UPTEC ES10 013
Examensarbete 30 hp April 2010
Design and Construction of an EV Driveline Prototype with an Integrated Flywheel
Nils Finnstedt
Teknisk- naturvetenskaplig fakultet UTH-enheten
Besöksadress:
Ångströmlaboratoriet Lägerhyddsvägen 1 Hus 4, Plan 0
Postadress:
Box 536 751 21 Uppsala
Telefon:
018 – 471 30 03
Telefax:
018 – 471 30 00
Hemsida:
http://www.teknat.uu.se/student
Abstract
Design and Construction of an EV Driveline Prototype with an Integrated Flywheel
Nils Finnstedt
Research shows that flywheels have a significant potential for improving the performance of EV (Electric Vehicle) drivelines. Flywheels can be used as power buffers that even out the energy flow between the primary energy storage device and the EV traction motor. This improves the potential energy density and extends the lifetime of the primary energy storage device of the EV.
In this degree project a prototype of a flywheel-buffered driveline was constructed.
The flywheel chosen was an electric motor/generator constructed at the Division of Electricity at Uppsala University. Lead acid batteries were used as the primary energy storage device in the driveline and the traction motor was a DC-motor.
Two DC/DC buck converters were designed for the driveline. The first limited the current from the batteries to the flywheel and the second controlled the power from the flywheel to the traction motor. Both converters were controlled by
microcontrollers. The current limiter was controlled by a hysteresis controller and the DC-motor power was regulated manually, under the constraint of a maximum current PI-controller. The buck circuits were simulated in MATLAB Simulink prior to their construction.
The performance of the driveline was satisfactory, despite the poor efficiency of the DC-motor. The results showed that the efficiency of the flywheel and the power converters was relatively high and that the flywheel had excellent power-buffering properties.
Examinator: Kjell Pernestål Ämnesgranskare: Hans Bernhoff
Handledare: Janaína Gonçalves och Johan Lundin
Sammanfattning
Forskning har visat att svänghjul har potential att förbättra prestanda hos drivlinor för elfordon.
Svänghjul kan användas för att jämna ut effektflödet mellan drivlinans primära energilager och dess drivmotor. Det primära energilagret kan vara optimerat för hög energitäthet istället för hög effekttäthet om effekten från det är utjämnat och maxeffekten reducerad. För batterier, som är det vanligaste primära energilagret för elfordon, ökar också livslängden och förlusterna sjunker om de kan leverera en konstant effekt istället för den varierande effekten med de höga maxströmmarna som drivmotorn kräver. Svänghjul är lämpliga att använda som energibuffertar i drivlinor för elfordon eftersom de både har goda effekthanterings- och energilagringsegenskaper.
En prototyp av en elektrisk drivlina med ett integrerat svänghjul har designats och konstruerats i detta examensarbete. Detta har varit en del av ett större forskningsprojekt på Avdelningen för Elektricitetslära på Uppsala Universitet. Det svänghjul som användes i drivlinan var en specialdesignad elektrisk motor/generator som tidigare konstruerats på avdelningen. Drivlinan bestod förutom svänghjulet av blybatterier, en DC-motor som drivmotor, en last som mekaniskt belastade DC-motorn, samt ett antal elektriska kretsar för att kontrollera effektflödet mellan drivlinans olika delar.
De elektriska kretsarna som designades i arbetet var två DC/DC-konverterare. Den ena hade syftet att begränsa strömmen fån batterierna till svänghjulets drivsystem och den andra att kontrollera effekten från svänghjulet till DC-motorn. Kretsarna designades något olika för att jämförelser av olika systemlösningar skulle kunna göras. Båda kretsarna kontrollerades av digitala mikrokontrollers. Kretsarna datorsimulerades innan de konstruerades.
Bortsett från DC-motorns verkningsgrad, visade mätningar att drivlinans prestanda var god.
Svänghjulet och dess drivkretsar visade sig ha relativt hög verkningsgrad och dess förmåga att jämna ut effektflödet från batterierna till drivmotorn var mycket god.
Table of Contents
1 Introduction... 1
2 Background ... 3
3 Aims and objectives... 4
4 Method ... 5
5 Theory ... 6
5.1 Electric machines ... 6
5.1.1 DC-Machines ... 6
5.1.2 PM Synchronous Machines ... 7
5.2 Batteries ... 8
5.3 Power Converters... 8
5.3.1 DC/AC ... 8
5.3.2 AC/DC ... 8
5.3.3 DC/DC ... 9
5.4 Electronic components... 10
5.4.1 Switches ... 10
5.4.2 Diodes ... 11
5.4.3 Microcontrollers... 11
5.4.4 Gate Drivers ... 12
5.5 Snubber Circuits... 12
5.6 Control Methods ... 12
5.6.1 Hysteresis Control... 12
5.6.2 PID Controlled PWM ... 13
5.7 Current Limitation ... 14
5.8 Power Quality ... 14
5.9 PCB Design... 14
6 Basic Choices of Design ... 16
7 Simulations ... 18
7.1 Current limiter simulations ... 18
7.2 Traction motor drive simulations... 20
8 Final Choices of Design... 23
8.1 Current limiter design ... 23
8.2 Traction Motor Controller Design ... 25
9 Results... 28
9.1 Power Converter Performance... 28
9.1.1 Current Limiter Performance... 28
9.1.2 Traction Motor Controller Performance ... 30
9.2 Driveline Performance ... 34
9.2.1 Steady State... 34
9.2.2 Drive Cycle ... 35
9.3 PCB ... 37
10 Discussion ... 39
11 Future Work ... 41
12 Conclusions... 42
Acknowledgements... 42
References... 43
Appendix... 44
A. DC-motor dynamics... 44
B. Power MOSFETs and IGBTs... 46
C. Snubber Design ... 48
D. Zieger-Nichols Rules ... 51
E. Microcontroller codes... 52
Current Limiter Code... 52
Traction Motor Controller Code ... 53
F. Switch Losses ... 57
Nomenclature
C Capacitor
Cb Snubber capacitor in parallel with battery Cd Snubber capacitor in parallel with diode Cs Snubber capacitor in parallel with switch Cw Drag Coefficient
Kp Proportional PID-coefficient I Current
Ia
IC
IL
IR
IS
Armature current Capacitor Current Inductor Current Resistive Load Current Switch Current
L Inductance M Torque R Resistance Ra Armature resistance Rs Snubber-resistor resistance Td Differential PID-coefficient Ti Integral PID-coefficient U
VL
Voltage
Inductor Voltage VR
VS
fn
Resistive Load Current Switch Voltage
Natural ringing frequency (Hertz) fo Cut-off frequency
n Motor speed
Φ Magnetic flux in air gap δ
ωn Natural ringing frequency (Radians) ζ Damping coefficient
Abbreviations
AC Alternating Current
CAD Computer-Aided Design
DC Direct Current
EMF Electro Motive Force ESR Effective Series Resistance
EV Electric Vehicle
FTP Federal Test Procedure
IGBT Insulated Gate Bipolar Transistors
MOSFET Metal-Oxide-Semiconductor Field-Effect Transistors PCB Printed Circuit Board
PI Proportional, Integral
PID Proportional, Integral, Differential
PM Permanent Magnet PWM Pulse Width Modulation
RC Resistor, Capacitor
RCD Resistor, Capacitor, Diode VRLA Valve Regulated Lead Acid
1 Introduction
One of the greatest challenges facing mankind today is to reduce the impact of human beings on the environment. Limiting climate change is essential for the future of human beings and other forms of life on this planet. One way of achieving this is to limit the use of fossil fuels. To do this as efficiently as possible, we need to both increase the percentage of alternative energy sources used, and to decrease our total energy consumption. One efficient method of decreasing our total energy consumption is to make new technology more energy efficient.
