ∑𝑟𝑖=0 𝑗 (2)
𝑖=0 𝑝𝑗 𝑠
order reduction techniques are available which are published
in earlier years some of them are Hankel Norm
Where, 𝑚𝑖 , 𝑞𝑖 are scalar constants for numerator and
Approximation, Stability Equation Method, Singular Value
𝑝𝑗, 𝑙𝑗 , are scalar constants for denominator of higher order and
Decomposition, Pade Approximation Technique, Routh
Stability Method, Pade Via Lanczos [8]. reduced order system, respectively. The objective is to find a
reduced 𝑟 𝑡ℎ order system model R(s) such that it retains the

IJISRT23MAY1465 www.ijisrt.com 1441

Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
important properties of the original higher system model G(s) The Equation May be written as
for the same type of inputs. The analysis is done by two
𝑘
different reduction method viz stability equation method and 1
𝑁𝑒 (𝑠) = 𝑏0 ∑𝑖=1 (1 + 𝑠 2 /𝑧𝑖2 ) (12)
differentiation method. The graphical and analytical analysis
of reduced order system is to be carried out. 𝑘
2
𝑁𝑜 (𝑠) =𝑏1 𝑠 ∑𝑖=1(1 + 𝑠 2 /𝑝𝑖2 ) (13)
III. METHODOLOGY
Where 𝑘1 and 𝑘2 are integer parts of n/2 and (n-2)/2
respectively.
For model order reduction various techniques are
proposed, but here we are using these two methodologies Here 𝑧12 < 𝑝12 < 𝑧22 < 𝑝22 ……… by ignoring the factor
which are: with higher magnitude of 𝑧𝑖 & 𝑝𝑖 , we get thee desired reduced
order ‘r’ of stability equation and it can be written as
A. Stability Equation Method
In this technique the reduced order transfer function is 𝑟
1
𝑁𝑒𝑟 (𝑠) = 𝑏0 ∑𝑖=1 (1 + 𝑠 2 /𝑧𝑖2 ) (14)
gained straight from the pole zero arrangement of stability
equations of the original higher order transfer function. Hence 𝑟
the order of the higher order transfer function of stability
2
𝑁𝑜 (𝑠) = 𝑏1 ∑𝑖=1 (1 + 𝑠 2 /𝑝𝑖2 ) (15)
equation can be reduced [14].
Where r1 andr2 are the integer parts of r/2 and (r-2)/2
Let us Assume that the Transfer Function of the Higher respectively. The reduced order numerator formed as:
Order System is:
𝑁𝑟 (𝑠) = 𝑁𝑒𝑟 (𝑠) + 𝑁𝑜𝑟 (𝑠) (16)
𝑏𝑚 𝑠 𝑚 +𝑏𝑚−1 𝑠 𝑚−1 + ………. 𝑏1𝑠+𝑏0
𝐺(𝑠) = (3)
𝑎𝑛 𝑠 𝑛 +𝑎𝑛−1 𝑠 𝑛−1 + …..….𝑎1 𝑠++𝑎0 Similarly, the Reduced Order Denominator is Obtained:

𝑁(𝑠) and 𝐷(𝑠) are the Numerator and Denominator of 𝐷𝑟 (𝑠) = 𝐷𝑒𝑟 (𝑠) + 𝐷𝑜𝑟 (𝑠) (17)
considered transfer function 𝐺(𝑠) respectively. By separating
the numerator 𝑁(𝑠) and denominator 𝐷(𝑠) into their even The Reduced Order Transfer Function R(S) is Written
and odd parts we get as:
𝑁𝑒 (𝑠)+𝑁𝑜 (𝑠) 𝑁𝑒𝑟 (𝑠)+𝑁𝑜𝑟(𝑠)
𝐺(𝑠) = (4) 𝑅(𝑠) = (18)
𝐷𝑒 (𝑠)+𝐷𝑜 (𝑠) 𝐷𝑒𝑟 (𝑠)+𝐷𝑜𝑟 (𝑠)

Where, As the result of the stability equation method, we get

stable reduced order models. It is observed that poles and
𝑁𝑒 (𝑠) = ∑𝑚
𝑖=2,2,4 𝑏𝑖 𝑠
𝑖
(5) zeros of lower models have a greater degree of dominance
than those of higher magnitude. In this method, the poles or
𝑁𝑜 (𝑠) = ∑𝑚
𝑖=1,3,5 𝑏𝑖 𝑠
𝑖
(6) zeros with higher magnitudes ore disregarded.

