VIETNAM MINISTRY OF EDUCATION AND TRAINING

HO CHI MINH CITY

UNIVERSITY OF TECHNOLOGY AND EDUCATION

LE TRONG NGHIA

RESEARCH ON METHODS TO IMPROVE LOAD SHEDDING

IN POWER SYSTEM

MAJOR: ELECTRICAL ENGINEERING

PROGRAM CODE: 9520201

DOCTORAL THESIS SUMMARY

HO CHI MINH CITY –2020

 A THESIS SUBMITTED IN FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY AT
HO CHI MINH CITY
UNIVERSITY OF TECHNOLOGY AND EDUCATION

The principal supervisor: Assoc. Prof. Dr. QUYEN HUY ANH

The second supervisor: Assoc. Prof. Dr. PHAN THI THANH BINH

Reviewer 1:
Reviewer 2:
Reviewer 3:

This thesis has been presented to

THE EXAMINATION COUNCIL FOR DOCTORAL THESIS
HO CHI MINH CITY
UNIVERSITY OF TECHNOLOGY AND EDUCATION

Ho Chi Minh City - 2020

 PUBLICATIONS

1. Trong Nghia Le, Huy Anh Quyen, Ngoc Au Nguyen. Application of fuzzy-
analytic hierarchy process algorithm and fuzzy load profile for load shedding
in power systems. International Journal of Electrical Power & Energy
Systems, Volume 77, pp. 178-184, May 2016. ISSN: 0142-0615. (SCIE).
2. Trong Nghia Le, Ngoc Au Nguyen, and Huy Anh Quyen. Emergency
Control of Load Shedding Based on Coordination of Artificial Neural Network
and Analytic Hierarchy Process Algorithm. Proc of IEEE Conferences on
System Science and Engineering (ICSSE), pp. 57-60, September – 2017; ISSN:
2325-0925 (Online) (IEEE Xplore).
3. Nguyen Ngoc Au, Le Trong Nghia, Quyen Huy Anh, Phan Thi Thanh Binh.
Sa thải phụ tải dựa trên nhận dạng nhanh ổn định động hệ thống điện. Tạp chí
khoa học công nghệ, Đại học Đà Nẵng, ISSN 1859-1531 – Số 11(129).2017
Quyển 2, 11-2017.
4. L.T. Nghia, T.T. Giang, N.N. Au, Q.H. Anh, Do. Ngoc. An. Emergency
Control of Load Shedding Based on Fuzzy- AHP Algorithm. International
Journal of Engineering Research & Technology, Vol. 6 Issue 09, pp. 185-191,
September – 2017; ISSN: 0974 –3154 (Print).
5. Nghia. T. Le, Anh. Huy. Quyen, Au. N. Nguyen, Binh. T. T. Phan, An. T.
Nguyen, Tan. T. Phung. Application of Dual Artificial Neural Networks for
Emergency Load Shedding Control. International Journal of Advanced
Computer Science and Applications (IJACSA), Vol. 11, No. 4, pp.74-82, 2020.
ISSN: 2158-107X (Print) ISSN: 2156-5570 (Online). (ESCI).
6. Nghia. T. Le, Anh. Huy. Quyen, Binh. T. T. Phan, An. T. Nguyen, and Hau.
H. Pham. Minimizing Load Shedding in Electricity Networks using the
Primary, Secondary Control and the Phase Electrical Distance between
