Artificial Intelligence Controlling Chopper Operation of Four


Download Artificial Intelligence Controlling Chopper Operation of Four


Preview text

SAHARUL AROF et al: ARTIFICIAL INTELLIGENCE CONTROLLING CHOPPER OPERATION . .

Artificial Intelligence Controlling Chopper Operation of Four Quadrants Drive DC Chopper for Low Cost Electric Vehicle

Saharul Arof ¹,² [email protected] Muhd Khairulzaman A.K¹ [email protected]
Jalil, J.A¹ [email protected]

H. Arof 3 [email protected]
Mawby, P.A2 [email protected]

Abstract—DC drive system for traction is predominantly powered by separately excited as compared to series dc motor and both drive systems are controlled by four quadrants dc chopper. However, the operation modes of the conventional HBridge four quadrants drive dc chopper for series motor are limited to driving and regenerative braking (only with the presence of residual magnetism), with no capability of reverse operation, field weakening, resistive braking and parallel mode. As there are six chopper operations required for considerations, it is thus necessary to have controller that is able to choose the appropriate operation with respect to the chopper input signals. Hence, a new Four Quadrants dc drive chopper for series motor in EV’s application, optimized by Artificial Intelligence is developed. This paper further describes the application of Artificial Intelligence with Self Tuning Fuzzy Logic (STFL), Neural Network (NN) and Adaptive Neuro-Fuzzy Inference System (ANFIS) controllers for controlling chopper operation of the proposed chopper. Both the controller and EV system were developed using MATLAB/Simulink. The AI algorithms were tested and a comparative analysis has been performed on the three methods of control. Simulation results showed that all the controllers are able to select the expected operation for the proposed chopper with respect to the test signals given. It is also observed that the appropriate mode of operation guaranteed better EV performance in terms of battery power consumption, vehicle speed and distance traversed.
Keywords—DC drive, EV and HEV, series motor, four quadrant
chopper, Artificial Intelligence
I. INTRODUCTION
In future, the electric motor propulsion system (i.e. electric vehicle) will replace the internal combustion system (i.e. mechanical combustion engine) due to its higher efficiency and zero carbon emission [1]. Unfortunately, electric vehicles (EV) are not yet affordable for many people. Hence, the advancement of super capacitor as an alternative to batteries is the catalyst. At present, super capacitor (as big as 12,000F) is available in the market. The super capacitor offers several advantages such as lower weight, cheaper cost and fast charging. However, super capacitor can only operate at low voltage (the most is 2.7V) and it is not suitable for ac drive (which is always operated

at high voltage). Oakley Research National Laboratory; a US based
research center has currently producing lower than 5V operating voltage dc motor to suit super capacitor voltage for electric vehicle application. Direct Current (DC) drives which are series and separately excited have once been used as prototypes or products for EVs and HEVs. Examples of such systems are those installed in Peugeot 106, Citroen Saxo, GM EV and Lada [7,8]. In those days, DC motors had disadvantages in size, weight, performance and reliability. A recent study undertaken at Oak Ridge National Laboratory [4] reported that DC motors are suitable for EV/HEV applications. Such motors shown in figure 1 possess high power output, higher efficiency, smaller size, less weight, long lasting carbon brush and commutator, low operating voltage (less than 50V) and use modular structure i.e. easy to replace parts like carbon brush, etc. [8,10,11,12]. Advanced brush technology used in latest DC motors allows the motor to be operated at low voltages [11,12] resulting in lower losses. Nowadays, the lifespan of the carbon brush and commutator of a DC motor is longer than the rotor bearing of an AC induction motor [9,11]. The brush can last until 30,000km while the commutator can endure 250,000 km before it flushes over [9,12]. The reliability of the softcommutated DC motor is now around 90% - 94% [9,10,11,12] while the spark reduction system can further extend the brush life [12]. The total cost of a drive system (AC or DC) for an EV depends on several factors such as the number of gate drive circuits, power electronics devices used, its cooling system, heat sinks, converter, batteries, motor and auxiliary devices. A DC drive system’s lower cost is mainly due to its simple circuit, controller design and control strategies [8,9,10]. DC choppers were introduced in the early nineteen sixty using force-commutated thyristor. DC choppers [2] are mainly used to drive dc motors while offering the capability of bidirectional power flow for both motoring and regenerative braking. For EV application using series motor, the common Half-bridge DC chopper offers no capability of regenerative braking, field weakening

