Embed Size (px)
Citation preview
1
1. Title Intelligent Power Management System Interim Report July 30, 1997
Table of Contents
1. Title Intelligent Power Management System ............................................................................................. 1 2. Abstract ........................................................................................................................................................ 3 3. Introduction ................................................................................................................................................. 3
3.1 High Speed Rail Transportation ............................................................................................................ 3 3.2 Locomotive Propulsion Options ............................................................................................................ 4 3.3 Flywheel Energy Storage ....................................................................................................................... 5 3.4 High Speed Rail Corridors ..................................................................................................................... 5 3.5 High Speed Train Set ............................................................................................................................. 6
4. Possible Optimizations ................................................................................................................................ 7 4.1 Prime Mover/Flywheel Sizing ............................................................................................................... 7 4.2 Energy/Fuel Efficiency .......................................................................................................................... 7 4.3 Flywheel Use ......................................................................................................................................... 7
5. Simulation .................................................................................................................................................... 9 5.1 Model Description ................................................................................................................................. 9 5.2 Results ................................................................................................................................................. 12
6. Demonstration ........................................................................................................................................... 14 6.1 Discussion ............................................................................................................................................ 14 6.2 Computer Controls .............................................................................................................................. 14
6.2.1 Philosophy of Control Communications ....................................................................................... 14 6.2.2 Analysis of Networked Control and Instrumentation ................................................................... 15 6.2.3 Review of Network Technology ................................................................................................... 19 6.2.4 Description of CAN Technology .................................................................................................. 22 6.2.5 Energy Management Organization ............................................................................................... 22 6.2.6 Information Communications ....................................................................................................... 25 6.2.7 Observations on the Energy Management Code ........................................................................... 29 6.2.8 Route Simulator and Track Load Simulator Processor ................................................................. 30 6.2.9 Network Communications Results ............................................................................................... 31
6.3 Power Control ...................................................................................................................................... 33 6.3.1 AC Drives ..................................................................................................................................... 33 6.3.2 Power Train .................................................................................................................................. 34 6.3.3 Load Simulator ............................................................................................................................. 39
6.4 Demonstration Results ......................................................................................................................... 39 7. Conclusion ................................................................................................................................................. 39 8. Recommendations ..................................................................................................................................... 39 9. Appendices ................................................................................................................................................ 40 10. Bibliography ............................................................................................................................................ 40
Table of Figures Figure 1. Route Speed Limit Profile ................................................................................................................ 6 Figure 2. Locomotive Power Train .................................................................................................................. 6 Figure 3. Block diagram of train simulation .................................................................................................... 9 Figure 4. Propulsion Model ........................................................................................................................... 10 Figure 5. Block diagram of the Locomotive Simulation Model. ................................................................... 11 Figure 6. Block diagram of the Speed Control Model ................................................................................... 11 Figure 7. Commanded Speed Compared to Track Speed Limit .................................................................... 12 Figure 8. Actual Velocity Compared to Track Speed Limit .......................................................................... 13 Figure 9. Net Power To Traction Motors ...................................................................................................... 13 Figure 10. Information relation diagram for distributed networked control. ................................................. 23 Figure 11. Demonstration train system control and instrumentation block diagram. .................................... 23
2
Figure 12. Energy management control LabVIEW display panel ................................................................. 24 Figure 13. CAN message format ................................................................................................................... 25 Figure 14. RTOS flow diagram ..................................................................................................................... 26 Figure 15. Demonstration train network with J1939 message routing. ......................................................... 28 Figure 16. Track load simulator functional block diagram ............................................................................ 31 Figure 17. Message timing for normal operations ......................................................................................... 32 Figure 18. Message timing for no response from Flywheel #2...................................................................... 32 Figure 19. CAN status update message. ........................................................................................................ 33 Figure 20. CAN status message update with no response from Flywheel #2 unit. ........................................ 33 Figure 21. VSI Schematic .............................................................................................................................. 33 Figure 22. Schematic of phase “A” of a 3-phase VSI .................................................................................... 34 Figure 23. Train energy control block diagram. ............................................................................................ 35 Figure 24. Open circuit terminal voltage as a function of field current. ........................................................ 38 Figure 25. DC bus voltage control MATLAB simulation model. ................................................................. 38
Table of Tables Table I.Comparison of control network technologies .................................................................................... 19 Table II. Evaluation matrix for selecting network technology. ..................................................................... 20 Table III. Master to PC data format. .............................................................................................................. 27 Table IV. PC to master data format ............................................................................................................... 27 Table V. Network message data formats. ...................................................................................................... 29 Table VI. Synchronous generator open circuit test results. ........................................................................... 37
3
2. Abstract The preliminary report on the investigation of Intelligent Power Management Systems addresses energy management issues for high speed trains that have renewable energy sources on board. The design issues that dictate the viability of flywheel energy storage on trains are the transfer capability of electronic converters, the magnitude of both prime mover and flywheel energy storage capacity, and the number of flywheels that energy storage is to be distributed among. This report discusses the simulation of trains to determine power ratings for both prime mover as well as flywheel sizing for trip time management and energy conservation.