Generator and Loads. International Journal of Advanced Computer Science
and Applications (IJACSA), Vol. 10, No. 2, pp.293-300, 2019. ISSN: 2158-
107X (Print) ISSN: 2156-5570 (Online). (ESCI).
7. Le Trong Nghia, Quyen Huy Anh, Phan Thi Thanh Binh, N Thai An, P H
Hau. A voltage electrical distance application for power system load shedding
considering the primary and secondary generator controls. International
Journal of Electrical and Computer Engineering (IJECE), Vol. 9, No. 5, pp.
3000-3009, October 2019. ISSN: 2088-8708. (SCOPUS Indexed Q2).
8. Trong Nghia Le, Huy Anh Quyen, Thi Thanh Binh Phan, Ngoc Au Nguyen,
Tan Phung Trieu. Select Location for Load Shedding In Power System. Proc
of IEEE Conferences on Green Technology and Sustainable Development
(GTSD), pp. 13-17, December 2018. (IEEE Xplore)
9. Le. Trong. Nghia, Quyen. Huy. Anh, P.T.T. Binh, Phung. Trieu. Tan. A
New Under-Frequency Load Shedding Method Using the Voltage Electrical
Distance and Artificial Neural Networks. International Journal of Advanced
Engineering, Management and Science (IJAEMS), Vol.5, No 3, pp. 171-179,
March 2019. ISSN: 2454-1311.
10. Le Trong Nghia, Huy Anh Quyen, Huu Kiet. Load Shedding based on
Fuzzy Logic and AHP Algorithm. International Journal of Engineering
Research & Technology, Vol. 5 Issue 04, pp. 395-402, April-2016; ISSN: 0974
–3154 (Print).
11. N.M. Tam, L.T. Nghia, N.H. Nhat, Q.H. Anh. The Optimal Location Load
Shedding using Electrical Distances. International Journal of Engineering
Research & Technology, Vol. 6 Issue 10, pp. 99-103, October – 2017; ISSN:
0974 –3154 (Print).
12. L.T. Nghia, T.T. Giang, Q.H. Anh, P.T.T. Binh, Bui.NguyenXuan. Vu.
Shedding Apply Neural Network and Power Sensitivity Theory. International
Journal of Engineering Research & Technology, Vol. 7 Issue 04, pp. 408-413,
April-2018; ISSN: 0974 –3154 (Print).
13. Le Trong Nghia, Quyen Huy Anh, Phan Thi Thanh Binh, Nguyen Trong
Tin. Minimizing The Amount of Load Shedding Considering The Primary
Control Of Generator. Journal of Technical Education Science, Vol. 49, pp.
58-66, June 2018; ISSN: 1859 –1272.
14. Tan. T. Phung, Nghia. T. Le*, Anh. Huy. Quyen, Hau. H. Pham, and An.
T. Nguyen. A Hybrid Artificial Neural Network - Genetic Algorithm for Load
Shedding in Power System. GMSARN International Journal, Vol 14, pp. 21-
28, 01-2020. ISSN:1905-9094. (SCOPUS Indexed).
INTRODUCTION