𝐷𝑒 (𝑠) = ∑𝑛𝑖=0,2,4, 𝑎𝑖 𝑠 𝑖 (7) B. Differentiation Equation Method-

This method was introduced by Gutman et al. [14]. This
𝐷𝑜 (𝑠) = ∑𝑛𝑖=1,3,5 𝑎𝑖 𝑠 𝑖 (8) method based on polynomial’s differentiation. The
differentiation approach of Gutman et al. may also be
considered a multipoint pade method; it is another stability
The roots of the even part of numerator 𝑁𝑒 (𝑠) and
preserving method. Any reduction technique in the frequency
denominator 𝐷𝑒 (𝑠) are called zeros 𝑧𝑖 (𝑠) while the odd part
domain that obtains the lower order models, as a result of this
of the numerator 𝑁𝑜 (𝑠) and denominator 𝐷𝑜 (𝑠) are called
novel connection between the pade and stability preserving
poles 𝑝𝑖 (𝑠). In this method, we reduce a polynomial’s less
methods.
significant factor by individual eliminating. Let us illustrate
the method by reducing the numerator and the denominator is
The coefficient of the reduced order transfer function is
reduced as well. The reduced order model is attained by the
produced by differentiating the reciprocal of the numerator
ratio of reduced numerator and denominator.
and denominator polynomial of the higher transfer functions
repeatedly. The differentiated reduced order is reciprocated
The Numerator is Separated as:
again and normalized reduced polynomials are used. Because
of the problem that zeros with big modules tend to be better
N(s) = 𝑁𝑒 (𝑠) + 𝑁𝑜 (𝑠) (9) approximation than those with a small module, the
straightforward differentiation is abandoned.
Were

𝑁𝑒 (𝑠) = 𝑏0 + 𝑏2 𝑠 2 + 𝑏4 𝑠 4 +… (10)

𝑁𝑜 (𝑠) = 𝑏1 + 𝑏3 𝑠 3 + 𝑏5 𝑠 5 + … … … (11)

IJISRT23MAY1465 www.ijisrt.com 1442

Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
 Algorithm of Differentiation- lower order state space matrix is again transformed into the
transfer function equation.
 The Reciprocal of the Transfer Function is Taken.
 To Get the Desired Reduced Order the Reciprocated The input and output state space models are inversely
Transfer Function is Differentiated. correlated with the controllability and observability
 The Reduced Transfer Function is Again Reciprocated Gramians. The controllability of the system is given by: -
 The Decreased Order System is then Subjected to the
∞ 𝑇
Steady State Correction. 𝑃 = ∫0 𝑒 𝐴𝑡 𝐵𝐵𝑇 𝑒 𝐴 𝑡 𝑑𝑡 (25)