The report considers the computational control needed for efficient and effective man-agement of this train system. A distributed networked control system is compared to a single central computer that is direct wired with instrumentation and control effectors. Issues of relia-bility and cost, both installation and maintenance are discussed. Bench mark requirements for processor and network performance and cost allow the identification of classes of network tech-nology suitable for critical train controls.
For the final report, the hardware and software used in the demonstration system will be described in sufficient detail to allow the reader to reproduce an identical system. The descrip-tion in this report includes discussions of power electronics, user interfaces with computer con-trols and network management schemes, and performance analysis of these systems.
3. Introduction The US Federal Railroad Administration has been pursuing the idea of using locomotives
with an on-board prime mover for high speed rail applications. Such transportation systems would not require the added cost of rail electrification. Gas turbines are preferred over diesel en-gines as prime movers for high speed rail in order to save weight. However, electric traction mo-tors are still preferred over fluid turbine drives.
This report presents preliminary results from a study of the possibility of adding flywheel energy storage to a high speed locomotive. The flywheels are charged whenever the locomotive is in regenerative braking mode and whenever the prime mover is producing more power than is needed to maintain the desired track speed yet operating in the optimal performance power range. The chief benefits to such a scheme are:
1) decreased trip time, 2) better acceleration at high speeds, 3) reduced prime mover power rating and weight, 4) reduced rail bed cost, and 5) improved fuel efficiency.
3.1 High Speed Rail Transportation Rail has long been a transportation option for both passengers and freight. Early locomo-
tives were based on a steam boiler fired by either wood or coal. The steam pressure was used to turn the drive wheels. These were eventually replaced with the diesel-electric locomotives com-monly used in North America today. The diesel-electric locomotive consists of a diesel engine that is the prime mover for a synchronous generator. Most modern locomotives have a 3 to 5000 HP synchronous generator, with the output rectified by a diode rectifier. The resulting DC bus then supplies the DC traction motors. A locomotive is a small (roughly 4 MW) power system on wheels.
4
Most of the rail traffic in North America consists of freight trains, which tend to be long, heavy and slow. There is also some passenger travel in the form of light and commuter rail sys-tems around urban areas and some government controlled long distance travel along lines owned by the freight railroad. Nearly all of these passenger systems are constrained by the civil speed limits which, for US rail lines, is 78 miles per hour. The chief exception to this rule is the North-east Corridor, where 120 mile per hour travel is allowed on limited access sections of track.
High speed rail transportation is getting increased attention as a possible alternative to air travel for business travel between cities that are between 150 and 400 miles apart. The chief ob-jective is to provide transportation between cities that are far enough apart to discourage auto-mobile travel and too close together for air travel to be efficient, generally in the range of 150 to 400 miles. Rail service would avoid the congestion and delays often present at major airports, and by reducing the traffic at the airports also relieve some of the congestion.
Fast, efficient high speed rail systems already exist in Europe and Japan. Examples would be the TGV in France and the TransRapid in Germany where trains often exceed 150 miles per hour. The European rail infrastructure is designed around all-electric locomotives, where the electric power to run the traction motors is provided through a pantograph and catenary supplied from an electric power distribution grid. These locomotives are typically referred to as all-electric locomotives. However, the catenary and utility interface infrastructure required for all-electric locomotives is not economically feasible in the US as studies in Florida, Texas, Califor-nia, the upper Midwest, and the Pacific Northwest have all shown. However, there is a project in Florida that is still moving ahead 1. Therefore, studies are being conducted on high speed loco-motives with on-board prime movers that drive synchronous generators. These locomotives are typically referred to as non-electric locomotives even though they use a small on-board auto-nomous electric power system.