DOI 10.5013/IJSSST.a.16.04.03

3.1

ISSN:1473-804x online,1473-8031 print

SAHARUL AROF et al: ARTIFICIAL INTELLIGENCE CONTROLLING CHOPPER OPERATION . .

and resistive braking operations. In this paper, a novel Four Quadrant DC chopper
topology is also introduced. The proposed chopper design is able to reduce power loss and generates fewer ripples. The proposed chopper is aimed to solve common problem on the speed control of series motor in which the speed decreases linearly when loaded.
The proposed dc chopper design is further optimized by utilizing the Artificial Intelligence operators such as Self Tuning Fuzzy Logic, Neural Network and Adaptive Neuro Fuzzy Inference System controllers in controlling the chopper operation. A comparative analysis has been performed on the three methods of control.
II. FOUR QUADRANTS DC DRIVE CHOPPER
A. Proposed Chopper Design
The proposed four quadrants chopper design shown in figure 2 consists of three IGBTs used as the main switches, field weakening and bridging IGBTs. LM is the motor inductance connected in series with the armature windings to smoothen the armature current of the motor and RBV is the brake series resistor.
The chopper has seven modes of operation that are driving, reversing, field weakening, parallel mode, regenerative braking, resistive braking, and generator. When the motor starts to rotate, driving mode is selected. As the motor speed increases above its nominal speed, field weakening mode is selected. In generator mode, the motor behaves as generator and generate power to charge the battery. A regenerative braking mode is selected to slower down the EV while the motor runs above its nominal speed, hence charging the battery at the same instant. In contrary, resistive braking mode is selected to slower down the EV while the motor runs below its nominal speed, as well as charging the battery at the same instant. If the motor tends to run with incrementing speed, i.e. EV is driven on steep hill; the parallel mode is selected to prevent drastic speed drops. The chopper is controlled by four sub controllers; i.e. for data distribution, chopper operation, subsequent and delay, and IGBTs firing.
B. Chopper Operation Controller
The chopper operation controller is designed to choose the most appropriate operation by processing signals obtained from the accelerator pedal, brake pedal, battery voltage, motor speed, and etc. Upon receiving these signals, the controller will generate other signals such as error and rate of speed.
Failing to pick the right most operation could cause chopper failure. If the load is too heavy due to passengers’ weight or due to climbing steep hill, the field weakening operation should be avoided.

SPEED & ENERGY

Figure 1. High power low voltage dc motor from ORNL

K1

K6 D3



Rd

V3 Rbv

1K2

2K3

K5

M/G

2K3

1K2

R1

D2

Lm

CN

D4

Lf K4

Rfw V2

K7 D1

V1

Figure 2. Proposed Four Quadrants Drive DC Chopper

1.2

FW SIGNAL

S1

1

SPEED

0.8

S2

E2

0.6

0.4

ENERGY

E1 0.2

0

0

500 1000 1500 2000 2500

Time (x 0.5ms)

Figure 3. Field weakening mode failure

DOI 10.5013/IJSSST.a.16.04.03

3.2

ISSN:1473-804x online,1473-8031 print

SAHARUL AROF et al: ARTIFICIAL INTELLIGENCE CONTROLLING CHOPPER OPERATION . .