The transport sector consumes large amounts of fossil fuels with low efficiency. It is, with the exception of trains, almost exclusively driven by combustion engines powered by fossil fuels. An alternative to the fossil fuel-powered combustion engine needs to be found to reduce the environmental impact from traffic.
Electric motors are an alternative to combustion engines in vehicles. Electric motors can with high efficiency be powered from renewable energy sources and do not cause emissions where they are being used. Electric vehicles (EVs) have been manufactured as long as other types of automobiles [1]. The most common primary energy storage system used in EVs is batteries. The low energy density of batteries has, however, until recently prevented the EV from competing with combustion engine-powered vehicles. In recent years tremendous improvements have been made in terms of the energy density of batteries, but the range of EVs and battery lifetimes are still low compared to the range and performance of vehicles driven by conventional combustion engines.
Research has shown that one way of increasing the lifetime and energy density of EV primary energy storages is to integrate power buffers in the drivelines, in the form of flywheels [2] [3]. A flywheel is a rotating mass that stores energy in the form of kinetic energy. The principle is that the average power of the EV’s traction motor is transferred to the flywheel from the primary energy storage, while the varying traction motor power can be provided by the flywheel if the vehicle accelerates. Power can also be transferred from the traction motor to the flywheel when the EV brakes. This principle prolongs the lifetime of batteries, as they can work at a smooth and optimal discharge rate and because the number of charge/discharge cycles can be reduced [4]. The energy density of the primary energy source can be increased by the flywheel, as it enables the primary energy source to be optimised for high energy density instead of for high maximum powers. Thus by integrating a flywheel in the driveline, the range of the EV can be extended. An electric machine is a suitable device for charging and discharging a flywheel in an EV driveline.
Simulations have shown that the average power needed to propel a car differs greatly from the instantaneous power needed [5]. Figure 1 shows the simulation results of the average and the instantaneous power (excluding internal losses) of a car with a mass of 1500kg, a dimensionless drag coefficient, Cw, of 1.35, and a frontal area of 1.73m2 when driving according to a standard FTP 75 (Federal Test Procedure) urban drive cycle. The fact that the average power is so much less than the maximum power indicates the advantages of incorporating a flywheel into the driveline.
0 200 400 600 800 1000 1200 1400 -30
-20 -10 0 10 20 30 40
Time [s]
Power [kW]
Power - US urban drive cycle, FTP 75
Min = -25kW Max = 33kW
Average = 2.2kW
Figure 1: Calculated power for US urban drive cycle
The amount of storable energy is an important design parameter for a flywheel. If the flywheel is only to be used as a power buffer, it does not have to be very large. Figure 2 shows simulations of the energy stored in a flywheel if it is charged/discharged with the powers illustrated in Figure 1 [5]. The fact that the difference between the maximum and minimum energy in the flywheel is only 0.34kWh means that the flywheel only needs to be able to store that amount of energy. A flywheel that can store more energy can, on the other hand, have the advantage that it can be used for buffering energy when the EV is being charged from the grid.
0 200 400 600 800 1000 1200 1400
0 0.5 1 1.5 2 2.5
Time [s]
Energy Storage [kWh]
0 200 400 600 800 1000 1200 14000
20 40
Speed [km/h]
Energy stored in flywheel and battery
Vehicle Speed Energy in Battery Energy in Flywheel 0.34 kWh
Figure 2: Energy in flywheel and battery
It is that the flywheel has both high specific energy and specific power that makes it suitable as an EV power buffer. In comparison with batteries, flywheels can handle a lot more power, and in comparison with capacitors, they can store a lot more energy. Figure 3 shows that flywheels fill the gap between batteries and capacitors [4]. Low losses and a very long lifetime are two other
advantages of flywheels.
Figure 3: Specific energy and power for electric energy storage devices
Purely mechanical flywheels have been used for a long time in many different applications. The technology of electrically powered flywheels for EV applications is, however, not especially mature technology. So far, only EV prototypes with integrated flywheels have been constructed and tested.
2 Background
Research has since 2005 been carried out on energy storage in flywheels for EV applications at the Division of Electricity at Uppsala University. The novelty of the concept developed in Uppsala is that the driveline is divided into two voltage levels. The windings in the flywheel machine are arranged so that they divide the electric system into one high voltage level and one low voltage level, similar to an electric transformer. The advantage of this is that batteries and fuel cells, which can be used as primary energy storages for EVs, work intrinsically at low voltage levels while traction motors work more efficiently at a higher voltage level [5]. The two-voltage-level flywheel allows the batteries to be connected to the low voltage side of the flywheel and the traction motor to the high voltage side.
The aim of the research project is to design the complete EV driveline prototype that is illustrated in Figure 4. So far two flywheel prototypes have been designed and constructed at Uppsala
in a driveline. The tests that were performed previously used a grid-connected power supply as a primary power source for the flywheel and a variable resistor as a load.