1. The urgency of the research

Frequency is an important specification in assessing the power quality of
the electricity system and must be maintained within permissible limits to ensure
the stable operation of the power system.

When serious problems occur, such as short circuits on generator buses,

transmission lines or transformers, it can cause power system frequency
instability. These incidents need to be detected quickly in order to make a
decision on with/ without load shedding for stable restoration of power system.

Previous studies, the calculation and simulation only considered one

operating load mode, but these studies did not consider many different load
modes.

In addition, the previous load shedding optimization studies only

considered single-target constraints, which are mainly technical constraints.
These studies have not considered the combination of economic and technical
criteria in a load shedding scheme.

Therefore, the thesis: "Research on methods to improve load shedding

in power system" partially overcomes the above limitations and meets the urgent
requirements in control, operation and protection of power system.
2. The objectives of the thesis
3. Object and scope of the study
4. Approach and research method
5. New contributions in terms of science and practical significance of the
thesis
6. The structure of the thesis

1
Chapter 1 OVERVIEW THE LOAD SHEDDING IN POWER
SYSTEM

1.1 Overview the power system problems

1.2 Overview the frequency control and load shedding
1.3 Load shedding selection factors
1.4 Literature Review
Studies of load shedding control focus on solving the problems: restore
power system stability with the fastest time, the smallest load shedding capacity,
minimizing costs of damage during power outages. There are many methods for
load shedding and restoring power system stability that have been developed by
researchers. The load shedding techniques are divided into three main research
areas: Conventional Load Shedding techniques, Adaptive Load Shedding
techniques, and Intelligent Load Shedding (ILS) [2].
1.4.1 Traditional load shedding
a. Under Frequency Load Shedding (UFLS) [9], [10] [11]
b. Under Voltage Load Shedding (UVLS) [16], [17] [18]
1.4.2 Adaptive Load Shedding [19], [21]
1.4.3 Intelligent Load Shedding
Intelligent Load Shedding (ILS) is a method of activating the under
frequency relays based on a dynamic variable intelligent load shedding program.
The main components of this program: knowledge bases, fault lists and
computational methods ILS [22] - [29].
a. Application of Artificial Neural Network (ANN) in load shedding [25],
[30], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42].
b. Application of fuzzy logic control in load shedding [43], [44], [45], [46],
[47]
c. Application of Adaptive Neural-Fuzzy Inference System (ANFIS) in load
shedding [48], [43], [49]
d. Application of Genetic Algorithms (GA) in load shedding [51], [52] [53],
[54], [55], [56], [57], [58], [59].
e. Application of PSO algorithm in load shedding [61], [62]

2
Chapter 2 EMERGENCY CONTROL OF LOAD SHEDDING IN
POWER SYSTEM

2.1 Introduction
2.2 Emergency control load shedding in power system
2.2.1 The proposed load shedding method [63], [64]
The principle model of the load shedding method based on the quick
identification of the state with / without load shedding is shown in Figure 2.1 [63]
and the detailed model is shown in Figure 2.2 [64].

Figure 2.1: The principle model of emergency control of load shedding

The first Artificial Neural Network (ANN1) will perform identification

with/without load shedding. If the output of ANN1 has load shedding, the
selector will activate to allow the second Artificial Neural Network (ANN2). The
ANN2 will perform identification of LSi load shedding strategies (i = 1, n) to
control the selection of load shedding strategies. These load shedding strategies
are built upon the AHP algorithm.
2.2.2 Training dataset and training ANN1
2.2.3 The load shedding control strategies based on AHP algorithm
2.2.4 Training the ANN2
2.2.5 Simulation - Test the proposed load shedding method on standard
power system diagram

The proposed load shedding method is simulated on the standard IEEE 10-
generator 39-bus test system with the support of PowerWorld GSO 19 software
for fault case at Bus no.25 and Bus no.32. Frequency simulation results and rotor

3
angle when performing load shedding according to the proposed load shedding
method are shown in Figure 2.9 and Figure 2.10.

Figure 2.9: The frequency of system after applying the proposed method (the
case study is short circuit at Bus no.32)

Figure 2.10: The rotor angle of the generators after applying the proposed
method (the case study is short circuit at Bus no.32)

The results of comparison between the proposed shedding method and the
traditional load shedding method [9] are presented in Table 2.21.

4
 Table 2.21: The comparison between the proposed load shedding method
and the UFLS method
Frequency Recovery Time Load shedding
(Hz) (s) (MW)
Short circuit at Bus 32
The proposed load 60,028 50 628,2
shedding method
The UFLS method 60,055 78 780,4
Short circuit at Bus 25
The proposed load
60,0455 40 438,9
shedding method
The UFLS method 60,0750 70 448,9

Analysis of simulation results in Figure 2.8, Figure 2.10, Figure 2.14,

Figure 2.15 shows the implementation of the proposed load shedding strategy to
keep the power system stable, restore frequency to allowable values. Figures 2.9,
Figure 2.13, Figure 2.16, Figure 2.17 and Table 2.21 show the recovered
frequency value in the range from 60,028Hz to 60,0455Hz. Meanwhile, the
UFLS method has more load shedding capacity than the proposed method. This
value is from 2.28% to 24.2%. In addition, the UFLS method also has a slower
frequency recovery time than the proposed method. This value is from 10s to 28s.
Thus, the quick identification of with/without load shedding has accelerated the
decision-making of load shedding. This helps to stabilize electrical system
frequency, the recovery frequency to allowable value and the recovery time faster
than the UFLS method.