C. Balance Truncation Method Similarly, the observability of the system is given by:

 Review on BTM- ∞ 𝑇
𝑄 = ∫0 𝑒 𝐴 𝑡 𝐶 𝑇 𝐶𝑒 𝐴 𝑡 𝑑𝑡
𝑇
(26)
This approach transforms the controllability Gramian
and the observability Gramian into a diagonal matrix for the
The two Lyapunov equations can be used to determine:
altered realisation [11]. Balance realisation is the name given
to this process. Balancing of a particular reality is the first
𝐴𝑃 + 𝑃𝐴𝑇 + 𝐵𝐵𝑇 = 0 (27)
stage in the balance truncation procedure.
𝐴𝑇 𝑄 + 𝑄𝐴 + 𝐶 𝑇 𝐶 = 0 (28)
Let us consider a LTI system in a state space form as: -
 Diminution by Balanced Truncation Technique: -
𝑋(𝑡) = 𝐴𝑥(𝑡) + 𝐵𝑢(𝑡) (19)
There is a transformation procedure that, for every
stable dynamic system, renders controllability and
𝑦(𝑡) = 𝐶𝑥(𝑡) + 𝐷𝑢(𝑡) (20)
observability equal and in diagonal form. The idea of system
observability and controllability provides the basis for the
Where for each ‘t’,
combination of singular value decomposition, principal
component analysis, and the balanced truncation approach.
𝑢(𝑡)€𝑅𝑞 input vector,
 Minimal Phase System: -
𝑥(𝑡)€𝑅𝑛 state vector,
Consider that (A, B) is observable, (A, C) is
controllable, and A, is stable. A matrix can be changed into if
𝑦(𝑡)€𝑅𝑝 output vector respectively.
it is symmetric and positive definite into the lower order
triangular matrix by Cholesky factorization method. The
The transfer function of original state space model is
Cholesky factor of lower triangular matrix are 𝐿𝑐 𝑎𝑛𝑑 𝐿𝑜 , of
obtained as: -
the P and Q.
𝐺(𝑠) = 𝐶(𝑠𝐼 − 𝐴)−1 𝐵 + 𝐷 (21)
𝑃 = 𝐿𝑐 𝐿𝑇𝑐 (29)
Here 𝐴, 𝐵, 𝐶 𝑎𝑛𝑑 𝐷 are the matrix of order 𝑛 ∗ 𝑛, 𝑛 ∗
𝑄 = 𝐿𝑜 𝐿𝑇𝑜 (30)
𝑞 𝑎𝑛𝑑 𝑝 ∗ 𝑞 in n-dimensional space R.
The matrix's singular value decomposition 𝐿𝑜 𝐿𝑇𝑐 ,
The objective is the higher order model to be
transformed into lower order model. The reduced order state
space equation is written as: - 𝐿𝑜 𝐿𝑇𝑐 = 𝑊∑𝑉 𝐻 (31)

𝑥𝑟′ (𝑡) = 𝐴𝑟 𝑥𝑟 (𝑡) + 𝐵𝑟 𝑢𝑟 (𝑡) (22) Here, W, V = orthogonal matrix, 𝑉 𝐻 =Hermitian

transpose
𝑦𝑟 (𝑡) = 𝐶𝑟 𝑥𝑟 (𝑡) + 𝐷𝑟 𝑢𝑟 (𝑡) (23)
The column of matrix u is referred to as a left single
vector and column of matric V is named as a single right
The state space matrix for reduced order into transfer
vector. The strength of controllability and observability for
function is written as: -
each specific state is obtained by Hankel singular values.
𝐺𝑟 (𝑠) = 𝐶𝑟 (𝑠𝐼 − 𝐴)−1 𝐵𝑟 + 𝐷𝑟 (24)
A dynamic system can be changed into a balanced
system by utilising the non-singular transformation T, which
The resulted error of output of both original and
is defined as
reduced model is kept low promising for input u(t).
1
The level of observability and control Gramians can be 𝑇 = 𝐿𝑐 𝑉 ∑−2 𝛼 (32)
used to measure observability and controllability in particular
state space directions. In this the transfer function equation of The balanced system matrix is obtained as.
higher order is transformed into equivalent state space model
then the matrix is reduced into the lower order matrix and the 𝐴~ = 𝑇 −1 𝐴𝑇, 𝐵 ~ = 𝑇 −1 𝐵, 𝐶 ~ = 𝐶𝑇 (33)

IJISRT23MAY1465 www.ijisrt.com 1443

Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
The transmission matrix D remain unchanged during the reduction process of original system. In a balanced system the
𝑃 ~ 𝑎𝑛𝑑 𝑄 ~ matrix became same and converted into diagonal matrix as