AC/DC/AC Energy storage
(battery, fuel cell gas turbine etc.)
AC/DC
High voltage High power
Low voltage Low power
Motor/
Generator
Flywheel
Figure 4: Driveline design
3 Aims and objectives
The aim of this degree project was to design and construct an EV driveline prototype in which one of the flywheels designed at the Division of Electricity could be integrated. The driveline needed to include the eight parts presented in the block diagram in Figure 5 below.
Figure 5: Drive line components
The flywheel controller, the flywheel and the rectifier are the parts of the driveline that had been previously constructed by the research group. These parts were not changed, except that the flywheel controller circuit was replaced by a Printed Circuit Board (PCB) which was designed and mounted as a part of the degree project. Apart from this, a battery system and a traction motor with a braking load were selected and obtained, and a current limiting circuit and a motor controller circuit were designed and constructed. To limit the scope of the project, it was decided to make the driveline unidirectional, which means that power can only be transferred in one direction: from the battery towards the load. The driveline was designed with a limited budget and should be seen as a first prototype that can help the process of making a more advanced system in the future.
The objective of the degree project was to construct a test driveline, which would enable interesting measurements to be made to give information about how a flywheel affects the dynamics of a driveline system. Earlier, only tests of the flywheel itself, and not an entire driveline system were made. A driveline is a complex system with many parts connected in series, and to understand its dynamics fully practical tests need to be made. Apart from the dynamics of the total system, the project can provide information about how power converter circuits should be designed. To demonstrate the performance of the system constructed, it was decided that a number of measurements should be made, once the construction was complete.
4 Method
This degree project was carried out according to the method presented below:
• Literature study within the relevant area
• Choice of a model to simulate
• Simulation of the model
• Evaluation of the simulations
• Design of a system to construct
• Construction of the system
• Measurements of the system performance
• Evaluation of the system
• Conclusions and observations
This method was chosen as it was considered to be the most efficient way to achieve the aims of the project. A literature study provides the basic knowledge about how the different parts in a system work. Simulations give knowledge about the dynamics of the specific system to be constructed, so that a suitable design can be chosen for a prototype. Finally the construction of a prototype gives knowledge about how the system works in reality and how the parts interact with each other. A literature study and simulations save a lot of time and expense when constructing a complex technical system.
The following report describes the way in which the method was carried out in practice, explains in further detail the motivation for the choices made and presents the results.
5 Theory
The degree project started with a literature study within the relevant area. This provided the background theory needed to choose a suitable design and to interpret the results. The relevant theory is summarised in the following section.
5.1 Electric machines
There are several different types of electric motors and generators. Three types of machines commonly found as traction motors in EVs are synchronous motors, asynchronous motors and DC- motors [6]. As the work in this thesis has been focused on a DC-motor, and the flywheel in the driveline is a three phase synchronous AC-motor, the theory will focus on these two motor types.
5.1.1 DC-Machines
As implied by their name, DC-motors can be fed directly from a DC power source. Apart from this, DC-motors are suitable as EV motors because they have a high starting torque and they are easy to speed control [7].
Like all rotating electric machines, the DC-motor has a rotor and a stator. A commutator that is integrated in the motor transmits the relative rotation between the magnetic field in the stator and the rotor in the DC-motor. In contrast to other types of motors, the DC-motor’s main magnetic flux is constant with respect to the stator, and rotating with respect to the rotor. The stator of a DC- motor contains field windings or field magnets if it is a permanent magnet (PM) machine. The rotor contains armature windings that are isolated from each other and placed in axial slots in the laminated rotor and connected to the commutator. The commutator consists of copper segments that are isolated from each other and are in contact with connectors on the motor via brushes. When the rotor rotates, the brushes slides across the different segments on the commutator so that different armature windings carry the armature current, which means that the magnetic field made by the armature windings shifts so that there is always a magnetic torque applied on the rotor.
Figure 6 shows the different parts of a DC-motor.
Field windings
Brushes
Stator Rotor
Commutator Figure 6: DC-motor
DC-motors can be magnetised either by permanent magnets or by electro-magnets. The field windings in electrically magnetized DC-machines can be fed in different ways. They can be fed by
a separate DC-source or by the same source as the armature, in parallel with the armature, in series or in a combination of both. The dynamics of the motor will depend on how the armature winding is fed. The traction motor used in the driveline constructed in this degree project is a compound motor. This means that there are two field windings; one is in series with the armature, called the series winding, and one in parallel with the armature and the series winding, called the shunt winding. More about how these windings affect the dynamics of DC-motors, and the dynamics of DC-motors in general, can be found in Appendix A.
There are many ways of controlling the output of a DC-motor. The currents through the different windings in the motor, that decide the torque and speed, can be regulated both by connecting resistors in series with them or by regulating the supply voltage to the motor. The outputs can be controlled via the inputs, manually or automatically or in a combination of the two.
5.1.2 PM Synchronous Machines
Synchronous electric machines are AC-powered machines in which the rotor rotates at the same speed as the magnetic field from the stator. All synchronous machines have magnets in their rotors.
PM machines are magnetised by permanent magnets.
Traditional PM synchronous machines, in contrast to traditional DC-machines, have the magnets in their rotor inside their stators. There are, however, other types of synchronous machines. In the driveline built in this degree project, the electric machine that works as a flywheel is an axial flux machine [8]. This means that the stator is placed axially between the two rotor discs that carry the permanent magnets. The stator is made of bakelite and carries two sets of windings; one high voltage winding and one low voltage winding. The machine can be seen in Figure 7.
Stator
Rotor
Figure 7: Flywheel machine
5.2 Batteries
Batteries are the most common energy storage devices used in EVs. There are several types of battery technologies that are suitable for different applications. Essential features of EV traction batteries are the following [9]:
- High energy density
- High charging and discharging power density
- Long lifetime with maintenance-free and high safety mechanisms
- Wide acceptance as a recyclable battery from an environmental standpoint - Price
Four types of batteries commonly found as traction batteries in EVs are lead acid, nickel cadmium, nickel metal hydride and lithium-ion batteries. However, in this degree project only lead acid batteries were used.
Lead acid batteries have a relatively low energy density (30-50Wh/kg), but are the most economical solution for larger power applications where weight is of little concern. In addition to being inexpensive, lead acid batteries are robust and the technology is well-established and used in many applications. There are two types of lead acid batteries: flooded lead acid batteries that require maintenance by periodic replenishment of distilled water, and valve regulated lead acid (VRLA) batteries that are maintenance free [10].