Input signals from accelerator and brake
pedal

Parameters Calculations & Loop Up table

Figure 4. Chopper Operation Process Flow
Figure 3 represents a condition of field weakening failure as a result of choosing the incorrect chopper operation mode. S1 (normalized) represents the motor speed during high load without field weakening action. In contrary, S2 represents the same condition with the field weakening action. E1 and E2 (normalized) represents the battery energy used to run the motor. When the field weakening mode takes action, the speed of the motor is observed to drop by 10% while theconsumption of the battery power increases for almost 200%. Hence, an incorrect selection and control of the chopper operation leads to a drop in the EV performance.
Several methods can be used to select the most appropriate chopper operation such as the Expert System IF-THEN rule, Fuzzy Logic, Neural Network (NN), Self Tuning Fuzzy Logic (STFL), and ANFIS. The chopper operation process flow is simplified in Figure 4. The Expert System IF-THEN rule is the common selection method used. However, a drawback of this method is that many rules have to be written depending on each type of chopper operations and input signals. Due to that, additional sensors such as car position sensors and weight sensors are needed in order to ease the control algorithm and to differentiate the chopper operation selection mode. This method requires lots of memory and effort to write the code. Hence, the AI control algorithms such as STLF is proposed as the selection method for the chopper operation and due to its linearity, the selection or decision making problems would be simplified.
III. CONTROL STRATEGY
AI operators such as Fuzzy Logic (FL), Neural Network (NN) and ANFIS have been widely used in dealing with nonlinear and complex system such as pattern recognition, forecasting, decision making, etc. Self Tuning Fuzzy Logic; an adaptive to conventional Fuzzy Logic has excellently been used as decision maker/operation selection for application such as in the elevator systems [12-13]. Figure 5 shows the process of determining the most suitable chopper operation for EV utilizing AI controller.

Four Quadrants
DC Chopper

System Feedback

STFLC/NN/ANFIS (AI Controller)
Performance Index Evaluation

Determine max and min values
Figure 5. Chopper Operation Block Diagram

Knowledge base

Data (Mf)base

If Then Rule b

SelfTuning
Module

Crisp Input

Fuzzification Module

Fuzzy Inference Engine

Defuzzification Module

Process

Crisp Output

Figure 6. Basic Concept of Fuzzy Logic
IV. SELF TUNING FUZZY LOGIC AS CHOPPER OPERATION
CONTROLLER
The basic concept of Fuzzy Logic is based on the concept of a crisp input and crisp output. Crisp means the actual data or parameter being used, described either in quantitative or qualitative parameters [14]. Between the crisp input and crisp output, all the processes are based on ‘fuzzy’ parameters converted at the beginning of the process. The full architecture is shown in Figure 6.
Aimed to simplify the computation, triangular shaped function is used to determine membership function. Five fuzzy variables are defined for each parameter. The use of five fuzzy variables is large enough to provide adequate approximation, save memory storage and reduce complexity in computation.

DOI 10.5013/IJSSST.a.16.04.03

3.3

ISSN:1473-804x online,1473-8031 print

SAHARUL AROF et al: ARTIFICIAL INTELLIGENCE CONTROLLING CHOPPER OPERATION . .

A. Self Tuning Mechanism
In the self- tuning fuzzy logic controller, the membership functions are modified or adjusted according to the error and rate of speed of the vehicle. Error is the value of speed compared to accelerator pedal, while rate of speed is the difference between current speed and the previous speed. STFL controller has seven sets of Fuzzy Logic operators which represents seven modes of chopper operations. When an accelerator pedal or brake command is registered, relevant data on the chopper and EV are needed for computation of input membership function. The signal inputs are then assessed to each of fuzzy operator which represents every single chopper operation mode. Every single fuzzy operator produce output, i.e. Performance Index (PI) which is later compared to each other. The operation with the highest PI is considered the most suitable operation to be picked as the current chopper operation.