𝛼1 0 0 0 0 0 0
0 𝛼2 0 0 0 0 0
0 0 … 0 0 0 0
𝑃~ = 𝑄~ = ∑ 𝛼 = 0 0 0 𝛼𝑟 0 0 0 (34)
0 0 0 0 𝛼𝑟+1 0 0
0 0 0 0 0 … 0
[0 0 0 0 0 0 𝛼𝑛 ]

Where 𝛼𝑖 𝑖 = 1,2, … … , 𝑟, 𝑟 + 1, … … 𝑛 are Hankel singular values, listed in decreasing order as 𝛼1 > 𝛼2 > 𝛼3 > ⋯ > 𝛼𝑟 ≫
𝛼𝑟+1 > ⋯ > 𝛼𝑛 .

To calculate the desired order (r) reduced model the balance system (𝐴~ , 𝐵 ~ , 𝐶 ~ , 𝐷 ~ ) is partitioned as

𝐴11 𝐴12 𝐵
𝐴~ = [ ] , 𝐵 ~ = [ 1 ], 𝐶 ~ = [ 𝐶1 𝐶2 ], 𝐷 ~ = 𝐷 (35)
𝐴21 𝐴22 𝐵2

The order of the reduced order matrix element which are used for computation are 𝐴11 (𝑟 ∗ 𝑟), 𝐵1 (𝑟 ∗ 𝑚) and 𝐶1 (𝑝 ∗ 𝑟).
Therefore, the simplified model's transfer function is: -

𝑅𝑟 (𝑠) = 𝐷 ~ + 𝐶1 (𝑆𝑖 − 𝐴11 )−1 𝐵1 (36)

 Non-Minimal Phase System: -

Assume that A is stable, for obtaining the balance realization of given non-minimal stable system which is continuous-time
system The square root approach was put out by Tombs and Postlethwaite. The observability and controllability Gramians for
non-minimal phase systems are positive semi-definite matrices. By using the Choleskey factorization method, this sort of matrix
cannot be converted into a lower order triangular matrix (L). Use is made to acquire the lower order triangular matrix of the
𝐿𝐷𝐿𝑇 decomposition of non-minimal phase systems. The information concerning 𝐿𝐷𝐿𝑇 breakdown is provided. The lower
triangular matrices 𝐿𝑜 and 𝐿𝑐 are obtained when the controllability and observability Gramians are obtained by equations (28) and
(29)

𝑃 = 𝐿𝑐 𝑑𝐶 𝐿𝑇𝐶 (37)

𝑄 = 𝐿𝑂 𝑑𝑂 𝐿𝑇𝑂 (38)

Instead of being positive definite, the lower triangular matrices are symmetric positive semi-definite. Decomposing the
lower triangular matrix into singular values 𝐿𝑇𝑜 𝐿𝑐 is given as

𝐿𝑇𝑜 𝐿𝐶 = [𝑈1 𝑈2 ][∑1 𝑉1 ∑2 𝑉2 ] (39)

IV. NUMERICAL EXAMPLE

Example1. Let us consider an 8th order transfer function equation:

The main purpose is to reduce the model of the eighth order to second order. Here we are using

18𝑠 7 +514𝑠 6 +5982𝑠 5 +36380𝑠 4 +

𝑇(𝑠) = (40)
𝑠 8 +36𝑠 7 +546𝑠 6 +4536𝑠 5 +22449𝑠 4 +

Various steps for obtaining the reduced order model, which are as follow: -

Step 1: Separating the N(s) and D(s) of the above system into even and odd parts.

