Minimization of Starting Energy Loss of Three Phase Induction Motors Based on Particle Swarm Optimization and Neuro Fuzzy Network

International Journal of Power Electronics and Drive System (IJPEDS)
Vol. 7, No. 4, December 2016, pp. 1038~1048
ISSN: 2088-8694, DOI: 10.11591/ijpeds.v7i4.pp1038-1048  1038
Journal homepage: http://iaesjournal.com/online/index.php/IJPEDS
Minimization of Starting Energy Loss of Three Phase Induction
Motors Based on Particle Swarm Optimization and Neuro
Fuzzy Network
Mahmoud M. Elkholy, Mohamed A. Elhameed
Electrical Power and Machines Department, Faculty of Engineering, Zagazig University, Zagazig, Egypt
Article Info ABSTRACT
Article history:
Received May 18, 2016
Revised Oct 24, 2016
Accepted Nov 5, 2016
The purpose of this paper is to minimize energy losses consumed by three
phase induction motors during starting with wide range of load torque from
no load to full load. This will limit the temperature rise and allows for more
numbers of starting during a definite time. Starting energy losses
minimization is achieved by controlling the rate of increasing voltage
and frequency to start induction motor under certain load torque within a
definite starting time. Optimal voltage and frequency are obtained by particle
swarm optimization (PSO) tool according to load torque. Then, outputs of the
PSO are used to design a neuro-fuzzy controller to control the output voltage
and frequency of the inverter during starting for each load torque. The
starting characteristics using proposed method are compared to that of direct
on line and V/F methods. A complete model of the system is developed using
SIMULINK/MATLAB.
Keyword:
Induction motor
Neuro-fuzzy network
Particle swarm optimization
Starting energy losses
Copyright © 2016 Institute of Advanced Engineering and Science.
All rights reserved.
Corresponding Author:
Mahmoud M. Elkholy,
Electrical Power and Machines Department, Faculty of Engineering,
Zagazig University, P.O. Box 44519 Zagazig, Al Bahr Street, Zagazig, Egypt.
Email: melkholy71@yahoo.com
1. INTRODUCTION
Three phase induction motors are used extensively in industry due to rigidness, less maintenance
and fault tolerant. When these motors are connected directly to supply, they draw large currents and dissipate
large amount of energy. This results in more voltage drop across the network elements and more heat for the
motor itself especially in multi-start operations [1].
Many researches discussed the problem of motor starting [2-8], some of which discussed soft
starting to produce less currents and no sudden torques [9, 10], soft starting may enhance starting energy, but
the save is not optimum. For example, in [11] the save is not more than 4 % at light load and there is a
negative save with torques greater than 60 % of full load torque. In [12], a neuro-fuzzy soft starter is used to
reduce energy losses by adjusting the firing angles of the thyristors of an AC voltage controller, but optimal
conditions and effect of frequency variation are not discussed. In [13], genetic algorithm is used to optimize
the energy of starting by defining the appropriate voltage ramp during starting, the save in energy reaches
20 %, but frequency variation is not investigated in the paper. Adaptive and optimal control of induction
motor using PSO and neuro fuzzy controller is proposed in [14-16].
In this paper, particle swarm optimization method is used to obtain the rate of change of both
voltage and frequency over the full range of load torques from no load to full load to have minimum starting
energy losses. There are constraints such as both applied voltage and frequency must equal to the rated values
at the end of starting period and the motor must be started within a certain specified time. The output of the
optimization process is used to design a neuro-fuzzy controller to control the voltage and frequency during
starting depending on load torque so that the proposed system is optimal and adaptive.