3.2 Locomotive Propulsion Options The typical North American locomotive is a small electric power system. A diesel engine
powers a synchronous generator. The output of the generator is immediately rectified to DC by diode rectifiers. This DC bus supplies DC traction motors which have series excitation for high torque production at low speeds, ideal for freight operation. Each of these motors is capable of regenerating when the train is braking, with the energy dissipated by a braking resistor bank.
The traditional diesel electric design has several drawbacks for high speed rail applica-tions. Foremost among these is the size and weight of the diesel engine, which effectively limits the top speeds attainable for a diesel locomotive.
Gas turbines engines are considered to be a better option for a lightweight prime mover for high speed rail. Gas turbines have been considered for both mechanical drive systems and for electric drive systems.
DC traction motors have significant drawbacks. The tractive effort applied to the rails is limited by the axle with the poorest adhesion between its wheels and the rails. Once that axle starts to slip, all of the drive axles must be throttled back to require a zero slip condition. Typi-cally, this is the front axle on the lead locomotive, since it will be seeing the wettest, dirtiest rails and is thus most likely to slip.
Significant improvements in locomotive performance can be made by utilizing induction motors driven by inverters in place of DC traction motors. A system with a single inverter per traction motor allows for individual adhesion control for each axle, significantly improving the efficiency in getting the power to the rail. In addition, the largest DC motor that will fit on a lo-
5
comotive is roughly 1000HP. However, a higher horse power induction motor will fit in the same space since it does not require space for the commutator. There has also been a significant move towards ac propulsion for freight locomotives as well.
3.3 Flywheel Energy Storage The idea of storing energy in the form of kinetic energy in a rotating inertia has been
known for centuries. Modern flywheels can be operated as part of a rotating electric machine and have the ability to compete with conventional electrochemical batteries 2,3.
A key objective of the energy storage system is to maximize the energy density of the storage system. The amount of kinetic energy stored in a rotating mass is directly proportional to its moment of inertia and to its rotational speed squared. Therefore, it is much more effective to increase the rotational velocity, rather than making a heavier flywheel with a larger diameter. Many flywheel energy storage systems are designed to rotate at speeds above 50,000 RPM, and most have a goal of exceeding 100,000 RPM.
In order to incorporate these high rotational velocities with an interface to an electric power system the flywheel is driven by a power electronic drive. The mechanical strength of the rotating assembly determines the peak energy storage, but the converter will determine how fast the flywheel can be charged or discharged; i.e., the maximum power transfer rate.
The flywheel can be used in transportation applications to store energy when the locomo-tive is resistive braking. This energy can then be used to accelerate the locomotive. So the inver-ter rating will need to be based on the peak energy transfer rate desired for braking and accelera-tion.
Another concern with flywheels for transportation is how the rotating mass of the flyw-heel effects the handling of the vehicle. This problem can be solved for high speed rail applica-tions by having many small counter-rotating flywheels on a locomotive.
3.4 High Speed Rail Corridors Ideally the high speed rail train sets will operate on a dedicated route where they won't
have to be scheduled around freight trains. However, this does not mean that the train simply ac-celerates to 120 MPH and stays at that speed until the next stop. A typical route, such as the Northeast Corridor or the Empire Corridor in New York will have a wide range of speed limits as shown in Figure 1. Each track section has speed limits determined by factors such as track curvature, grade crossings, slope, condition of track, and the area the track is passing through. Therefore, the train will vary its speed as it covers the route.
6
3.5 HigT
bine engier to prodtraction inerative btransfer pbus for cters and f4000 HP
h Speed TrThe high speeine for the prduce a DC binverters whbraking operpower betweases where tfor emergenc, and six pas
0
20
40
60
80
100
120
140
Spe
ed L
imit
(MP
H)
Fi
rain Set ed locomotivrime mover
bus, as shownich drive indration. The Deen the DC bthe braking ecy braking. Tssenger cars.