B. Adjusment of Membership Functions

Self tuning of the membership functions is done by reshaping the triangles of the fuzzy variables. Figure 7, 8, and 9 briefly describe how the membership functions are tuned. MC refers to the present M, and MN is the new M. Figure 7 is conditioned at medium load/weight. Figure 8 is conditioned at heavy load, i.e. when the vehicle is climbing steep hill. Figure 9 is conditioned at lighter load, i.e. when the vehicle is moving downhill. Five triangular-shaped membership functions were shown and five fuzzy variables were defined as Zero(Z), Small (S), Medium(M), Large (L),

and Very Large (VL) or Negative Big (NB), Negative small

(NS), Zero (Z), Positive Small (PS), and Positive Big (PB). Tuning is carried out basically by varying the upper, and the lower, bounds of the triangles. Five bounds were identified, and labelled asU1, U2, U3, U4, andU5, equally separated from each other by a length l along the universe of discourse. By changing l, and the bounds, the membership functions are adjusted accordingly.
Self tuning fuzzy logic controller analyzes the EV conditions to derive specific values used to shape the membership functions according to error and rate of speed. The middle boundU3(also labelled MC), is first defined for lighter, medium, and heavy load driving. The respective value is stored in a look up table and the value varied depending on the driving condition. Figure 10 shows how to obtain a new M and l values in order to initiate a correction.
Adjustment of length l is carried out for the five input parameters. The l value may also be computed by using equation 1. If (maximum – average) is bigger than (average – minimum) then:

, other wise

(1)

C. Selection of Fuzzy Rules
The fuzzy rules are prepared in the following form: If A and B and C, then D. Rules are prepared just for its specific

driving condition (Driving, Field Weakening, Generator, Regenerative Braking, Resistive Braking, Parallel and Reverse). For a Driving mode to be selected, the rules are written with respect to conditions that make driving mode will be possibly selected such as in Table 1. As mentioned before the real traffic situation is a mix of various driving pattern and very much dependent on accelerator pedal, braking pedal, motor speed, rate of speed, error and total weight of the vehicle. With the help of STFL, it is not necessary to have specific rules for lighter, medium, or heavy load conditions.

V. NEURAL NETWORK AS CHOPPER OPERATION
CONTROLLER

According to McCulloch and Pitts, neuron consists of input,

weights, neuron, and output as shown in figure 11 [14].The

neuron computes the weighted sum of the input signals and

compares the result with a threshold value, θ. If the net input

is less than the threshold, the neuron output is -1. However,

if the net input is greater than or equal to the threshold, the

neuron becomes activated and its output attains a value +1.

The neuron uses the activation function as described by

equation 2. The actual output of the neuron with a sign

activation function is further described by equation 3.



(2)

where

1



1





∅0

(3)

According to Rosenblatt, perceptron is the simplest form

of Artificial Neural Network (ANN) which consists of a

single neuron with adjustable synaptic weights and a hard

limiter. A single layer multi-input perceptron is shown in

Figure 12. The weighted sum of the input is applied to the

threshold which produced an output equal to +1 if its input

is positive, and -1 if it is negative [14].

The learning process is done by making some

adjustments in the weights to reduce the difference between

the actual and desired outputs of the perceptron. The initial

weights are randomly assigned and then updated to obtain

the output consistent with the training. The updating process

can be obtained from equation 4 [14].

,

1,2,3 … .. (4)

where P is the iteration. The perceptron output,Y(p) is increased if the error,e(p)is positive and Y(p) is decreased if the errror is negative. Each perceptron input contributes xi(p) x wi(p) to the total input X(p). If the input value xi(p) is positive, an increase in its weight wi(p) tends to increase perceptron output Y(p). If xi(p) is negative, an increase in wi(p) tends to decrease Y(p). This contributes to perceptron learning rule as described by equation 5[14].

DOI 10.5013/IJSSST.a.16.04.03

3.4

ISSN:1473-804x online,1473-8031 print

SAHARUL AROF et al: ARTIFICIAL INTELLIGENCE CONTROLLING CHOPPER OPERATION . .