IJPEDS ISSN: 2088-8694 
Minimization of Starting Energy Loss of Three Phase Induction Motors Based on … (Mahmoud M. Elkholy)
1039
2. MATHEMATICAL MODEL
Using Laplace transformation the voltage equations of three phase squirrel cage induction motor in
d-q frame are [17]:
( )( ) ( ) ( ) ̇( ) ( ) ̇ ( ) ( )
(1)
( )( ) ( ) ( ) ̇( ) ( ) ̇ ( ) ( )
( )
( )( ) ( ) ̇ ( ) ( ) (3)
( )( ) ( ) ̇ ( ) ( ) (4)
Where , , and
is the d-axis stator voltage, is the q-axis stator voltage, is the d-axis stator current, the q-axis
stator current, is d-axis referred rotor current, is q-axis referred rotor current, is the self inductance
of stator phase winding, is the self inductance of rotor phase winding and is the mutual inductance
between stator and rotor windings. Thed-q flux linkages of stator and rotor are defined by:
(5)
(6)
(7)
(8)
The electromagnetic torque equation is:
( ) (9)
The motor mechanical equation is:
( ) ̇( ) (10)
Where; is the number of poles, is the electromagnetic torque of the motor, is the load torque, is the
moment of inertia Kg.m2
and is the friction coefficienct. The model of three phase squirrel cage induction
motor is developed using SIMULINK /MATLAB to solve above nonlinear equations and to study the
dynamic performance characteristics of the motor during starting. The SIMULINK dynamic model of the
motor is shown in Figure 1.a. The starting energy losses ( ), stator copper losses ( ) and iron losses
( )of the motor are defined as:
∫( ) (11)
(12)
(13)
Where; is the stator phase current, is the referred rotor phase current, is the stator phase voltage, is
the resistance of stator phase winding, is the referred resistance of rotor phase winding and is the core
loss resistance. Motor voltage and frequency are changed during starting according to the equations:

 ISSN: 2088-8694
IJPEDS Vol. 7, No. 4, December 2016 : 1038 – 1048
1040
(14)
(15)
Where; are constants.
The maximum values of voltage and frequency are the rated values of the motor. The ratio of voltage to
frequency must be limited to the ratio of rated voltage and frequency to prevent motor saturation
and overheating. Figure 1.b shows the SIMULINK model with these variable voltage and frequency.
a. SIMULINK Model of Three Phase Induction Motor b. SIMULINK System Scheme
Figure 1. Developed SIMULINK Model
3. OPTIMUM VOLTAGE AND FREQUENCY VARIABLES USING PSO
In this part PSO method is used to determine the constants Kv1, Kv2, Kf1 and Kf2 in Equations
(14) and (15). The objective function is Equation (11) of starting energy losses with the inequality
constraints of:
a. .
b. Motor voltage and frequency must equal rated values at the end of starting time.
Figure 2 shows the flow chart of the PSO operation, for a certain load torque, a swarm of 24 agents
is initialized, for each agent the motor dynamic model is operated and the objective function is evaluated.
New position of Agents is determined according to their velocities, their best position and the best position of
the swarm. Agent's velocity in swarm is updated according to Equation (16):
( ) ( ) (16)
where is velocity of agent i at iteration k, is weighting function, is weighting coefficients, is
random number between 0 and 1, is the current position of agent i at iteration k, is best position of
agent i, and is best position of the swarm. The weighting function is given by:
(17)
Where is initial weight, is final weight, is maximum iteration number, and is current
iteration number. According to Shi and Eberhart [18, 19], the following parameters are appropriate and do
not depend on optimization problems:
(18)

1041
Simulation has been carried out by using SIMULINK/MATLAB for 220/380 V, 1.1 kW, 50 Hz
three phase induction motor having: = 5.15 Ω, = 3.75 Ω, =0.5887 H, =0.5887 H, =0.5568 H,
=2 and J=0.05 kg.m2
. PSO are used to obtain the optimum voltage and frequency variables, maximum
number of iterations is = 50. The process is repeated for a load torque range from 0.2 to 3 N.m in a
step of 0.2 N.m. For each load torque, PSO is run for five times and the best solution is recorded. The optimal
values of , , and are listed in Table 1. Figure 3 and Figure 4 show the PSO fitness function
results at load torque of 1 and 2 N.m respectively. The optimum values of motor voltage and frequency
obtained by PSO are shown in Figure 5 and Figure 6. The optimum values of voltage and frequency are
varied with different values of load torque, therefore a neuro-fuzzy technique is used to adapt theseoptimum
with load torque.
Table 1. Optimal Motor Voltage and Frequency Parameters
TL(N.m) 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
16.153 15.873 15.06 14.574 14.1776 14.123 13.98 13.827 13.483 13.173 12.517 12.279 11.407 11.274 11.266
58.702 61.595 69.627 75.411 79.806 83.253 87.311 91.556 92.11 96.958 100.368 107.552 114.336 115.359 120.034
5.232 5.015 4.854 4.69 4.49 4.333 4.17 4.019 3.808 3.645 3.416 3.281 3.041 2.884 2.735
11.741 12.32 13.925 15.082 15.961 16.651 17.47 18.317 18.423 19.392 20.135 21.513 22.898 23.095 24.035
Start
Read motor and swarm initial data
Initialize the position of each agent
Run Simulink model
Evaluate the objective function
Update agents position
Iter ≤ iter_max
Yes
End
No
Write the best solution
Figure 2. Flow Chart of the PSO Method
Figure 3. Variation of PSO Fitness Value with
Generation at Load Torque 1 N.m
Figure 4. Variation of PSO Fitness Value with
Generation at Load Torque 2 N.m

