PROSIDING SEMINAR NASIONAL RITEKTRA 2011 ISBN: 978-602-97094-2-1

System Improvement and Fuzzy Logic Controller Design to Extend Battery Life-Time of Electric Bike System Ferdian Adi Pratama, Prianggada Indra Tanaya & Maralo Sinaga Department of Mechatronics, Faculty of Engineering Swiss German University, EduTown BSDCity Tangerang 15339, INDONESIA. Email: [email protected]

[email protected] [email protected] Abstrak Sepeda listrik (e-bike) merupakan alat transportasi perkotaan yang murah enerji. Pengembangan prototipe e-bike berbasis micro-controller telah dilakukan secara bertahap. Penambahan Logika Fuzzy pada sistem kontrol ini bertujuan untuk meningkatkan umur pemakaian baterai disamping juga memperbaiki sistem kontrol micro yang telah diterapkan. Beberapa rancangan logika fuzzy disimulasikan untuk mendapatkan arsitektur yang sesuai. Rancangan logika fuzzy yang dihasilkan diterapkan pada e-bike dan dievaluasi kinerjanya berdasarkan kriteria waktu dan jarak tempuh. Kata kunci: logika fuzzy, simulasi, umur baterai, e-bike, micro-controller, kinerja sistem.

1. INTRODUCTION Due to the rapid advancing technology nowadays and our dependency on fuel, the demand for smarter, adaptive and more efficient vehicles is increasing. An Electric Bike has been designed and constructed in SGU (Swiss German University). There have been three preceding developments of this SGU electric bicycle. During the first development [2] in November 2006, the electric bike was built using an ordinary bike and using microcontroller ATMEGA8 as the controller. Next, the second developer [3] made some major improvements in braking system and motor controller in July 2007. During the third development [4], the electric bike was improved with regenerative braking and battery charging in January 2008. The application of fuzzy controller in electric vehicle has proved that the fuzzy logic controller is sturdy [5]. A fuzzy logic controller has been applied in electric bike using PSoC (Programmable System-onChips) as the microprocessor [6]. Also, a fuzzy controller has been developed using

microcontroller, to optimize gears shifting [7].* By applying a method of Artificial Intelligence (A.I), which is Fuzzy Logic to Electric Vehicle [1], this article reports the 4th development, which applying fuzzy logic to increase the battery efficiency of the Electric Vehicle. Further developments in mechanical and electrical aspects of the E-bike are needed to improve the stability, security and efficiency. 2. EXPERIMENTAL PROCEDURE 2.1 Fuzzy Logic Fuzzy logic is a way of thinking, applied as an Intelligent System using linguistic terms, which offers great flexibility for different types of inputs, such as linear and non-linear inputs. Fuzzy logic is the derivative of Boolean logic, which allows us to assign value in the 'grey area' between 1 and 0. Fuzzy logic is able to handle vagueness and imprecision, since it does not require precision, noise free inputs, and can be programmed fail safe if a feedback sensor quits or destroyed, which makes it robust as a 76

PROSIDING SEMINAR NASIONAL RITEKTRA 2011 ISBN: 978-602-97094-2-1

controller. Linguistic values are defined from membership functions, which allow us to specify the degree of the inputs. Exhaustive IF-THEN-ELSE pattern are used to formulate the inputs membership functions of a particular system. And then, the inputs were processed by Inference Engine, resulting crisp output, which is the output value that we can use. 2.2 Structure of Fuzzy Logic The input will undergo 3 processes in the fuzzy controller:
Fuzzification
Inference
Defuzzification There are 2 methods of Inference process available: Mamdani & Sugeno mehod. Mamdani method requires us to calculate the Center Of Gravity (COG) of the membership function. Instead of using COG technique, Sugeno method use a single spike, a singleton, as the membership function of the rule consequent, which requires us to calculate the Weighted Average of the membership functions. Therefore, the result of using Mamdani method is more accurate compared with Sugeno method. In this paper, Sugeno method is used to determine the PWM value. It is sufficient for micro-controller based system, the calculation is simpler, and processing time will be smaller. 2.3 Limitations and Assumptions There are 2 methods of Inference process available: Mamdani & Sugeno mehod. Mamdani method requires us to calculate The E-bike is limited in a few things, which matters during the test drives. First, the regenerative braking is going to be deactivated during this thesis work, due to the following reasons: • the electrical braking is already off initially due to short circuit • during the simulation, the throttle