Numerator part:

Using equation (10) and (11) the numerator further divided: as

Even part:

IJISRT23MAY1465 www.ijisrt.com 1444

Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
𝑁𝑒 (𝑠) = 514𝑠 6 + 36380𝑠 4 + 222088𝑠 2 + 40320 (41)

This can be written as: -

514𝑠 2
𝑁𝑒𝑟 (𝑠) = 36380𝑠 4 ( + 1) + 222088𝑠 2 + 40320 (42)
36380

Odd part:

𝑁𝑜𝑟 (𝑠) = 18𝑠 7 + 5982𝑠 6 + 122664𝑠 3 + 185760𝑠 (43)

Also, the odd part is: -

18𝑠 2
𝑁𝑜𝑟 (𝑠) = 5982𝑠 5 + ( + 1) + 122664𝑠 3 + 185760𝑠 (44)
5982

As per stability equation concept, the factors which have large magnitudes can be neglected. Therefore, the reduced order
numerator is from eq. (16) is:

𝑁(𝑠) = 𝑁𝑒𝑟 (𝑠) + 𝑁𝑜𝑟 (𝑠) (45)

𝑁𝑟 (𝑠) = 185760𝑠 + 40320 (46)

Denominator part:

Even part:

𝐷𝑒 (𝑠) = 𝑠 8 + 546𝑠 6 + 22449𝑠 4 + 118124𝑠 2 + 40320 (47)

The eq. (46) can be written as:

s2
Der (s)  546s (  1)  22449s4  118124s 2  40320
6
(48)
546
After neglecting the factors which have large magnitudes the equation (46) becomes:

Der (s)  118124 s 2  40320 (49)

Odd part:

Do (s)  36s 7  64536s 5  67284 s 3  109606 s (50)

Equation (49) can be written as:

36s 2
Dor (s)  64536s ( 5
 1)  67284s3  109606s (51)
64536
Reducing further eq. (50) we get:

Dor (s)  109606s (52)

From equation (17) the reduced denominator is

Dr (s)  Der (s)  Dor (s) (53)

Hence the reduced 2nd order model of the system using stability equation method is:
185760𝑠+40320
𝑇2𝑟(𝑠) = (54)
118121𝑠 2 +109606𝑠+40320

IJISRT23MAY1465 www.ijisrt.com 1445

Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165

Fig 1 Comparison of Step Response of Reduced Order Model and Original Higher Order using Stability Equation Method

Example:2 Consider the equation (39) for Model order reduction using differentiation method.

18𝑠 7 +514𝑠 6 +5982𝑠 5 +36380𝑠 4 +

𝑇(𝑠) = (55)
𝑠 8 +36𝑠 7 +546𝑠 6 +4536𝑠 5 +22449𝑠 4 +

The reduction of polynomials Nominator and Denominator are separate and by taking the reciprocal of both.

The reciprocated numerator is:

dN(s)
= 18 + 514s + 5982s 2 + 36380s 3 + 122664s 4 + 222088s 5 + 185760s 6 + 40320s 7 (56)
ds

Differentiating with respect to ‘s’ the numerator is:

DNr (𝑠)
= 514 + 11964𝑠 + 109140𝑠 2 + 490656𝑠 3 + 1110440𝑠 4 + 1114560𝑠 5 + 282240𝑠 6 (57)
𝑑𝑠

Similarly, differentiating the numerator up to desired reduced order, the second order numerator is:

𝑁2𝑟 (𝑠) = 133747200 + 203212800 (58)

The original reduced numerator is obtained by again reciprocal of 𝑁2𝑟 (𝑠)

𝑁2𝑟 = 133747200𝑠 + 203212800 (59)

The reciprocal of denominator is:

𝐷𝑟 (𝑠) = 1 + 36𝑠 + 546𝑠 2 + 4536𝑠 3 + 22449𝑠 4 + 67284𝑠 5 + 118124𝑠 6 + 109606𝑠 7 + 4032057) (60)

Differentiating with respect to ‘s’

𝑑𝐷𝑟 (𝑠)
= 36 + 1092𝑠 + 13608𝑠 2 + 89796𝑠 3 3364420𝑠 4 + 708744𝑠 5 + 767242𝑠 6 + 322560𝑠 7 (61)
𝑑𝑠