 ISSN: 2088-8694
1042
Figure 5. Variation of Motor Phase Voltage with Time at
Different Load Torques
Figure 6. Variation of Motor Frequency with Time at
Different Load Torques
4. ADAPTIVE NEURO-FUZZY CONTROLLER FOR MOTOR INVERTER
Sugeno fuzzy system [20], is more compact than Mamdani system and is used for adaptive
techniques for constructing fuzzy models. These adaptive techniques can be used to customize the
membership functions so that the fuzzy system best models the data. The concept of adaptive neuro-fuzzy
inference system ANFIS is used to obtain the required voltage and frequency values. The optimal values of
voltage and frequency depend on the load torque. A neuro-fuzzy network is trained to perform this
adaptation. The input to this network is load torque, and the outputs are the four constants of voltage
and frequency equations. An ANFIS is built for each constant in the voltage and frequency equations,
i.e , , and . The four networks have the same structure of a hidden layer with 6 neurons and are
trained for 1000 epochs, Table 2 shows the training statistics for these networks. Figure 7 shows the training
pattern for the networks. The output of the ANFIS is compared with the output in the training set and is
shown in Figure 8. A Gaussian membership functions are used to fuzzy input data and a linear function is
used for the output data. A back propagation algorithm is used to adapt membership functions so that the
error goal between targets and outputs of ANFIS is achieved. Complete Simulink model of the system with
neuro-fuzzy controller is shown in Figure 9.
Table 2. Training Statistics of ANFIS Networks
Parameter Max. error Mean error Standard Deviation
Kv1 3.1422 ×10-4
4.9687 ×10-6
1.6883 ×10-4
Kv2 3.4055 × 10-4
1.4188 ×10-5
1.7557 ×10-4
Kf1 2 ×10-3
1.6843 ×10-6
1.3 ×10-3
Kf2 8.0540×10-5
2.8264×10-6
3.5745×10-5
Figure 7. Variation of , , and ANFIS
Error with Training Epochs
Figure 8. Variation of , , and ANFIS
Output, target

1043
Figure 9. SIMULINK Model of the System with Neuro-Fuzzy Controller
5. RESULTS AND DISCUSSION
Simulation has been carried out by using SIMULINK/MATLAB for two different motors. Motor A
is a 220/380 V, 1.1 kW, 50 Hz three phase induction motor having the following parameters: = 5.15 Ω,
= 3.75 Ω, =0.5887 H, =0.5887 H, =0.5568 H, =2 and J=0.05 kg.m2
. Motor B is a 220/380 V,
2 hp, 50 Hz three phase induction motor having the following parameters: = 5.6 Ω, = 4.965 Ω,
=0.0.2611 H, =0.2611 H, =0.2445 H, =4 and J=0.2 kg.m2
.
5.1. Results of Motor A
In this section three groups of results are presented for motor A. The first group represents results
when the motor is started by the proposed method where voltage and frequency are changed as given in
Equations (14) and (15) with constants , , and as listed in Table 2. The second group
represents results of DOL method. The third group represents results when the motor is started with V/F
method where frequency is increased with ramp rate from 0 Hz to rated frequency of 50 Hz in 10 s (this is the
factory setting of EUROTHERM DRIVES 605 Series Frequency Inverter) and voltage is given as:
(19)
Figure 10.a shows the variation of motor energy losses with the three starting methods at different
load torques ranges from 0.2 N.m to 3 N.m with a step of 0.2 N.m. It's shown that the proposed technique
reduces starting energy losses much more than DOL and V/F starting. This reduction in starting energy losses
leads to energy saving especially in multi-starting application.
The variation of motor speed and developed torque with time at different load torques are shown in
Figures 10.b and 10.c respectively. In case of V/F starting method with high load torques, and at low
frequency the developed torque is lower than load torque so that the motor failed to start and motor starts
when the frequency reaches its rated value.
The variation of stator current with time at load toque of 0.2 N.m in case of the three methods of
starting is shown in Figure 10.d. The starting current in case of V/F method is lower than that of the proposed
and DOL methods but the starting time is the highest. Starting energy losses with proposed method are the
lowest one at all range of load torques as shown in Figure 11. The percentage saving in starting energy losses
of proposed method compared to other methods are shown in Table 3 and calculated as:
(20)
(21)