will be constantly pulled and no significant braking will be done to test the E-bike Second, the load sensor, which is to determine the weight passenger automatically, was not integrated in the Ebike, because of difficulty of integrating the sensor in the bike. So the passenger weight must be manually inputted and modified through the program. Third, during the coding of the fuzzy logic in the C compiler, there will be no preprocessor used. Preprocessor is a header files included in the C code, to ease the coding -in this case Fuzzification, Inferencing, and Defuzzification-. There are a few well-made fuzzy logic preprocessor available in the Internet with certain prices. Nevertheless in this thesis work, the fuzzy logic will be coded from scratch, yet has lower complexity compared with the one coded with preprocessor and still easy to understand. The motor driver of this E-bike is going to use Pulse Width Modulator (PWM) as the motor driver. Assumptions have also been made for simplicity of the test drives. First, it was assumed that during test drives in normal mode, the throttle would be opened 100% (full throttle). Second, it was assumed that during test drives in normal mode, when the battery reach 4 bars, it will be the end of the test drives, because when the battery reach 4 bars, it will be in the range within 32V until 33V, which is unable to run the motor at proper speed, especially when loaded. Third, to ease the debugging, the LCD will be programmed to show the battery life- time, ADC value of the throttle, and the ADC PWM value which result from the fuzzy controller calculation. 2.4 Implementation The motor driver of this e-bike is still using Pulse Width Modulator (PWM), which is the same with the previous e-bike 77

PROSIDING SEMINAR NASIONAL RITEKTRA 2011 ISBN: 978-602-97094-2-1

design. The Fuzzy controller will be designed and simulated using MATLAB(R), will be programmed in C language with CodeVisionAVR(R) as the compiler.

configurations are listed in table 1 and table 2 below.

3. FUZZY LOGIC DESIGN CONSIDERATIONS [1] 3.1 Triangle Configurations Inputs and output parameters were considered first. Two input parameters were used, which are battery voltages and passenger weight. PWM will be the output for the Fuzzy Controller. The goal of this thesis were considered as the fuzzy rules were constructed, which is to make the battery life as long as possible, but also still possible to run the e-bike as efficient as possible. persamaan ditulis dalam kurung dan ditempatkan rata kanan (lihat contoh). Figure 1 shows the fuzzy controller main diagram, containing 2 inputs, passenger weight and battery voltages, and will resulting 1 output, PWM.

Figure 2: Membership function of Passenger Weigth for design 1 and 2 Table 1 Configurations of Fuzzy design 1 & 2 Design 1-2 Name Type Params

Light trapmf 40 40 46 52

Name Type Params

Low trimf 25 25 32 34

Weight Medium trimf 46 55 64 Battery Medium trapmf 32 34 36

Heavy trapmf 55 67 80 80 High trimf 35 37.5 40 40

Table 2 Configurations of Fuzzy design 3 Design 3 Name Type Params

Light trimf 34 40 55

Name Type Params

Low trimf 25 30 33

Weight Medium trimf 46 55 61 Battery Medium trapmf 31.5 33.5 35 36.5

Heavy trapmf 58 61 80 80 High trimf 34 38 40 40

All of membership functions above are used in MATLAB®, trimf represent the triangle membership functions, and trapmf represent the trapezoidal membership functions. Figure 1: Fuzzy Controller Main Diagram

The linguistic variables for the passenger weight are divided in 3 categories, light, medium, and heavy. Trapezoidal function was chosen to represent the fuzzy characteristic for all the weight categories. In this work, 3 (three) different Fuzzy configurations of design will be presented. Design 1 & 3 will have the same rules but different triangle configurations, and Design 1 & 2 will have the same triangle configurations but different rules. The light membership function is shown in figure 2 in red line. The other designed membership functions

3.2 User-Defined Rules The user-defined rules for Fuzzy design 1 & 3 are listed in table 3 below, and rules for Fuzzy design 2 is listed in table 4. Table 3 Rules of Fuzzy design 1 & 3 No. 1 2 3 4

1st & 3rd Fuzzy Design Rules IF Weight Operator Battery Light Medium Heavy Heavy

OR AND AND

Low Medium High

THEN PWM Slow Medium Fast Fast

78

PROSIDING SEMINAR NASIONAL RITEKTRA 2011 ISBN: 978-602-97094-2-1

Table 4 Rules of Fuzzy design 2 No. 1 2 3 4 5 6 7 8 9

IF Weight Light Medium Heavy Light Medium Heavy Light Medium Heavy

2nd Fuzzy Design Rules Operator Battery AND AND AND AND AND AND AND AND AND

Low Low Low Medium Medium Medium High High High

THEN PWM Slow Slow Slow Slow Medium Fast Medium Fast Fast

3.3 Simulation Result

Using MATLAB®, all design of fuzzy logic is simulated (figure 3,4, and 5). Those figures show the simulation, where two inputs; battery condition (V) and passenger weight (kg), and the output PWM signal in term of pulse. 4. TEST DRIVE RESULT [1] Since the bicycle is developed using several modes, such as NORMAL (fuzzy disable) and SMART modes (include 3 fuzzy logic design), the test drive will compare 4 graph results between the distance traveled and the battery voltage level. 4.1 Fuzzy Logic Disable

Figure 3: 3-D I/O Relationship of Design 1

Initial test drive was conducted to analyze the initial battery performance. The E-bike was tested with 60 kg passenger. The E-bike was using three 12V Lead-Acid batteries as the power supply. The test drives started at battery around 38.5V and stopped at battery around 32V-33V, because at around 32V the E-bike runs very slow. The 0,7% error measurement from multimeter is also taken into account. The linearized equation has been added in the graph.