5
 Chapter 3 LOAD SHEDDING METHOD BASED ON
APPLYING FUZZY-AHP ALGORITHM

3.1 Introduction
3.2 Fuzzy technique and operating law [69]
3.3 Overview the Fuzzy – AHP algorithm [72]-[74].
3.4 Simulation - Test the proposed load shedding method on the standard
IEEE 10-generator 39-bus test system
To compare the efficiency of the load shedding method based on the
Fuzzy-AHP algorithm and the load shedding method based on AHP algorithm,
the proposed algorithm is tested on the standard IEEE 9-generator 37-bus test
system [75] in both case studies. In case of study, a generator at Bus no.4 is
disconnected, respectively the system is operating in the state of 70%, 80%, 90%
and 100% of maximum load.
3.4.1 Load shedding method based on application of AHP algorithm
3.4.2 Load shedding method based on application of Fuzzy-AHP
algorithm [76].
Similar to the case study presented in section 3.4.1, a generator at Bus no.
4 is disconnected while the system is operating at different load levels. This
causes the generation power to be limited and the frequency value less than the
permissible value. Follow steps of Fuzzy-AHP algorithm model which is
presented in Section 3.3.

Rank in ascending order the importance of load units. In this ranking, the
lower weight load will have priority for load shedding in control strategies. The
load shedding will be carried out until the restore frequency value is greater than
the allowable value. Calculation results for the load levels are presented in Table
3.34.
Fuzzy approach for load profile corresponds to the rate of 70%, 80%, 90%
and 100% of the maximum power load.

6
 Figure 3.1: Fuzzy approach for load profile

In case of studying the load operating at 83% of maximum load, the fuzzy
approach results show the value of µ2> µ3, so choose CS2 and the control strategy
CL2.

The proposed shedding method is compared with the load shedding

according to AHP algorithm. The standard IEEE 9-generator 37-bus test system
is applied to test case studies. In both cases, the frequency is restored to the
allowed value and reached a value close to the rated frequency. However, the
Fuzzy-AHP method has smaller total load shedding power. Because, the Fuzzy-
AHP algorithm helps in more precise weight selection. Specifically, the load
shedding capacity decreased from 9.44% to 10.5%. In the competitive electricity
market, this helps to reduce damages from power outages and makes a lot of
economic sense. In this case, the frequency control processes of the generators
are considered. The details of this process are presented in Chapter 4. The
comparison of load shedding results between AHP method and Fuzzy-AHP
method are presented in Table 3.38.

7
 Table 3.31: The comparison of load shedding results between AHP method
and Fuzzy-AHP method
Load
shedding AHP algorithm Fuzzy-AHP algorithm
method
Control Control Control Control Control Control Control Control
strategy strategy strategy strategy strategy strategy strategy strategy
1 2 3 4 1 2 3 4
Loads are L2, L4, L2, L4, L2, L4, L2, L4, L3, L2, L3, L2, L3, L2, L3, L2,
cut L25, L5 L25, L5 L25 L25 L4, L8 L4, L8 L4 L4
Load
shedding 127,35 143,69 161,64 179,62 113,98 130,11 146,38 162,65
(MW)
Recovery
Frequency 59,94 59,99 59,95 60 59,91 59,95 59,86 59,89
(Hz)
Recovery
35s 40s 44s 50s 35s 40s 40s 45s
time (s)

When considering the frequency recovery time, both methods have the
following characteristics: the higher the load, the longer the frequency recovery
time. However, the Fuzzy-AHP method has a faster frequency recovery time than
the AHP method.