Degree of Membership

Z 1

S

M

L

VL

0

Speed

S1

S2 S3 S4 S5

S1- S5 is adjustable, not fixed

Figure 7. Self Tuning Membership Function conditioned at medium load

Degree of Membership

VS

S

M

L

VL

1

l

0 U1

Universe of Discourse

U2

U3 U4

U5

Figure 8. Self Tuning Membership Function conditioned at heavy load

Degree of Membership

VS S M L VL 1
l

Universe

0

of Discourse

U1 U2 U3 U4 U5

Figure 9: Self Tuning Membership Function conditioned at lighter load

DOI 10.5013/IJSSST.a.16.04.03

3.5

ISSN:1473-804x online,1473-8031 print

SAHARUL AROF et al: ARTIFICIAL INTELLIGENCE CONTROLLING CHOPPER OPERATION . .

START
Determine error and rate of speed of vehicle
Determine the average value according to look up table
Determine max and min values

(ave –min) (max – ave)

l = (max – ave)/2

l = (ave – min)/2

Operation Mode Driving Field weakening Regen brake Rheostat brake Generator Parallel

END

Figure 10: Self tuning Algorithm TABLE 1. Signals classified for chopper mode selection rules in general

Acc pedal Bigger than zero Medium to high NIL NIL Low to medium Medium to high

Brake signal Zero Zero Medium to high Zero to medium HIGH Zero Zero

Speed Zero to medium Medium to high Medium to high Zero to High Medium to high Medium to high

Rate of speed NIL Zero to positive Zero to negative Zero to negative Zero to negative Negative

Error positive positive NIL NIL negative positive

Figure 11. Neural Network Neuron

Figure 12. Neural Network Perceptron

DOI 10.5013/IJSSST.a.16.04.03

3.6

ISSN:1473-804x online,1473-8031 print

SAHARUL AROF et al: ARTIFICIAL INTELLIGENCE CONTROLLING CHOPPER OPERATION . .

1

(5)

where is the learning rate. Perceptron has limitation,

hence multilayer neural networks (NN) is an alternative.

Multilayer perceptron is a feedforward neural network with

one or more hidden layers as shown in Figure 13. Generally,

it consists of an input layer of source neurons, a minimum

of one hidden layer of computational neurons, and an output

layer of computational neurons. The input layer accepts

input signals and redistribute to all neurons in the hidden

layers,while the output layer accepted output signals from

the hidden layer and forms the output patterns. The weight

of the neurons represents the features hidden in the input

patterns. This multilayer feedforward NN is trained with the

back- propagation algorithm such as the LM technique [14].

Hence, the NN controller shown in Figure 13 is used as the

chopper operation controller. The NN controller

receivesfive input signals and producessix output signals

which represents the six modesof chopper operation. Each

output is transformed and represented by the Perfomance

Index (PI).

The PI will then be compared to each other. The highest

PI will be selected as the current chopper mode of operation.

In order to teach NN controller to work as the chopper

operation controller, a set of input and output data must be

fed to the NN controller for training. This is called

supervised learning [14]. Thus, the input and output data for

teaching NN controller is recorded from STFL controller.

Input data is the same data being fed to STFL controller,

while output data is the expected chopper operation mode

represented in coded Binary data. The number of iterations

and hidden layers were set during training and adjusted

accordingly in order to get the best performance. The

process of teaching the NN controller is represented in a

process flow block diagram as shown in Figure 14.

VI. ANFIS AS CHOPPER OPERATION CONTROLLER
Jang’s ANFIS six-layer feedforward NN is also used and tested as the chopper operation controller. Figure 15 represents the ANFIS architecture corresponds to the first order Sugeno fuzzy model with two Inputs and one output [14].
The input layer which is represented by equation 6. The neuron then passes the external crisp to layer 1.
(6) Layer 1 is the fuzzification layer. Neuron performs fuzzification using bell activation function as represented by equation 7.
(7)

Figure13. Multilayer Neural Network START
Load input and output data from STFLC
Set number of neuron in hidden layers
Set number of teaching iterations NO
Performance acceptable
YES STOP Figure 14. Neural Network training process flow diagram
Figure 15. ANFIS Architecture

DOI 10.5013/IJSSST.a.16.04.03

3.7

ISSN:1473-804x online,1473-8031 print

SAHARUL AROF et al: ARTIFICIAL INTELLIGENCE CONTROLLING CHOPPER OPERATION . .