5.3 Power Converters
Power converters are electronic systems that convert power flows from one current and voltage level to another. The following section will provide a theoretical discussion of the three different types of power converters that are involved in this degree project.
The three converters used are based on the same components. They are composed of switches, diodes, drivers, capacitors, inductors, resistors and control systems. Capacitors and inductors are passive components that store energy in electric fields and magnetic fields, and are used to stabilise voltage and current.
5.3.1 DC/AC
DC/AC converters, or inverters, convert direct power to alternating power. The DC/AC converter used in the driveline constructed in this degree project, powers the variable speed flywheel machine on its low voltage side. This means that it needs to vary both the output frequency and the voltage.
The DC/AC converter used is controlled by a microcontroller and consists of an IGBT bridge and an output filter [11].
5.3.2 AC/DC
AC/DC converters, or rectifiers, convert alternating power to direct power. The rectifier used in the driveline constructed in this degree project is used for extracting DC power from the three-phase flywheel machine on its high voltage side.
The AC/DC converter used in the driveline is a full-wave passive three-phase rectifier. It is constructed with six power diodes that are connected in a bridge as shown in Figure 8 [12].
Diode bridge 3-phase AC source
Load
Figure 8: Passive rectifier
When the rectifier is fed by a three-phase AC current the diodes that conduct are the two connected to the phases with the momentarily highest and lowest phase voltages. The output DC voltage from the rectifier can be smoothened by capacitors in parallel with the load.
5.3.3 DC/DC
As implied by the name, DC/DC converters are power converters that convert DC power at one level of current and voltage to another. Some converters increase the voltage and decrease the current, while others do the opposite. There are also DC/DC converters that both can increase and decrease voltage depending on how they are controlled. The types of DC/DC converters designed in this degree project are step-down converters.
Buck Converters
Buck converters, or step-down converters, are converters from which one can obtain an output signal (current or voltage) which is lower than the input. How they work can be understood by studying Figure 9 that shows a circuit diagram of a buck converter.
DC source Switch
Load Diode
Inductor
Capacitor
Figure 9: Buck converter
When the switch of the converter is conducting, the inductor current will increase and the inductor will produce a negative voltage as it builds up energy in a magnetic field. When the switch opens, the energy in the inductor will produce a positive voltage that will power the load so that the current flows through the diode instead of the power source. The capacitor filters the output voltage.
The output voltage of the buck circuit increases when the switch is closed and decreases when it is open. If the inductor current never goes to zero, the average output voltage will be a linear function of the average ratio of the switch position [13]. For example, if the switch is closed 50 % of the
total switching period, the average voltage across the load will be 50 % of the power source voltage.
Figure 10 shows typical buck converter waveforms. It shows a sequence where the converter switch switches from its on-position to its off-position two times. VL is the voltage across the inductor, IL is the current through the inductor, VR is the voltage across the resistive load, IR is the current through the resistive load, VS is the voltage across the switch, IS is the current through the switch and IC is the current through the capacitor.
Figure 10: Buck converter waveforms
5.4 Electronic components
Electric systems are built of many different connected parts. In this section the electronic devices used in the power converters built in this degree project will be examined.
5.4.1 Switches
Discrete power switches are essential components of power converters. There are many types of switches, all of which have different advantages and disadvantages. All switches used in switched mode power converters are semiconductors. The reason for this is that they have much shorter switch times than any mechanical switch, which is of great importance in reducing switching
losses. Values that are important when selecting a switch can be summarised in the following list [14]:
- Maximum current carrying capability - Maximum voltage blocking capability
- Forward voltage drop during ON and its temperature dependency - Leakage current during OFF
- Thermal capacity
- Switching transition times during both turn-on and turn-off - Capability to stand dV/dt when switch is OFF or during turn-off - Capability to stand dI/dt when switch is ON or during turn-ON - Controllable dI/dt or dV/dt capability during switching transition - Ability to withstand both high current and voltage simultaneously - Switching losses
- Control power requirement and control circuit complexity
As mentioned, switching losses occur when switches have long switching times. Ideally, a discrete switch has no losses. During the switching process, however, for a short period of time some current will pass through the switch and there will be a voltage drop across it simultaneously. The power that is burnt off during the switching processes defines the switching losses. The longer time the switching process takes, and the more often it happens, the larger the losses will be.
All types of switches are controlled by control signals, some based on current and some on voltage.
Both types of switches used in this degree project, MOSFETs and IGBTs, are controlled by voltage. When a certain positive voltage is applied to their control input, known as gate, relative to their cathode, they will conduct. If the gate is at the same or lower potential than the cathode, the switch will block the current from running from the anode to the cathode. A comparison between IGBTs and MOSFETs can be found in Appendix B.
The switching time of a voltage-controlled switch is determined by the time it takes for the gate voltage to rise and fall. Even though both MOSFETs and IGBTs have electrically isolated gates, it takes some time to change their potential, as they have built-in parasitic capacitors, both between the gate and the anode and between the gate and the cathode. The switching time is the time it takes to charge or discharge these capacitors. To achieve a short switching time, a switch with a low gate charge, which is the charge that is required to charge both the gate capacitors, should be chosen.
5.4.2 Diodes
Diodes are passive semiconductors that conduct current in one direction and block current in the opposite direction. Power diodes are useful in applications such as rectifying circuits and to provide current paths for inductive loads [15]. Diodes conduct when the voltage of the anode, in relation to the cathode exceeds a certain voltage named knee voltage, often around 0.7V [15]. If the voltage is increased above this level, the conducting current increases dramatically.
5.4.3 Microcontrollers
The switches in a power converter system need a system that controls their operations. If the power flow in the converter is to be automated, an automatic control system needs to be implemented. A
control system needs components that can interpret the system outputs and transform them into an input signal. A suitable component for this is the microcontroller.
A microcontroller is a small computer on a single integrated circuit. It contains a central processing unit, a clock, I/O ports, and a memory. The microcontroller can be programmed, which gives great freedom to the user. Suitable microcontrollers for power switch control also have analogue-to- digital converters that make it possible for them to interpret analogue measurement signals.