To obtain the desired reduced order the denominator is differentiated again and again, the reduced 2 nd order denominator is:

𝑑𝐷𝑟 (𝑠 )
= 85049280 + 552414240𝑠 + 812851200𝑠 2 (62)
𝑑𝑠

Again, the reduced order denominator is reciprocated:

𝐷2𝑟 (𝑠) = 85049280𝑠 2 + 552414240𝑠 + 812851200 (63)

IJISRT23MAY1465 www.ijisrt.com 1446

Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
The reduce second order transfer function of the original higher order system from equation (50) and (54) is:
133747200𝑠+202312800
𝑇2 (𝑠) = (64)
85049280𝑠 2 +552414240𝑠+812851200

Applying steady state correction to reduced order model

40320
𝑆𝑆𝑂 = =1 (65)
40320

203212800
𝑆𝑆𝑅 = = 0.25 (66)
812851200

1
𝐾2 = =4 (67)
0.25

So, the final reduced second order transfer function is shown below:
534988800𝑠+812851200
𝑇2 (𝑠) = (68)
85049280𝑠 2 +552414240𝑠+812851200

Fig 2 Comparison of the step response of the original system with the reduced system using differentiation equation method

Example: -3 considering the equation (39) for balanced truncation reduction technique

(69)

With the use of equation (28) and (29) the Lyapunov controllability Gramians (60) and observability Gramians (69)
equations are shown in table1.

The reduced order state space matrices of the system are

−7.3260 2.1350 −4.2220

𝐴1 =[ ] 𝐵1 =[ ] 𝐶1 =[−4.222 0.2378] 𝐷1 = [0] (70)
−2.1350 −0.0379 −0.2378

The equation(x) is used to calculate the second order reduced model:

𝑇𝑏 (𝑠) = 𝐷1 + 𝐶1 (𝑠𝐼 − 𝐴)−1 𝐵1 (71)

17.77𝑠+4.548
𝑇2 (𝑠) = (72)
𝑠 2 +7.364𝑠+4.836

Applying steady state correction to reduced order model.

IJISRT23MAY1465 www.ijisrt.com 1447

Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
From Eqn.(62) the SSO is:
40320
𝑆𝑆𝑂 = = 1 (73)
40320

4.548
𝑆𝑆𝑅 = = 0.94 (74)
4.836

1
𝐾2 = = 1.06 (75)
0.94

So the final reduced order transfer function system is shown below and fig.(3) shows the step response of reduced and
original system.

{18.8362𝑠+4.820}
𝑇2(𝑠) = {𝑠2 (76)
+7.364𝑠+4.836}

Fig 3 Comparison of the step response of the original system with the reduced system using balanced truncation method

Table 1 The Lyapunov Observability and Controllability of Higher Order System

Sr. No. Controllability and Observability Eq. No.
1 0.0214 0.0000 −0.0001 −0.0000 0.0000 −0.0000 −0.0000 0.0000
0.0000 0.0001 −0.0000 −0.0000 −0.0000 0.0000 −0.0000 −0.0000
−0.0001 −0.0000 0.0000 0.0000 −0.0000 −0.0000 0.0000 0.0000
−0.0000 −0.0000 0.0000 0.0000 −0.0000 −0.0000 0.0000 0.0000 (60)
P=
0.0000 −0.0000 −0.0000 −0.0000 0.0000 −0.0000 −0.0000 −0.0000
−0.0000 0.0000 −0.0000 −0.0000 −0.0000 0.0000 0.0000 −0.0000
−0.0000 −0.0000 0.0000 0.0000 −0.0000 0.0000 0.0000 −0.0000
[ 0.0000 −0.0000 0.0000 0.0000 −0.0000 −0.0000 −0.0000 0.0000 ]