8
Chapter 4 CALCULATION THE MINIMUM LOAD SHEDDING
POWER CONSIDERS PRIMARY AND SECONDARY CONTROL
PROCESS

4.1 Introduction
4.2 Overview the power system frequency response
4.3 Frequency adjustment process when there is a failure in power system
4.4 The primary frequency control in power system
In case the generation unit is equipped with a governor, the power
generation characteristics are shown in the (A) characteristic and Figure 4.3.

Figure 4.3: Frequency adjustment process in power system

4.5 The secondary frequency control in power system
4.6 Calculate the minimum load shedding capacity to restore frequency to
the permissible value
4.6.1. Purpose of calculating minimum load shedding capacity
The calculation of the minimum load shedding power PLS min ensures the
least amount of load shedding, reduce the least damage to customers and restore
frequency to the permissible value.
4.6.2. Formulas to calculate the minimum load shedding capacity
9
 When a generator is disconnected, the power difference between the
generation and the load PL will cause frequency difference. At this time, the new
power balance equation is presented as follows:
n 1 n 1
PL  PD   PGi   PPrimary control (4.6)
i 1 i 1
n 1  PGn,i f1
n 1
PL   PGi  PD   . (4.7)
i 1 i 1 Ri f0
n 1  P
n 1
f Gn , i f1
PL   PGi  ( 1 ).PL .D   . (4.8)
i 1 f0 i 1 Ri f0
n 1
f n 1 P
PL   PGi  ( 1 )( PL .D   n ,i )
G
(4.9)
i 1 f0 i 1 Ri
n-1 n 1 P
With PL  PL -  PG and   P .D   Gn ,i
i L
i 1 i 1 Ri

From formula (4.9): PL  f1 . (4.10)

fn

When the system considers the secondary control power to restore the
frequency, the new power balance state with the new frequency value f2, formula
(4.6) becomes:
n 1 n 1
PL  PD   PGi   PPrimary control  PSecondary control max (4.11)
i 1 i 1

Where, PSecondary control max is the maximum power of secondary control.

This power is determined by the (4.12) equation:
k
PSecondary control max   ( PGm, j  PPrimary control, j ) (4.12)
j 1
Where, PGm , j is the maximum generation power of the jth secondary
frequency control generator, PPrimary control, j is the jth primary control power of
the generator.
After performing the secondary control process but system frequency has
not yet restored to the permissible value fcp, load shedding must be performed to
restore the frequency, the PLSmin is calculated according to the following formula:

10
 n 1 n 1
PL  PD  PLS min   PGi   PPrimary control  PSecondary control max (4.13)
i 1 i 1
n 1 n 1
PLS min  PL  PD   PGi   PPrimary control  PSecondary control max (4.14)
i 1 i 1
n 1 f cp n 1 PGn ,i f cp
PLS min  PL   PGi  .PL .D   .  PSecondary control max (4.15)
i 1 f0 i 1 Ri f 0

Where: f cp  f 0  f cp is the allowable frequency attenuation.

Equation (4.14) is reduced to the following formula:

f
PLS min  PL  cp .  PSecondary control max
(4.16)
f0
4.7 Calculation and test on standard power system diagram
The method to calculate the load shedding capacity was tested on the
standard IEEE 9-generator 37-bus test system [75]. This calculation applies the
formula (4.16) to calculate the minimum load shedding capacity to restore
frequency to the acceptable value.

The calculated minimum load shedding capacity PLoad shedding min is 17,64MW.

Based on the PLoad shedding min, calculation and simulation results have ensured
that the power system frequency can be restored to the allowable value of 59,7Hz.

11
Chapter 5 LOAD SHEDDING METHOD CONSIDERS THE
MULTIPLE METHODS COORDINATION

5.1 Introduction
5.2 The load shedding distribution at the load buses is based on the PED
concept
5.2.1 Phase Electrical Distance (PED)
PED calculation method is performed according to the following steps:

Step 1: Extract the Jacobian matrix [ J P ].