START
Load input and output data from STFLC and NN

Set numbers of MF and EPOCH

NO
Performance acceptable

YES STOP

Figure 16. ANFIS training flow diagram

where is the input and is the output of neuron i in Layer 1. ai, bi and ci are parameters that control the center, width and slope of the bell activation function [14]. Layer 2 is the rule layer corresponds to fuzzy Sugeno-type rule. Rule antecedents are evaluated by the operator product in ANFIS [14].



(8)

is the input and is the output of rule neuron i. Layer 3 is the normalization layer.





̅

(9)

Layer 4 is the defuzzification layer.

1

2

̅

1

2

(10)

Layer 5 is the summation of neuron.



∑̅

1

2 (11)

ANFIS learns these parameters and tunes membership functions by itself, thus no prior knowledge on the rules is required. ANFIS uses a hybrid learning algorithm that combines the least square estimator and the gradient descent method [14]. ANFIS takes five input signals and one output signal for the chopper operation controller. The PI is replaced with IF-THEN rules output range selector. ANFIS controller output data is compared to a look up table that specifies the range of value representing the chopper operation mode. The input and output data for training ANFIS controller is recorded from STFL and NN controller, however the input data are preprocessed and normalized while the output data is transformed into six values between 0 and 1represents chopper mode of operation. The process of training ANFIS controller to act as chopper operation controller can be represented in a process flow block diagram as shown in Figure 16.

VII. SIMULATION AND ANALYSIS
The performance of the proposed fuzzy-based strategy applying STFL controller can be evaluated on the real hardware system or simulated systems. A real hardware system is difficult to implement, time consuming and expensive. In contrary, a simulated system is easily developed, time-saving and cost effective. It provides a convenient platform to test the algorithms and to simulate the process of chopper and EV operations, without causing too many risks in terms of the costs spent and losses brought to the system in case of algorithm failure. Furthermore, various parameters can be simply adjusted and optimized, hence speeding up the simulation process towards achieving the desired results.
In this work, a computer simulation is performed as a method to test the AI algorithms and to study the controller performance and effectiveness. For this purpose, MATLAB/Simulink is used to develop the controller and the system. The simulation model used to study the chopper behavior is shown in Figure 17. Some parameter values required by the simulation software to simulate the system are provided by the user.
The car is tested according to the accelerator signal (signal 9), brake signal (signal 7) and also according to earth profile(signal 8). Each of the test signal is shown in Figure 18, 19, and 20. The performance of STFL, NN and ANFIS as chopper operation controller is also depicted in Figure 18, 19 and 20 respectively. Simulation results showed that all the three controllers are all able to select the expected operation for the Four Quadrant DC chopper with respect to the test signals given.
It is observed that all the controllers selected driving mode during start up. As the vehicle speed increasesand due to accelerator demand, STFL controlleris the first to select the field weakening mode followed by ANFIS and NN.However, when the accelerator signal is let low, the generator mode are selected by all the controllers. When the accelerator signal is back to maximum and while the speed is still high, the field weakening mode is again selected by all the three controllers.
When brake command is activated at high vehicle speed and due to vehicle inertia, the regenerative mode should be operated.As the vehicle speed drops further, the resistive braking mode is expected. However, NN controller tends to extend the regenerative braking mode much further hence no resistive braking mode can be observed. When brake command is released and replaced with low driving command, and while the vehicle speed is low STFL controller selected the driving mode while both NN and ANFIS controllers selected the field weakening mode. As the vehicle moves downhill while the driving command is low,all the controllers are expected to select generator mode. STFL controller is the last toselect the generator modeas compared to NN and ANFIS controllers.