2 0.0000 0.0000 0.0000 0.0000 0.0001 0.0003 0.0002 0.0000

0.0000 0.0000 0.0002 0.0013 0.0042 0.0074 0.0059 0.0009
0.0000 0.0002 0.0024 0.0148 0.0500 0.0907 0.0754 0.0153
0.0000 0.0013 0.0148 0.0913 0.3145 0.5888 0.5250 0.1407 (61)
Q=109 *
0.0001 0.0042 0.0500 0.3145 1.1135 2.1838 2.1344 0.7392
0.0003 0.0074 0.0907 0.5888 2.1838 4.6028 5.0745 2.2276
0.0002 0.0059 0.0754 0.5250 2.1344 5.0745 6.5581 3.5768
[0.0000 0.0009 0.0153 0.1407 0.7392 2.2276 3.5768 2.3825]

IJISRT23MAY1465 www.ijisrt.com 1448

Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Table 2 Time Response Analysis of Step Responses of Reduced2nd Order System
Reduction Reduced 2nd order Rise Settling Peak Peak Overshoot
Techniques Time Time Time
Balance {18.8362𝑠 + 4.820} 0.0528 5.9677 2.3329 0.4442 134.0648
Truncation 𝑇2(𝑠) = 2
{𝑠 + 7.364𝑠 + 4.836}
Stability 185760𝑠 + 40320 0.6863 8.7186 1.5659 2.6800 56.5926
Equation 118121𝑠 2 + 109606𝑠 + 40320
Method
Differential 534988800𝑠 + 812851200 0.2438 1.7198 1.1093 0.6541 10.9293
Equation 85049280𝑠 2 + 552414240𝑠 + 812851200
Method

V. CONCLUSION [7] S. R. Potturu and R. Prasad, "Model Reduction of LTI

Discrete-Time Multivariable Systems using Pade
It can be seen that the balance truncation method's Approximation and Stability Equation Method," in
reduced order equation's response and the result of the 2019 6th International Conference on Control,
original higher order equation are similar. Decision and Information Technologies (CoDIT),
2019, pp. 67-72.
A model reduction method based on Stability equation [8] S. K. Gautam, S. Nema, and R. K. Nema, "Model
method, Differentiation equation method and balance Order Reduction of Interval Systems Using Routh
truncation method (BTM) is presented in this research. A Approximation with Mid-Point Concept and Stability
numerical example that compares the time responses of the Equation Method," in 2021 IEEE 2nd International
original system with the reduced system evaluates that the Conference On Electrical Power and Energy Systems
rise time, settling time, peak and peak time of the balanced (ICEPES), 2021, pp. 1-5.
truncation technique is low as compared to the SEM and [9] K. Sudheswari, R. V. Ganesh, C. Manoj, A. Ramesh,
DEM. and K. Manoz Kumar Reddy, "Model-Order Reduction
and Reduced Controller Design Using Routh
Where, the overshoot is higher in BTM technique Approximation and Factor Division Method," in Soft
compared to SEM and DEM. Computing Techniques and Applications, Singapore,
2021, pp. 165-170: Springer Singapore.
REFERENCES [10] A. Gupta and A. K. Manocha, "Comparative Analysis
of Different Balanced Truncation Techniques of Model
[1] L. Li, "Coprime factor model reduction for discrete- Order Reduction," in International Conference on
time uncertain systems," Systems & Control Letters, Innovative Computing and Communications,
vol. 74, pp. 108-114, 2014. Singapore, 2021, pp. 453-463: Springer Singapore.
[2] A. Vendl and H. Faßbender, " Model Order Reduction [11] H. Zeipel, T. Frank, M. Wielitzka, and T. Ortmaier,
for Unsteady Aerodynamic Applications," Proc. Appl. "Comparative Study of Model Order Reduction for
Math. Mech., vol. 13, no. 1, pp. 475-476, December Linear Parameter-Variant Thermal Systems," in 2020
2013 2013. IEEE International Conference on Mechatronics and
[3] C. Seshukumar, B. Dasu, and M. S. Kumar, "A Hybrid Automation (ICMA), 2020, pp. 990-995.
Reduction Technique for Transformer Linear Section [12] A. Sikander and R. Prasad, "A New Technique For
Model," in International Journal of Engineering Reduced-Order Modelling of Linear Time-Invariant
Research and Technology, 2013, vol. 2: IJERT. System," IETE Journal of Research, pp. 1-9, 2017.
[4] S. Satakshi, S. Mukherjee, and R. C. Mittal, "Order [13] K. Salah, "A novel model order reduction technique
reduction of linear time invariant continuous systems based on artificial intelligence," Microelectronics
using mixed approach," in Power, Control and Journal, vol. 65, pp. 58-71, 2017/07/01/ 2017.
Embedded Systems (ICPCES), 2012 2nd International [14] P. Kumar and S. K. Chaudhary, "Novel approach in
Conference on, 2012, pp. 1-5. classical pade approximation," in 2017 7th
[5] A. Sikander and R. Prasad, "Reduced order modelling International Conference on Cloud Computing, Data
based control of two wheeled mobile robot," Journal of Science & Engineering - Confluence, 2017, pp. 667-
Intelligent Manufacturing, pp. 1-11, 2017// 2017. 670.
[6] C. N. Singh, D. Kumar, and P. Samuel, "Order [15] U. Zulfiqar, M. Imran, and A. Ghafoor, "Cross-
reduction of interval systems using direct truncation Gramian based frequency-weighted model order
and stability equation method," in 2017 International reduction technique," Electronics Letters, vol. 52, no.
Conference on Advances in Mechanical, Industrial, 16, pp. 1376-1377, 2016.
Automation and Management Systems (AMIAMS), [16] S. K. Tiwari and G. Kaur, "Model Reduction by New
2017, pp. 363-368. Clustering Method and Frequency Response
Matching," Journal of Control, Automation and
Electrical Systems, journal article pp. 1-8, 2016.