1
Step 2: Inverse elements in the Jacobian matrix [ J P ], get the [ J P ]
matrix

Step 3: Apply the formula (5.4) to calculate Dp(i,j)

Step 4: Filter the matrix to find the relationship between generator buses
and load buses

Step 5: Ranking the loads is done according to the increasing PED value
between the load buses and the outage generator.

The formula for calculating the load shedding capacity at the load buses
according to PED is presented in the formula (5.7) [86]:
DP ,eq
PLSi  .PLS min (5.7)
DP ,mi
With 1 (5.8)
DP ,eq 
1
im D
P , mi

5.2.2 Simulation – test the proposed method on standard power system

diagram
The PED relationship between JO345 # 1 generator and load buses is
shown in Figure 5.1.

12
Figure 5.1: The PED relationship between JO345 # 1 generator and load buses

The efficiency of the proposed load shedding method is compared with the
UFLS method [9]. The comparison results of frequency and phase angle of the
proposed method and the UFLS method are presented in Figure 5.2 and Figure
5.3.

Figure 5.2: The comparison of frequency after load shedding of the proposed
method and UFLS method

13
Figure 5.3: The comparison of the rotor angle after load shedding between the
proposed method and the traditional method

The proposed shedding method has much smaller load shedding power than
UFLS method, specifically the load shedding power reduces from 82,93MW to
17,64MW. Here, the recovery frequency value of the proposed method is lower
than the UFLS method. However, this value is still within the allowable and
acceptable range (59,7Hz). Especially, the recovery time of the phase angle of
the proposed method is equivalent to the UFLS method.
5.3 The load shedding distribution at the load buses is based on the VED
concept
5.3.1 Voltage Electrical Distance (VED)
Formula (5.12) can be considered as a definition of the VED between two
buses i and j [81], [87].
DV  i, j   DV  j , i    Log  ij *  ji  (5.12)
Formula to calculate the distribution of load shedding at buses based on
VED [88]:
DV ,eq
PLSi  .PLS min (5.13)
DV, mi

Where, m is the mth generator; i is the ith load; PLSi is the amount of load
shedding (MW); PLS min is the minimum amount of load shedding to restore
frequency to permissible value (MW); DVmi is the VED of the ith load corresponds

14
to the generator outage; DV,eq is the VED equivalent of all load and generator
buses.
1 (5.14)
DV ,eq 
1
im D
V , mi

5.3.2 Simulation - test the proposed method on standard power system

diagram
The proposed load shedding method is compared with the UFLS method
[9]. The load shedding method based on the VED, the voltage value at the nodes
is always greater than or equal to the voltage value at the load nodes when load
shedding based on PED. At several load nodes such as Bus 10, Bus 13, Bus 16,
Bus 18, Bus 21, Bus 27, Bus 47, Bus 48, Bus 53, the recovery voltage values are
higher than the load shedding method based on PED. Thereby, it shows the
effectiveness of the recovery voltage value of the VED-based load shedding
method.

The comparison results show that in both methods of load shedding, the
recovery frequency values of the two methods are almost equal due to the same
amount of shedding power. However, the recovery voltage value of the VED-
based load shedding method is better than the PED-based load shedding method.
The phase angle parameter of the PED-based load shedding method is faster than
the VED-based load shedding method. Therefore, depending on the case of
voltage recovery or phase angle recovery, one of two methods of load shedding
can be chosen.

5.4 The load shedding method considers the combination of many methods
applying AHP algorithm and expert system

The flowchart of a combination many methods to rank and distribute the

amount of shed power for the loads is shown in Figure 5.11.

The procedure and steps are as follows:

15
 Step 1: Determine the number of criteria to be satisfied when load
shedding.

Step 2: Calculate the value of each criterion and data normalization.

Step 3: Apply the AHP algorithm to calculate the weight of each criterion.