DOI 10.5013/IJSSST.a.16.04.03

3.8

ISSN:1473-804x online,1473-8031 print

SAHARUL AROF et al: ARTIFICIAL INTELLIGENCE CONTROLLING CHOPPER OPERATION . .

If the driving command is set high again, the controller should then select the field weakening mode.It is also observed that the three controllers succesfully chose the expected mode. As the vehicle drives ona steep hill, the speed is expected to drop. Hence, the parallel mode is expected to be selected. Parallel mode is capable to overcome the climbing steep hill effect (motor speed drops drastically), however this capability consumes more battery power. It is also observed that STFL controller is the first to activate Parallel mode followed by ANFIS and later by NN. Finally, as the vehicle regained its speed the field weakening mode is expected to be selected.ANFIS controller is observed to be the first converted to field weakening mode followed by STFL controller.
Figure 21, 22 and 23 show the results of STFL with respect to battery state of charge (SOC), vehicle speed and distance traversed as compared to NN and ANFIS chopper operation controllers. It is observed that performance of STFL controller the least as it consumes more battery power. The maximum speed of the EV controlled by STFL controller is the lowest as compared to the other two AI operators. The distance traversed of EV powered by the STFL controller is the shortest among the three controlling methods.
VIII. CONCLUSION
A new Four Quadrants dc drive chopper for series motor in EV’s application optimized by Artificial Intelligence is developed in MATLAB/Simulink. Results showed that dc drive series motor has potential to be utilized in EV with the proposed chopper design. Hence, Artificial Intelligence with STFL controller, NN and ANFIS are able to be used as the Four Quadrant Chopper Operation Controller due to their capability to select the most appropriate chopper operation. Comparatively, ANFIS controller works better for EV performance i.e. higher vehicle speed and longer distance traversed, while NN controller is suitable for saving battery power. In contrary, results showed that STFL controller has poor EV performance i.e. lowest vehicle speed and shortest distance traversed as well as highest battery consumption. Thus, STFL controller requires further tuning in order to achieve better EV performance. Conclusively, the EV performance can further be optimized with an appropriate tuning of the controller.
REFERENCES
[1] YiminGao and MehrdadEhsani, “Design and Control Methodology of Plug-inHybrid Electric Vehicles” IEEETrans. Ind. Electron, VOL. 57, NO. 2, pp 633-640, Feb 2010.
[2] Michael H. Westbrook, The Electric and Hybrid Electric Vehicle, SAE, 2001.
[3] Iqbal Husain, Electric and Hybrid Electric Vehicles, Design fundamentals, CRC Press, 2003.

Accel3

Acc_ped
Energy Management Subsystem

Accelerator Accel2

Accel4

Wmot

DC CCHOPPER1 TL

DC CCHOPPER

RefDriveTorque m otor_torque

CarSpeed

Accelerator Car speed (km/h) Driv e torque (ref erence, measured)

Power

Accelerator Car speed (km/h)

Driv e torque (ref erence, measured)

Power (Motor, Battery )

Car Motor speed (rad/s)

Vehicle Dynamics

TL distance trav ersed CarSpeed

Accel1 Input
NN

Fuzzy

soc [eff]
energy
V I

Scope2

Electric Car (EC) Power Train
Figure 17. MATLAB/Simulink model

Speed & Chopper Signals

SELF TUNING FUZZY LOGIC CONTROLLER

10 8

9

7

5

SPEED

01
-5 0

3

2

45

6

500

1000

Time(x 10ms)

Figure 18. STFLC Chopper operation Result

NEURAL NETWORK CONTROLLER

10 8

9

7

5

SPEED

Speed & Chopper Signals

0

1

2

3

45

-5

6

0

500

1000

Time(ms)

Figure 19. NN Controller

[4] Oak Ridge National Laboratory “Advanced Brush Technology for DCMotors” 2009. Available:
http://peemrc.ornl.gov/projects/emdc3.jpg
[5] Oak Ridge National Laboratory “Soft-Commutated Direct Current (DC) Motor”2009. Available: www.ornl.gov/etd/peemrc

DOI 10.5013/IJSSST.a.16.04.03

3.9

ISSN:1473-804x online,1473-8031 print

SPEED & CHOPPER SIGNALS

SAHARUL AROF et al: ARTIFICIAL INTELLIGENCE CONTROLLING CHOPPER OPERATION . .