 ISSN: 2088-8694
1044
(22)
In modified V/F method the voltage and frequency are boosted by 10 % at starting to improve the
starting torque as:
( ) ( )
(23)
Table 3. % Saving in Starting Energy Losses of Motor a of Proposed Method Compared to DOL and V/F
Methods
Load
Torque
(N.m)
% Saving in Starting Energy losses
compared to DOL
compared to V/F
% Saving in Starting Energy
losses compared to modified
V/F
2.0 43.9243 43.7223 22.4533
2.3 44.3023 34.2.90 23.0389
2.. 40.4424 37.4994 28.6780
2.3 42...23 .4.4339 36.1000
4.2 09..7.7 .4.0.94 44.4024
4.0 03..742 .4..444 52.3407
4.3 07..940 .2.2442 54.3077
4.. 0..92.3 33.324. 52.8746
4.3 07.7949 37.3790 52.3544
0.2 07.7.34 3...4.2 51.0622
0.0 03..44. 3..9277 50.1629
0.3 09.2430 33.39.. 48.9659
0.. 42.3442 33..444 48.3310
0.3 43.3370 3..0433 49.5671
4.2 43.3020 33.499. 51.0327
Figure 10.a Variation of Motor Energy Losses with
Time at Different Load Torques (motor A)
Figure 10.b. Variation of Speed with Time at Different
Load Torques (motor A)
Figure 10.c Variation of Developed Torque with Time at Different Load Torques (motor A)

1045
Figure 10.d Variation of Stator Phase Current with Time at Load Torque of 0.2 N.m (motor A)
Figure 11. Variation of Starting Energy Losses with Time at Different Load Torques (motor A)
5.2. Results of Motor B
In this section the same results are repeated for motor B. The variation of motor energy losses, speed
and motor torque are shown in Figures 12.a to 12.c respectively. It’s shown that the motor energy losses of
proposed method are the smallest. In V/F method, the motor fails to start when loaded by a load torque
equals or greater than 5 N.m which is 50% full load up to the motor voltage and frequency are increased to
rated values. Therefore, the motor energy losses of V/F method are the highest.The variation of motor current
with time at no load torque is shown in Figure 12.d. The starting current of V/F method is the lowest one but
the developed torque is small so that the motor starting time is high and the corresponding energy losses is
the highest one as shown in Figure 13. The optimal values of constants , , and of this motor
from the PSO as listed in Table 4. The percentage saving in starting energy losses of motor B of proposed
method compared to other methods are shown in Table 5.
Figure 12.a. Variation of Motor Energy Losses with
Time at Different Load Torques (motor B)
Figure 12.b. Variation of Speed with Time at
Different Load Torques (motor B)