IJISRT23MAY1465 www.ijisrt.com 1449

Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
[17] S. K. Tiwari and G. Kaur, "An improved method using
factor division algorithm for reducing the order of
linear dynamical system," Sādhanā, journal article vol.
41, no. 6, pp. 589-595, 2016.
[18] D. Srivastava and D. Chandra, "Model Order
Reduction of Finite Element model," in 2016 2nd
International Conference on Advances in Electrical,
Electronics, Information, Communication and Bio-
Informatics (AEEICB), 2016, pp. 27-31.
[19] J. Singh, C. B. Vishwakarma, and K. Chattterjee,
"Biased reduction method by combining improved
modified pole clustering and improved Pade
approximations," Applied Mathematical Modelling,
vol. 40, no. 2, pp. 1418-1426, 2016.
[20] A. Sharma, H. Sharma, A. Bhargava, and N. Sharma,
"Power law-based local search in spider monkey
optimisation for lower order system modelling,"
International Journal of Systems Science, vol. 48, no.
1, pp. 150-160, 2017/01/02 2016.
[21] K. Salah and A. Adel, "Model order reduction using
fuzzy logic algorithm," in 2016 28th International
Conference on Microelectronics (ICM), 2016, pp. 13-
16.
[22] N. Razmjooy and M. Ramezani, "Model Order
Reduction based on meta-heuristic optimization
methods," in 1st International Conference on New
Research Achievements in Electrical and Computer
Engineering, 2016, pp. 1-6: IEEE.
[23] A. Rawat and S. K. Jha, "Analysis of model order
reduction based on Mikhailov criterion," in 2016
International Conference on Electrical, Electronics,
and Optimization Techniques (ICEEOT), 2016, pp.
3601-3605.
[24] D. Oyaro and P. Triverio, "TurboMOR-RC: An
Efficient Model Order Reduction Technique for RC
Networks With Many Ports," IEEE Transactions on
Computer-Aided Design of Integrated Circuits and
Systems, vol. 35, no. 10, pp. 1695-1706, 2016.

IJISRT23MAY1465 www.ijisrt.com 1450