Step 4: Calculate the combined weight of each load. This value is

calculated by multiplying the weight of each criterion by the value of each
normalized criterion.

Step 5: Ranking and distribution the amount of load shedding capacity at

load nodes based on combined weight.

Figure 5.11: The flowchart performs a combination of methods for ranking and
distributing the amount of shed power for loads
5.4.1 Criterion 1: The loads importance factor
5.4.2 Criterion 2: PED
The purpose of calculation is to focus load shedding at the load nodes
around or near the generators that have stopped working. The PED of two buses
16
is calculated using the procedure suggested in section 5.2.1. The PED of each
load node is normalized using the equation (5.15):
D (i, j )
WDP (i , j )  25 P (5.15)
 DP (i, j)
1

Where, WSP (i , j ) is the PED from the ith bus to the jth outage generator after
normalization; DP (i, j ) is the PED before normalization.
5.4.3 Criterion 3: VED
The VED between two buses is calculated using the proposed procedure in
section 5.3.1. The calculation results of VED are shown in Table 5.6. The VED
of each load node is normalized using the formula (5.16):
D (i, j )
WDv (i , j )  25 v (5.16)
 Dv (i, j )
1

Where, WDv (i , j ) is the VED of the ith bus to the outage generator after
normalization; Dv (i, j ) is the VED from the ith load bus to the generator.

The load shedding problem in this thesis considers the combination of 3

criteria as follows:

Criterion 1: The loads importance factor.

Criterion 2: PED from the load to the generator

Criterion 3: VED from the load to the generator

If the expert's opinion is perfect and M is built completely as mentioned,

the thesis applies the method of calculating eigenvalue of the M matrix, the
values W:

MW=nW, W=(w1, w2, w3), n is eigenvalues of M (5.17)

Applying the above theory and expert opinion, the matrix M is built as
follows:

17
 1 3 2 

M  1/3 1 1/2
1/2 2 1 
The eigenvector of the M judgment matrix has the following values:
W   W1 , W2 , W3   0,53962;0,16342;0, 29696
T T

Thus, there are 3 criteria to consider a load. If the importance of the criteria
is different, for each load applies the theory [89]: From the weighted values for
each criterion Wi of the W matrix, recalculate the weighted values in each load
using distribution multiplication, then adding the weights of the aspects together
with each load bus will get the final weight, the formula is calculated as follows:
 A~ (1 ,..., n )   Wi .WD, j
n
(5.18)
i 1

Where, i is the combined weight value of each load, Wi is the weighted

values of the M matrix, WD , j is the value of WFAHP, WDv (i , j ) , WDP ( i , j ) .

Based on the combination weights in the ranking table, the amount of load
shedding at the buses will be calculated as follows:
eq
PLSi  .PLS min (5.19)
A~
I

Where, PLSi is the amount load shedding power at the buses;  eq is the
equivalent weight of all load buses, A ~ is the combined weight at the i bus; PLSmin
th
i

is the total minimum load shedding power.

The comparison result between the proposed method and the UFLS method
is presented in Figure 5.12.

18
Figure 5.12: The frequency after load shedding based on the proposed method
and UFLS method

The UFLS method has better frequency response than the proposed load
shedding method. However, the recovery frequency value of the proposed load
shedding method is still within the allowable value.

The comparison result of the rotor angle of generator before and after load
shedding is shown in Figure 5.13. The rotor angle of generator after load
shedding according to the proposed method and the UFLS method is shown
Figure 5.14.

19
 Figure 5.13: The rotor angle of generator before and after load shedding
according to the proposed method

Figure 5.14: The rotor angle of generator after load shedding according to the
proposed method and UFLS method

The comparison results of the load shedding methods are presented in

Table 5.10.