 ISSN: 2088-8694
1046
Figure 12.c. Variation of Developed Torque with Time
at Different Load Torques (motor B)
Figure 12.d. Variation of Stator Phase Current with
Time at No Load torque (motor B)
Figure 13. Variation of Starting Energy Losses with Time at Different Load Torques (motor B)
Table 4. Optimal Motor Voltage and Frequency Parameters of Motor B
TL(N.m) 0 2.5 5 7.5 10
13.573 10.1234 8.0915 9.9001 9.3768
84.271 128.0304 148.0665 208.2964 208.7867
5.2718 4.2573 3.5152 2.1579 1.8459
16.8559 25.6061 29.7303 41.6615 41.7989
Table 5. % Saving in Starting Energy Losses of Motor B of Proposed Method Compared to DOL and V/F
Methods
Load Torque
(N.m)
compared to DOL
compared to V/F
compared to modified V/F
0 49.03.0 03.4409 04.0709
2.5 42.0..4 43.3332 09.2444
5 ..4394 30.7.97 33.47.3
7.5 ..3333 47.99.3 34.4233
10 7.0020 44.42.4 43.39.0
6. CONCLUSIONS
In this paper minimization of starting energy losses of three phase squirrel cage induction motor is
achieved by controlling the rate of change of applied voltage and frequency during starting. The optimum
values of voltage and frequency rates are obtained by particle swarm optimization (PSO) to have minimum

1047
starting energy losses with constraints of keeping the rated V/F ratio, reaching rated voltage and frequency at
steady state and limiting starting time to 10 sec. The suitable values of voltage and frequency are varied
according to load torque during starting. Therefore, optimum values are obtained over a wide range of load
torque from no-load to full load, a neuro-fuzzy controller is designed to adapt the rate of change of voltage
and frequency. To validate the proposed method, it is applied to two three phase induction motors, starting
energy losses in both cases are lower than that of direct on line method and conventional V/F method at
different load torques.
REFERENCES
[1] M. Mohammadi, A. Mohammadi Rozbahani, S. Abasi Garavand, M. Montazeri, H. Memarinezhad. Fuzzy Bang-
Bang Control Scheme of USSC for Voltage Sag Mitigation due to Short Circuits and Induction Motor Starting in
Distribution System. International Journal of Power Electronics and Drive System (IJPEDS). 2014; 4(4): 451-460
[2] Zhou Jing-hua, Chen Cheng, Zhang Xiao-wei and Chen Ya-ai. Reducing Voltage Energy-saving Control Method of
Induction Motor. International Conference on Electrical Machines and Systems, Busan, Korea, 2013; 2159-2162.
[3] Alexei V. Udovichenko. New Energy Saving Multizone Alternating-Voltage Soft Starters of Induction Machines.
Micro/Nanotechnologies and Electron Devices (EDM), 2011 International Conference and Seminar of Young
Specialists, Erlagol; 2011: 415-419.
[4] Nafeesa K and Saly George. Optimization of Starting Performance of Thyristorized Static Switch Fed Three Phase
Induction Motor. Power Electronics, Drives and Energy Systems (PEDES) & 2010 Power India, 2010 Joint
International Conference; 2010: 1-5.
[5] Vlad, I. Campeanu, A. Enache, S. Enache and M.A. Energy. Aspects and monitoring of asynchronous motors
starting.Optimization of Electrical and Electronic Equipment (OPTIM) 2014 International Conference, Bran; 2014:
306-31.
[6] Rusnok, S. Sobota, P. Mach, V. Kacor, and P. Misak. Possibilities of program EMTP - ATP to analyze the starting
current of induction motor in frequent switching.Electric Power Engineering (EPE), 16th
International Scientific
Conference, Koutynad Desnou; 2015: 614 – 619.
[7] Danang Wijaya, F. Kusumawan and S.A. Prabowo, H. Reducing induction motor starting current using magnetic
energy recovery switch (MERS).Information Technology and Electrical Engineering (ICITEE), 2014 6th
International Conference, Yogyakarta; 2014: 1 – 6.
[8] Piao Run-hao, Zhao Haisen, Zhang Dongdong and Li Jiaxuan. Analytical method for starting performance
calculation of induction motors considering skin effect and leakage flux saturation. Electrical Machines and
Systems (ICEMS), 17th
International Conference, Hangzhou; 2014: 135 – 138.
[9] Li Shue and Fue Chao. Design and Simulation of Three Phase AC Motor Soft Starter. Intelligent System Design
and Engineering Applications (ISDEA), 3rd
International Conference on Intelligent System Design and Engineering
Applications; 2013: 554-557.
[10] Li Kai, Chen Xing Lin and Tang Qiang. Dynamic Identification and Control of IM Soft- Start Using ANN.
Industrial Technology, 2008. ICIT 2008. IEEE International Conference; 2008: 1-6.
[11] Frede Blaabjerge, Johne K. Pederson, Soren Rise and Hans Henrik Hansen. A Comparative Study of Energy Saving
Benefits in Soft starters for Three Phase Induction Motors. Industry Applications Conference, 1995. Thirtieth IAS
Annual Meeting, IAS '95, Conference Record of the 1995 IEEE; 1995: 367-374.
[12] Syed Abdul Rahman Kashifa and Muhammad Asghar Saqiba. A Neuro Fuzzy Application: Soft Starting of
Induction Motors with Reduced Energy Losses. Electric Power Components and Systems. 2012; 40 (12): 1339-
1350.
[13] I. Ya. Braslavsky, A.V. Kostylev and D.P. Stepanuk.Energy consumption optimization during starting of thyristor
voltage converter induction motor system. International Symposium on Power Electronics, Electrical Drives,
Automation and Motion, SPEEDAM.2006: 11- 14.
[14] Sakuntala Mahapatra, Raju Daniel, Deep Narayan Dey and Santanu Kumar Nayak.Induction motor control using
PSO-ANFIS. International Conference on Computer, Communication and Convergence (ICCC 2015). 2015:
753–768.
[15] Xin She Yang. Natural Inspired Metaheuristic algorithm. Luniver Press, University of Cambrodge, United
Kingdom, second edition, 2010.
[16] Mahmoud M. Elkholy, Mohammed A. Elhameed. Braking of Three Phase Induction Motors by Controlling
Applied Voltage and Frequency Based on Particle Swarm Optimization Technique. International Journal of Power
Electronics and DriveSystem (IJPEDS). 2015; 5(4): 520-528.
[17] John Chiasson. Modelling and high performance control of electric machines. John Wiley & Sons, Inc., 2005
[18] Shi Y and Eberhart R. A modified particle swarm optimizer. Proceedings of IEEE International Conference on
Evolutionary Computation (ICEC’98), Anchorage, IEEE Press. 1998: 69–73.
[19] Shi Y and Eberhart R. Parameter selection in particle swarm optimization. Proceedings of the 1998 Annual
Conference on Evolutionary Programming, San Diego, MIT Press, 1998.
[20] Jang, J.S.R. ANFIS: Adaptive-Network-based Fuzzy Inference Systems. IEEE Transactions on Systems, Man and
Cybernetics. 1993; 23(3): 665-685.