Figure 4: 3-D I/O Relationship of Design 2

Figure 6: Fuzzy Logic Disable (NORMAL mode)

Figure 6 shows the batteries performance without fuzzy logic. The Ebike is able to travel at maximum distance approx. 36km in 2 hours 40minutes. This result will be compared with the test drives with fuzzy logic controller. Figure 5: 3-D I/O Relationship of Design 3

79

PROSIDING SEMINAR NASIONAL RITEKTRA 2011 ISBN: 978-602-97094-2-1

4.2 Fuzzy Logic Enable (Smart-mode)

Table 5 Comparison 4 test drive conditions No Fuzzy

Figure 7: Fuzzy Logic Design 1

The E-bike is able to travel at maximum distance approx. 38 km in 3 hours. This result will be compared with the test drives with other fuzzy logic design.

Figure 8: Fuzzy Logic Design 2

Fuzzy Design1

Fuzzy Design2

Fuzzy Design3

Slope (Absolute Value)

0.2576

0.239

0.2633

0.2448

Max. Distance (Approx.)

36km

38km

36km

38km

Total Time

2:40:07

3:00:21

2:52:26

2:54:28

From the table above, Fuzzy Design 1 shows the lowest slope, which makes it better than the others. Comparing the maximum distance, Design 1 & 3 shows a better performance compared with other the other two test drive conditions. For the total time, comparing Fuzzy Design 1 and 3, the E-bike is able to travel around 38km for Fuzzy design1 in 3 hours, but during tested using Fuzzy Design 3, the E-bike can travel around 38km in 2hours 54 minutes, which is 6 minutes faster than Fuzzy Design 1. For the equally covered distance, Fuzzy Design 3 is preferable compared with design 1 due to the faster travel time, which is desirable. 5. CONCLUSION

Figure 9: Fuzzy Logic Design 3

Comparing 4 kinds of test drive conditions (based on above result), overall the lowest absolute value of the equation indicates that the design shows the best battery life time. Besides, the maximum distance also affecting the battery lifetime. The total time represents the maximum duration of the E-bike can travel. The comparison table is shown in Table 5 below.

In mechanical aspect, the Center of Gravity of E-bike design was analyzed, a new battery box and control box has been redesigned to relocate the Center of Gravity, improving its stability. In electrical aspect, the motor driver has been integrated with protective components to protect the motor driver from overcurrent. The E-bike has been integrated with 2 modes of operation, NORMAL and SMART MODE. For the program, three different configurations of fuzzy controller has been designed, simulated and analyzed. The most appropriate design has been determined, which is Fuzzy Design 3.

80

PROSIDING SEMINAR NASIONAL RITEKTRA 2011 ISBN: 978-602-97094-2-1

6. FURTHER DEVELOPMENT There are things that can be improved furthermore to improve the reliability of the bike, such as: 1. Test the fuzzy logic controller with regenerative braking ON 2. Develop a DC servo amplifier as an alternative to PWM

4.

REFERENCES 1. Pratama, Ferdian A. (2010). System Improvement and Fuzzy Logic Controller Development to Extend Battery Life-Time of Electric Bike System. Bachelor Thesis. Department of Mechatronics, Engineering Faculty, Swiss German University, BSD, Tangerang. 2. Dhanu, W. (2006). Designing and Constructing an Electric Bicycle based on Microcontroller. Bachelor Thesis. Department of Mechatronics, Engineering Faculty, Swiss German University, BSD, Tangerang. 3. Adibrata, J. (2007) Further Development and Performance 8. Proceedings, (pp. 208-212)

5.

6.

7.

Analysis of an Electric Bicycle based on Microcontroller. Bachelor Thesis. Department of Mechatronics, Engineering Faculty, Swiss German University, BSD, Tangerang. Yustisiawan, A. (2008). Development of Regenerative Braking and Battery Charging for An Electric Bicycle. Bachelor Thesis. Department of Mechatronics, Engineering Faculty, Swiss German University, BSD, Tangerang. Hachicha, M., Masmoudi, N., and Kamoun, L. A Robust Fuzzy Torque Control System for Electric Vehicles Application. Tunisie: National Engineering School of Sfax, Laboratory of Computer Science and Industrial Electronic of Sfax. Chi, Y.L., Wai, H.L., and Chang, Bruce. (2006). Applying fuzzy logic to an electric bicycle. Taiwan: Wu Feng Institute of Technology. Lin, T.Y. (2003). Apply fuzzy logic to smart-bike controller design. ISUMA '03 Proceedings of the 4th International Symposium on Uncertainty Modelling and Analysis,

81