20
 Table 5.10: The results of the comparison of the load shedding methods
Recovery time Recovery Load shedding
(s) Frequency (Hz) (MW)
Under
Frequency Load
10 59,87 82,83
Shedding
(UFLS) method
The proposed 17,64
18 59,7
method

The frequency response of the UFLS method is better than the proposed
method. However, the recovery frequency value of the proposed method is still
within the allowable and acceptable range (59,7Hz). In addition, the proposed
method has smaller load shedding capacity (65,19MW) than the UFLS method.
This minimizes damage caused by generator outages, and satisfies economic-
technical constraints: the importance of the load, the PED, and VED. Besides,
the recovery time of the phase angle of this proposed method is equivalent to the
UFLS method. The reason is large load shedding at the load nodes close to the
generator outage, making the phase angle recover faster. So it proved the
effectiveness of the proposed method.

21
Chapter 6 CONCLUSIONS AND FUTURE WORK

6.1 Conclusions
Based on the research objectives and tasks, the thesis has completed the following
scientific contents:

- The thesis has analyzed and evaluated the previous load shedding
studies. Thereby, we can see the existing problems in order to propose methods
to improve load shedding in power system.

- The dissertation has proposed the method of load shedding in emergency

situations such as short-circuit on the buses or on the transmission lines of power
system. This is based on building a fast identification system with/without load
shedding and layering strategy of load shedding based on neural network and
AHP algorithm application. The results of calculation and simulation on the
standard IEEE 10-generator 39-bus test system show that the implementation of
the proposed load shedding strategy has supported the electrical system to keep
frequency stability after the incident. Load nodes are ranked based on the
importance factor and calculated by the AHP algorithm. The proposed shedding
method has smaller load shedding capacity, and faster frequency recovery time
than UFLS method. In addition, the electrical system has been kept stable and the
frequency of the power system has been restored to permissible values.

- The thesis has researched and presented the load shedding method based
on consideration of the load's importance factor. The proposed load shedding
method applied Fuzzy-AHP algorithm to determine the importance factor of the
load and prioritize the less importance of load shedding. Thereby, this method
has supported to minimize the economic damage caused by load shedding. In
addition, blurring the load profile has helped reduce the number of load shedding
control strategies. The simulation results on the standard IEEE 9-generator 37-
bus test system show that in the case of research, the frequency restored to the
allowed value.

22
 - The thesis presents the calculation of the load shedding included in the
primary and secondary control factors of the generators will minimize the amount
of load shedding and restore system frequency value back to the allowable range.
When we tested on the IEEE 9-generator 37-bus test system, the proposed
method has smaller load shedding than the UFLS method.

- The thesis has proposed the distribution of the load shedding capacity at
the load nodes based on the concepts of PED and VED. The test results on the
IEEE 9-generator 37-bus test system have shown good results. Although the
capacity of the proposed load shedding method is smaller than the UFLS method,
the recovery time of the phase angle, and the voltage values are equivalent to the
UFLS method.

- The thesis has proposed the method of load shedding based on

consideration of many criteria such as the importance factor of the load, PED,
and VED. The AHP algorithm was applied to calculate the weight of each
criterion and the combined weight. The results of calculation and simulation
show that the proposed load shedding method enables to restore frequency to the
allowable value. This method has smaller load shedding capacity than UFLS
method. Thereby, it helps minimize the damage caused by power outages, and at
the same time satisfy the economic - technical constraints criteria.

6.2 The Future Work

- In the research scope, the thesis only considers the case of equal increase of
the load. The topic has not considered the load that changes continuously over
time. The future work should consider the load models that change continuously
over time to make the problem more diverse.

- Considering the load in the scope of this thesis has not considered the
percentage of the load according to the reliability of power supply. The future
work should consider the percentage of load of category 1, load of category 2,
and load of category 3.

23
 - The damage consideration in the scope of this thesis only considers the load
damage. The future work should consider all costs: generation, backup sources,
as well as the impact of FACT devices.

- Building a neural network can self-learn and self-update training data to

meet the changes of power system.

24