5.4.4 Gate Drivers
Gate drivers are the interface between control systems and high power electronic systems. As switching losses depend on the time taken to charge and discharge the gate capacitors of power switches, they should not be driven directly by the logic outputs from devices such as microcontrollers. A gate-driving circuit for a voltage-controlled switch is a circuit that is made for injecting or removing the gate charge fast. The larger currents the driver can handle, the faster the gate charge can be injected or removed and the more efficient the power circuit will be. It is therefore of great importance for the efficiency of the power system to have a well-designed gate driver circuit [16].
5.5 Snubber Circuits
Snubber circuits, or snubbers, are small circuits that are added in many power converter circuits to protect sensitive semiconductor devices from being damaged. Snubbers also fulfill other tasks. The improvements snubber circuits can make to a power converter system are summarised in the following list [17].
- Reduce or eliminate voltage or current spikes - Limit dI/dt or dV/dt
- Transfer power dissipation from the switch to a resistor or a useful load - Reduce total losses due to switching
- Reduce electro magnetic interference by damping voltage and current ringing
There are many different types of snubbers that are specialised for the different functions listed above. The snubber that has been constructed in this degree project has been optimised to fulfill two functions: to limit voltage spikes and to reduce ringing. A description of what causes these problems and how snubber circuits should be designed to solve them, can be found in Appendix C.
5.6 Control Methods
There are several ways to control power systems. Different control strategies suit different power systems and different control parameters. In all switched mode circuits, the positions of the switches can be controlled. The two control strategies used in this degree project are hysteresis control and PI-controlled PWM.
5.6.1 Hysteresis Control
Hysteresis control is a controlling strategy that is suitable for controlling current and voltage in buck converters. It works through measuring the output that is to be controlled and comparing it with a reference value. If the value of the output is larger than the reference value, the switch is turned off, and if it is smaller, the switch is turned on. To prevent the switch from chattering when
it tries to keep the output at the reference value, a hysteresis dead band is created around the reference value [18]. This means that the switch is turned off when the output is slightly above the reference value and turns on when it is slightly below it. The larger the dead band is, the slower the switching frequency will be, at the expense of causing a larger ripple in the output.
Hysteresis control is easy to implement and very robust. It gives an optimal input-output response time, and the overshoot is eliminated [18].
A drawback of the hysteresis control strategy is that it puts high demands on the output measurements. The measuring frequency needs to be a magnitude higher than the desired switching frequency and the measuring noise a magnitude smaller than the hysteresis band.
5.6.2 PID Controlled PWM
Pulse width modulation (PWM) is a commonly used method for modulating switching devices. The basic idea is that a switching frequency is defined, and the duty ratio, which is defined as the time that the switch is on in relation to the PWM period, is varied to change a given system output.
If the duty ratio is to be regulated automatically by some kind of feedback system from the output, a controlling function must be defined that relates the duty ratio to the error of the output. The simplest regulator is a P-regulator (proportional) that adjusts the duty ratio proportionally to the output error. More sophisticated regulators can also integrate the error and form a PI-regulator (proportional, integral), or can also take the derivative of the error in what is known as a PID (proportional, integral, differential) regulator.
A lot can be said about the PID regulator. The basics can, however, be understood fairly easily. The P term of the regulator works by forming a control signal by multiplying the output error with a constant Kp. A large Kp constant gives a fast regulator, but can make the output oscillate. The effect of the P term of the regulator is large when the error is large, but small when the error declines. In many systems, the error does not converge to zero if only a P regulator is used. This is the reason why PI controllers are used. The I term of the regulator forms the control signal by multiplying the integral of the output error with a constant Ti. If the output error does not converge, the accumulated error grows and the PI controller reacts. The D term of the regulator is used to increase its speed without making the output oscillate. It works by multiplying the derivative of the output error with a constant Td. [19].
In order to get a good response from a P, PI or PID regulator it is essential to tune it. If a mathematical model of the regulated system can be derived, there are several analytical methods to find suitable values for the proportional, integrating and derivative constants. If a mathematical model cannot be found, e.g. if the system is very complex, there are several experimental methods that can be used to tune the regulator. One method that can be used is the Zieger-Nichols rules for tuning PID controllers [20]. This method can be found in Appendix D.
One of the advantages of PWM is that the duty ratio does not have to be regulated as fast as the switching frequency. This means that the output can be sampled at lower frequency than that required for hysteresis control. It also means that there will be time to filter the measurements and consequently eliminate their noise.
One of the drawbacks of PWM regulation is that it is not as fast as hysteresis control. If the speed of the PID regulator is increased by using a too large proportional constant, instability problems with large overshoots will occur [18].
5.7 Current Limitation
When an electric motor is started, the current can exceed the maximum level that its power system can handle. This is due to that the back EMF, that normally limits the motor current, is low when the motor is staring as it is proportional to the speed of the motor. Because of this, many motor drives have some kind of current-limiting system that works when the motor shall be started. The easiest way to limit the current to a motor is to connect resistors in series with it. Another way is to make a buck circuit that decreases the voltage applied to the motor when the current level is too high. Using a buck circuit to limit the current limits the losses in comparison to using resistors.
5.8 Power Quality
Ideally the power in electrical systems is delivered by single frequency AC or DC voltages and currents. In reality this is not the case. Voltages and the currents in all power systems contain both noise and harmonics. The content of power distortions in a system defines the power quality. The power quality issues that have been taken in account in this degree project are the current ripple created by the DC/DC-converters and the voltage spikes created when switches turn off.
There are many reasons why high power quality is desirable.
Current and voltage ripple leads to decreased efficiency and shortens the lifetime of power systems [21]. Many power electronic devices are sensitive to voltage distortion and can malfunction or shut down if the voltage quality is too bad. Moreover, voltage distortion leads to current distortion in all types of loads. Bad current quality leads, amongst other things, to increased losses. As resistive losses are given by I2R, the total resistive energy loss in a system will increase the more uneven the current is. In addition to this, due to the skin effect, the resistance in a conductor increases with the frequency. This means that the higher the frequency is, the more the current is concentrated to the surface of the conductor and less of its cross section area is utilized [21]. Another reason for reduced efficiency in systems with bad current quality is the fact that the magnetic flux will vary around the conductors, if the current in them varies. Varying magnetic flux leads to hysteresis losses and eddy losses in the iron in electric machines. The hysteresis losses are proportional to the frequency of magnetic flux variations and the eddy losses to the frequency squared [21].