10 8
7
5

ANFIS CONTROLLER 9
SPEED

0

1

2

-5 0

3 4 5

6

500

1000

TIME (x10ms)

Figure 20. ANFIS

Legend: 1. Driving 3. Generator 5. Resistive Braking 7. Brake Signal 9. Accelerator Signal

2. Field Weakening 4. Regenerative Braking 6. Parallel Mode 8. Earth Profile

100 99 98 97 960
60 50 40 30 20 10
00

ANFIS

CONTROLLERS

SELF TUNING FUZZY

NEURAL NETWORK

500

1000

TIME(ms)

1500

Figure 21. Battery State of Charge

CONTROLLERS

ANFIS SELF TUNING FUZZY

NEURAL NETWORK

200

400

600

800

1000

1200

1400

TIME(ms)

Figure 22. Vehicle Speed

DISTANCE TRAVERSE(km)

0.14 0.12
0.1 0.08 0.06 0.04 0.02
0 0

CONTROLLERS

ANFIS

NEURAL NETWORK

SELF TUNING FUZZY

200

400

600

800

1000

1200

1400

TIME(ms)

Figure 23. Distance Traverse

[6] Heinrich, Walter Rentsch, Herbert Dr.-Ing, ABB Industry: “Guide to Variable Speed Drives,” Technical Guide No. 41180 D-68619 LAMPERTHEIM, Germany, 3ADW 000 059 R0201 REV B (02.01), DCS 400 / DCS 500 / DCS 600: ABB 2003.
[7] Ramesh C.Bansal, Birla Institute of Technology and Science, Pilani, India. Electric Vehicle”, Handbook Of Automotive Power Electronics and Motor Drives, Taylor &Francise group CRC Press, 2005.
[8] M.H.Rashid, “Power Electronics, circuits, devices and applications”, 3rd ed.: Prentice Hall, 2004.
[9] T.W. Ching, “Soft-switching Converters for EV Propulsion”, Journal of Asian Electric Vehicles, vol 5, no 2, December 2007.
[10] N. Senthil Kumar, V. Sasasivam, M. Muruganandam, “A Low Cost Four-quadrant Chopper – Fed embedded DC Drive Using Fuzzy Controller,” Journal of Electric Power Components and System, vol 35, August 2007, pp. 907 – 920.
[11] Dheeraj Joshi R.C. Bansal, “Performance Evaluation ofMultiquadrant DC Drive Using Fuzzy-Genetic Approach,” Journal of Electrical System (JES) 2009,vol 5, no 4, December 2009.
[12] Jafferi Jamaludin, Nasrudin Abd. Rahim, “An Elevator control group system With a Self tunig Fuzzy Logic Group Controller”, IEEETrans. Ind. Electron, VOL. 57, NO. 12, DECEMBER 2010.
[13] Jafferi Jamaludin, Nasrudin Abd. Rahim “Development of self tuning fuzzy logic controller for intelligent control of elevator system”, Engineering Applications of Artificial Intelligence,Volume 22, Issue 8, December 2009, Elsevier.
[14] Negnevitsky, M. 2005. Artificial intelligence: A guide to intelligent systems. Essex, UK: Addison-Wesley.

BATTERY STATE OF CHARGE (%)

SPEED (km/h)

DOI 10.5013/IJSSST.a.16.04.03

3.10

ISSN:1473-804x online,1473-8031 print

Preparing to load PDF file. please wait...

0 of 0
100%
Artificial Intelligence Controlling Chopper Operation of Four