 ISSN: 2088-8694
1048
BIOGRAPHIES OF AUTHORS
Mahmoud M. Elkholy received Bachelor of Engineering (B.E) degree (with honor) from Zagazig
University, Egypt in 1994 under the specialization of Electrical Machines and Power
Engineering, Master of Science degree from Zagazig University, Egypt 1998 under the
specialization of Electrical Machines and Doctor of Philosophy (Ph.d) in the year 2001 from
Zagazig University, Egypt in the Dept. of Electrical power and Machines Engineering. He has
18 years of experience in academia and research at different positions. Currently he is an
Assistant Professor, Faculty of Engineering, Zagazig University, Egypt. His interest includes
control the steady state and dynamic performance of electrical machines, artificial intelligence
and renewable energy
Mohamed A. Elhameed received the B.E. degree (with honors) from Zagazig University-faculty
of Engineering, Zagazig, Egypt in electrical power and machines engineering in 1996, Master
degree in 2000 in the field of electrical power system from the same institute, and the Ph. D.
degree from Zagazig University, Egypt, in 2004, in the field of electrical power system. He has
been assistant professor, Faculty of Engineering, Zagazig University, Egypt. His current interest
includes electrical machines modeling and control, artifitial intelligence, renewable energy,
optimization and FACTS devices.

Minimization of Starting Energy Loss of Three Phase Induction Motors Based on Particle Swarm Optimization and Neuro Fuzzy Network

More Related Content

What's hot (20)

Similar to Minimization of Starting Energy Loss of Three Phase Induction Motors Based on Particle Swarm Optimization and Neuro Fuzzy Network (20)

More from IJPEDS-IAES (20)

Recently uploaded (20)

Minimization of Starting Energy Loss of Three Phase Induction Motors Based on Particle Swarm Optimization and Neuro Fuzzy Network