There are many ways to improve the power quality in a power system. There are various types of filters, both passive and active, that can be implemented. Filters can, however, create losses themselves. In the case of switched DC-DC converters, increasing the switching frequency or the size of filter inductors can decrease the current ripple. The switching frequency and the size of the filter should be optimised to minimise the total losses in the system.
5.9 PCB Design
A printed circuit board (PCB) is a board specially designed for a certain circuit, on which components can be soldered. The board consists of a substrate and copper traces that are printed on it. The substrate is an isolator that provides a structure that physically holds the circuit components and the printed wires in place [22]. The wires can be placed on both sides of the substrate, and also
inside the substrate in multi-layer PCBs. The conductors in the different layers can be connected to each other through holes called vias.
PCBs are manufactured industrially by photoplotters and CNC machines. A digital description of the board design is needed in order to manufacture it. This description is made by the PCB designer with a computer-aided design (CAD) program [22].
6 Basic Choices of Design
The first choices of system design were made based on the literature study, earlier experience, design criteria and practical limitations. The following section gives a description of and motivation for the basic choices that were made before any simulations of the system, or practical experiments were carried out. The parts of the driveline that these choices affected are presented in the list below:
• Battery
• Current limiter
• Traction motor controller
• Traction motor
• Load
The traction motor used was a compound DC-motor rated 10kW at 60V and 205A. This was chosen for practical reasons. The motor was originally bought for other purposes but fitted well in the system. No modifications were made to the motor, except for a mounting frame and a shaft coupling between the motor and its load.
The mechanical load used to brake the motor was a DC-generator rated 1.9kW. The electrical load used to brake the generator was a variable resistor rated 0-630Ω. These loads were chosen mostly for practical reasons. Both the generator and the variable resistor were found in the lab where the driveline was built. Beside the convenience of using these loads, they were suitable in the driveline.
Both the braking torque and the output power could easily be adjusted by regulating the field current in the generator and the resistance of the resistor.
Lead acid batteries were chosen for powering the driveline. Four 12V batteries connected in series gave an output voltage of 48V. This was enough to bring the flywheel to a speed which gave a high voltage output of 80V. As the DC-motor was rated at 60V, this low voltage level was suitable. The batteries chosen were 45Ah gel batteries. Lead-acid batteries were chosen as they are relatively cheap, safe and easy to charge. 45Ah was a suitable value for the energy content of the batteries, as it allows the flywheel to be run for the time it takes to do the tests likely to be done during one day (~4 hours at 500W average input power). Gel batteries which are a type of VRLA batteries were chosen as they are maintenance free.
Two buck circuits were chosen for the current limier and the DC-motor controller. Buck circuits were chosen as they could perform what was required of the two power converter systems and as they are efficient. The control systems were based on microcontrollers, as they are cheap, fast and robust.
The current and the voltage quality are important design parameters of a power system. Finding the optimal level of the power quality needs a lot of measurements. These measurements could not be made before the driveline had been constructed. However the power quality had to be set to some level for the power converters to be designed for. A current ripple of 0.5A was chosen as a suitable value to aim for. This value can easily be changed in the future when an investigation of the losses has been made.
Table 1 summarises the basic choices of design that were made in the beginning of the project.
Table 1: Basic choices of design
Battery VRLA batteries 48V
Power converters Buck converters
Power quality 0.5A maximum DC current ripple
Traction motor Compound DC-motor
Driveline load DC generator with variable load resistor
7 Simulations
After a literature study on the types of power circuits required in the driveline, and a decision about their topography, the circuits were simulated on a computer. The aim of the simulations was to understand the dynamics of the system and to discover what results to expect when certain values of the components in the system were changed. The simulations also provided an indication of which components should be ordered. This saved both time and money.
The simulations were made in MATLAB Simulink, which is a toolbox for MATLAB made for modelling and simulating dynamic systems. The programming interface is graphical and consists of block programming tools.
The method used to make the simulations of the buck circuits was to control the switches by feedback loops from current-measuring device in the system. The current measurements gave an output signal that was converted to a logic input for the switch via a hysteresis control block. This had the advantage that for a certain current quality, given by the hysteresis band width, and a certain current, given by the hysteresis band level, the switching frequency could be found. If the frequency was too high, larger inductors could be added to the system, as they make the current level more stable.
7.1 Current limiter simulations
To reduce the complexity of the simulations some simplifications of the system were made. The load for the current limiter was in reality the flywheel, driven by the DC/AC-converter. This was approximated as a constant resistive load, with the same resistance as the windings of the flywheel.
In reality, the load had inductance and a back EMF, which were ignored here. The inductance of the flywheel is, however, small (0.19mH), as it is coreless and the back EMF is small in the start-up phase, when the current limiter will be used. As the inductance of the windings and the back EMF make the current level more stable, the approximation can be seen as a worst case scenario.
Designing the current limiter so that it can limit the current without being dependent on inductance from the load increased the reliability of the system.
Other simplifications made include the imperfections in the used components. For example the reverse recovery time of diodes and parasitic inductances in conductors were ignored. These are very important factors for the system, but as the simulations were not used for studying transients, these simplifications were possible.
Many different simulations were made and the model used is shown in Figure 11.
Figure 11: Simulink current limiter model
Figure 12 and Figure 13 show two different plots from a simulation of the current limiter. Figure 12 shows the control signal to the switch and the inductor current measured by the system during the beginning of a simulation. Figure 13 shows the load current and load voltage during the same time period and simulation as in Figure 12. The simulation is made for a hysteresis band between 9.75A and 10.25A, a source voltage of 48V, an inductor of 3.75mH and a capacitor of 4.7mF.
0.5 1 1.5 2 2.5
x 10-3 9
9.5 10 10.5
Hysteresis controlled current
Measured current (A)
Time (sec)
0.5 1 1.5 2 2.5
x 10-3 0 1
Switch control signal
Figure12: Inductor current and control signal
0.5 1 1.5 2 2.5 x 10-3 8.8
9 9.2 9.4 9.6 9.8 10 10.2
Load current and voltage
Load current (A)
Time (sec)
0.5 1 1.5 2 2.5
x 10-3 11 11.5 12 12.5 13
Load voltage (V)
Figure 13: Load current and voltage
The simulations showed that the current ripple through the resistive load was smallest if the current measuring device was placed between the diode and the capacitor. If the device was between the capacitor and the load, the current ripple was the same as the hysteresis band. If it was placed before the capacitor, the capacitor evened out the current and decreased the ripple.
The simulations also showed that the greater the filter capacitance was, the smaller the voltage ripple across the load was, up to the level when the capacitor current increased and decreased relatively linearly with respect to the time. Increasing its size of the capacitor above this level did not improve the power quality significantly. The smaller the hysteresis band used, the smaller the capacitor needed to be.
Figure 12 and Figure 13 show that the switching frequency of the current limiter needs to be around 4.9 kHz to keep the inductor current ripple within 0.5A.
7.2 Traction motor drive simulations
Simplifications were also made in the simulations of the DC-motor controller. The motor was simplified by two parallel current paths with the impedance of the DC-motor’s shunt winding and its series and armature winding. As in the current limitation circuit, the back EMF and the imperfections in the components were ignored. The inductance of the motor was, however, not ignored as it is much larger than in the flywheel. The simplifications were motivated in the same way as regards the current limiter.
As the motor was simplified to its equivalent impedance, no control system based on the output of the motor could be simulated. This was the reason why the simulations were made by controlling the input to the motor. The current was controlled in the same way as in the simulations of the current limiting circuit. However, this was not a problem, as the objective of the simulations was to give information about the power converter system, not the motor output dynamics. Simulations were performed both with and without an extra inductor in the buck circuit. As the motor is an
inductive load, an extra inductor was not strictly required. Another difference between the current controller buck circuit and the motor control buck circuit was the filter capacitor.
The model used for the motor controller simulations is shown in Figure 14.
Figure 14: Simulink DC-motor controller model
Figure 15 and Figure 16 show two different plots of a simulation of the motor control power system. Fig15 shows the control signal to the switch and the current measured by the control system during the beginning of a simulation. Figure 16 shows the armature current and voltage during the same time period as in Figure 15. The simulations were made for a hysteresis band between 19.75A and 20.25A, a source voltage of 60V without an extra inductor in the buck circuit.
0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 x 10-4 19
19.5 20 20.5
Hysteresis controlled curent
Measured current (A)
Time (sec)
0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4
x 10-4 0 1
Switch control signal
Figure 15: Total motor current and control signal
0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4
x 10-4 19
19.5 20
20.5 Armature current and voltage
Armature current (A)
Time (sec)
0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4
x 10-4 -10 0 10 20 30 40 50
Armature voltage (V)
Figure 16: Armature current and voltage
The simulations showed that it was not suitable to use filter capacitors anywhere in the motor control buck circuit. As all parts of the load are inductive, it was not suitable to add a filter capacitor in parallel with any part of the load. Connecting a capacitor in parallel with an inductor causes a resonance phenomenon that is associated with increased losses.
Figure 15 and Figure 16 show that the switching frequency of the DC-motor controller needs to be around 34 kHz to keep the current ripple within 0.5A without any filter inductor or capacitor in the buck circuit.
8 Final Choices of Design
After the computer-based simulations, practical experiments were made. Before the experimental process started further designing parameters were chosen. These decisions were necessary prior to starting the experiments, as a lot of components needed to be ordered. Some of the decisions were motivated by the simulations and some by practical reasons. As the designing of the driveline was part of a research project, some choices were motivated by the fact that interesting conclusions could be drawn by choosing and evaluating certain system solutions. As two buck-converters were constructed, different system solutions could be made for them both and their function could be compared. The following section describes the choices that were made after the simulations were completed and during the experimental process. Table 2 summarises the system solutions for the two buck-converters.
Table 2: Final choices of design
Part of the Power Converter Current Limiter Traction Motor Controller
Switch IGBT MOSFET
Buck output filter Diode, inductor, capacitor Diode
Snubber circuit No Yes
Switch driver IR2110 IR2110
Microcontroller dsPIC30F 2010 dsPIC30F 2010
Current sensor Hall Effect sensor Hall Effect sensor
Control strategy Hysteresis control PWM, PI maximum-current control 8.1 Current limiter design
The current limiter was the first of the two DC/DC converters to be designed. This fact affected its design as efforts were made not to make it too complex. If the first converter was simple, more time could be spent on the second converter, when experience from the first one had been gained.
The switch that was used in the current limiting circuit was an IGBT. This was partly decided as the research group that the degree project was made for, had experience of the actual IGBT.
Another reason for choosing an IGBT was that the current limiting switch only switches when the flywheel is in its start-up phase. The rest of the time the switch is continuously conducting. This means that the most important performance of the switch is its static conducting performance, which means that the IGBT is a more suitable switch than a MOSFET. IGBTs are also in general rated for higher voltages than MOSFETs. This makes the requirement for voltage-spike-reducing snubber-circuits smaller when using IGBTs, which was a reason for using an IGBT in the first converter to be designed. The actual IGBT that was used was a module containing two transistors, which meant that the body diode of one of the switches could be used as a protection diode in the buck circuit. The IGBT was a SKM600GB066D made by Semikron, rated at 600V and 690A (continuously).
By studying the simulations it can be seen that if the current smoothening inductor in the current limiting circuit is 3.75mH, and the current ripple is set to a maximum of 0.5A, a switching frequency of 4.9kHz is required. This is a suitable switching frequency for the IGBT as it can be made both higher or lower if the specifications of the power quality are changed. On the basis of this conclusion, an inductance value of 3.75mH was used in the experimental setup. The filtering
capacitor was set to 4.7mF, as the simulations showed that this was adequate. The inductors that were used were chosen as they were readily available and had suitable inductance values. The conductors of the inductors were, however, under dimensioned for the application.
The IGBT driver and microcontroller were also chosen because of earlier experience. The microcontroller chosen was a dsPIC30F 2010 made by Microchip. It is a 16-bit digital signal controller with a 10-bit 1 Msps Analog-to-Digital Converter. The driver was an IR2110 made by International Retifier. It can deliver an output current of 2.5A and has a high side and a low side that can be used for driving two switches. However, only the low side was used in the setup. The driver voltage was set to 15V.
Hysteresis control was used for the current controller. This was mainly motivated by the fact that it is fairly straightforward to implement. The code made for the hysteresis controller can be found in Appendix E.
A Hall Effect current sensor was used. This was motivated by the fact that they have good accuracy and do not create excessive losses in the power circuit ass when measuring current with a shunt resistor. The current limiter was designed so that the level at which the current was limited could be adjusted with a potentiometer connected to the microcontroller. The current sensor was a HAL 50-s made by LEM. Figure 17 shows a circuit diagram and Figure 18 shows a photo of the current limiter circuit.
Input
Hysteresis Control System
Controller Flywheel
3.75 mH L
IGBT
4.7 mF 48V C
BATTERY
Current Sensor
Figure 17: Current limiter circuit diagram
Capacitor
Hysteresis Control System Inductor
Current Sensor
IGBT
Figure 18: Photo of current limiter circuit
8.2 Traction Motor Controller Design
Many of the design parameters chosen for the DC-motor controller differed from those chosen for the current limiter. This decision was made because interesting conclusions could be made by comparing different system solutions.
One difference was that a MOSFET was chosen for the switch in the motor controller buck circuit.
The choice was motivated by the fact that the switch in the motor driving circuit has to switch under all conditions, except for when maximum current is required, and therefore the dynamic performance of the switch is very important. Also, as a MOSFET was chosen, it was possible to have a sufficiently high switching frequency so that no extra inductor was required in the power converter. The chosen MOSFET was an IXFN140N20P made by IXYS. Its rating is 200V and 140A (continuously).
As the MOSFET has a faster switching time than the IGBT and has a lower voltage rating, measurements showed that a snubber circuit was required in combination with the MOSFET.
The first snubber to be designed was intended to reduce voltage peaks. Four 22μF polypropylene capacitors were connected in parallel with the power source. This reduced the voltage spikes essentially, but to make them even smaller a 10nF capacitor was connected in parallel with the MOSFET. As a MOSFET is an effectively resistive conductor, the discharge of this capacitor should not be a problem. This will have to be proved by testing the lifetime of the switch. The 10nF
capacitor was chosen because it reduced the current spike essentially. If this reduction had been made by increasing the size of the capacitors in parallel with the power source, their size would have had to be increased substantially. This would have been a more costly and space-consuming solution than connecting the capacitor in parallel with the MOSFET.
To reduce the ringing, an extra capacitor and resistor were connected in parallel with the MOSFET according to the damping theory presented in Appendix C. The ringing frequency was measured to 7MHz. This ringing frequency in combination with the 10nF voltage snubber capacitor in parallel with the switch, means, according to Equation C3 (found in Appendix C), that the damping resistor should be 2.27Ω. A 10Ω thick film power resistor was used as a damping resistor, as it was readily available. Equation C6 motivated that a 15nF capacitor was used as high pass filter capacitor. An attempt to build a RCD-snubber was made. The result, however, was not satisfactory.
The same control system hardware was chosen as for the current limiter. The same microcontroller, current sensor and gate driver were used in both systems. The software was, however, designed differently. As a comparison between hysteresis control and PWM control is of interest, it was decided to use a PWM control for the motor controller. As the time for the degree project was limited, no closed loop control system for the motor speed, torque or power were made. However, a closed loop controller for limiting the maximum current to the motor was constructed, for protection reasons.
The motor controller was designed so that the motor torque and speed could be regulated by varying the duty ratio of the switch via a potentiometer connected to the microcontroller, on the condition that the current did not exceed its maximal value. If the potentiometer was adjusted to a duty ratio that would lead to a current larger than the maximum current, the duty ratio was automatically adjusted by a PI-controller.
If the current exceeded its maximum value, which was set in the microcontroller code, the PI controller regulated the duty ratio so that the current remained on the maximum level until the duty ratio was decreased manually with the potentiometer or the back EMF of the motor reduced the current below the maximum level. The PI regulator was tuned by using Zieger-Nichols rules, which can be found in Appendix D. The critical value of the proportional regulating constant Kcr that first gave sustained oscillations of the output was 0.2. If the proportional constant was set to 0.2, the period of the oscillations Pcr was measured to 1.2ms. According to Table D1 in Appendix D the values for the PI constants should be set to Kp=0.45*Kcr and Ti=Pcr /1.2. The constants were consequently set to Kp=0.09 and Ti=0.001. The duty ratio of the PWM signal was updated every other PWM cycle. The code made for the PWM controller can be found in Appendix E. Figure 19 shows a circuit diagram and Figure 20 shows a photo of the traction motor controller.
M
MOSFET
..
10 μF 15 μF C C
Traction Motor
Rectifier Output
2.2 μF C
.. 2.2 μF C
.. 2.2 μF C 2.2 μF ..
C ..
PI - PWM Control system
Current Sensor
10 Ω R
..
..
Figure 19: Traction motor controller circuit diagram
Current Sensor
Figure 20: Photo of traction motor controller circuit
MOSFET and Diode Snubber Capacitors
Circuit Braker
PI – PWM Control System
9 Results
As the main objective of this degree project was to design and construct a driveline, the main result is the actual driveline prototype. Figure 21 shows a photo of the final driveline prototype. To demonstrate the functioning of the prototype some measurements of the system performance have been carried out. Many more measurements can be made in the future. The measurements that have been made so far are described below.
Figure 21: Photo of driveline
Traction Motor
9.1 Power Converter Performance
The power converters are important parts of the driveline prototype. They affect both the efficiency and the dynamics of the system. Some measurements of the performance of the current limiter and the traction motor controller are given below. As the other two power converters were not designed as part of this degree project, no measurements of their performance have been made.
9.1.1 Current Limiter Performance
As mentioned, when the DC-current to the flywheel reached a certain level it was restricted within the limits of a hysteresis band with a buck converter. Figure 22 shows the inductor current in the buck converter, which is the current to the flywheel controller, and the control signal to the IGBT, when the current was limited to 10A.
Flywheel Batteries
Generator
Current limiter
Flywheel Controller
Rectifier
Motor
Controller
Load
Figure 22: Current limiter performance
Figure 22 shows that the switching frequency is around 14.2 kHz when the current is limited between approximately 9.7A and 10.3A. The switching frequency controlled by the hysteresis controller varies, partly because of varying back EMF from the flywheel, and partly because of noise on the current sensor signal. The current sensor used for measuring the current illustrated in Figure 22 was a Hall Effect sensor, as was the one used by the control system. The current measurements made by the control system were made with a sample rate of 170 kHz.
Figure 23 shows the emitter-collector voltage and the gate-collector voltage during a turn-on sequence for the IGBT.