F
0
igure 1. Rou
ve being studthat powers n in Figure 2duction tractDC bus also bus and the fenergy transThe train be.
Figure 2. Loc
50D i
ute Speed Lim
died for simua synchrono
2. The DC bution motors. supplies mu
flywheel. A bfer rate exceing studied h
comotive Po
stance (M iles
mit Profile
ulation purpous alternatous connects These motor
ultiple flywhbraking resiseeds the capahas two loco
ower Train
100s)
poses consistor which in tthe prime mr drives are
heel invertersstor is still nacity of the fomotives, ea
150
ts of a gas tuturn feeds rec
mover to fourcapable of res, which can
needed on theflywheel invch rated at a
0
ur-ctifi-r ege-
n e DC
ver-about
7
4. Possible Optimizations There are several areas where performance of a high speed train that can be optimized.
The possible areas that can be optimized are total trip time and energy consumption over the course of the route. Further options become available when flywheels are added to the locomo-tive, especially in a system that utilizes 30 to 40 small flywheels. The decision to use many ver-sus a few flywheels depends upon the state of flywheel construction technology as well as the efficiency of scale. Inherent limitations of flywheels point to the conclusion that, from an energy management for optimization perspective, many smaller flywheels have a distinct advantage. This issue is discussed in greater detail in the following sections.
4.1 Prime Mover/Flywheel Sizing The addition of the flywheels provides several areas where the performance of a high
speed rail locomotive can be optimized. The first is in the power rating of the prime mover, which in this case is a gas turbine. Since gas turbines have a relatively narrow operating band for peak operational efficiency, the flywheel can be used to level the loading on the turbine to keep it in its most efficient state.
However, the addition of the flywheel provides the possibility for even greater savings. The flywheels can be charged any time the power demanded to keep the train at the desired speed is below the optimal output level of the gas turbine. This condition will occur when slow-ing for stops, slowing for curves, going downhill, and so on. The energy in the flywheels can be used to accelerate the locomotive to the desired speed more quickly. One of the limitations for the high speed rail locomotive is its ability to accelerate quickly when it is already running at high speeds, since it is already near the power limits of the prime mover, so there is little excess power for acceleration.
The flywheels can supplement the prime mover in these cases, allowing the train to ac-quire the legal speed more quickly and hence maintain time schedules. Taking this a step further, the flywheels could provide some if not all of the additional energy needed to accelerate at the higher speeds, allowing a somewhat smaller prime mover, one that is sized to just maintain the train at a set maximum speed. The flywheels can allow it to accelerate around this maximum or even exceed it for a short time.
4.2 Energy/Fuel Efficiency Another area where the flywheels will have a significant impact is in the area of optimiz-
ing the energy/fuel needs over the length the route. The rate of acceleration / deceleration be-tween the differing speed limits along the route can be programmed to obtain optimum fuel effi-ciency. Use of the flywheels to capture energy from braking improve upon this further.
However, the control scheme for the locomotive needs to be able to adjust as the train progresses down the route to adapt to any changes from the original plan. If the train is behind schedule, the goal will be to get to the next stop on time, or with minimal delay. In this case fuel efficiency is secondary to time, since many transit systems may be economically penalized from reduced ridership by adverse passenger attitude for late arrivals.
4.3 Flywheel Use In a locomotive with 30 to 40 small flywheels it will also be possible to choose how
many flywheels to use at a given time and which ones might be best to charge. Flywheels gener-ally work best as energy storage devices when they are spinning above roughly 50% of maxi-
8
mum rated speed. Since there are some rotational losses with flywheels at present, especially in the bearings, there are some steady-state losses present in the system. In addition, the power con-verters interfacing the flywheel to the DC bus will also experience losses.
Therefore, there are cases where a relatively small amount of energy is available to be stored in or retrieved from the flywheels thus allowing optimal sizing the flywheel storage sys-tem on the fly. It may be more efficient to only activate a few of the flywheel inverters instead of all of them. The route that the train will follow is known in advance, including speed limits, curve information, and grade information. Thus it is possible to manage the energy budget for the flywheels and choose which ones are discharging/charging at any moment.
For some routes it may even be possible to have several flywheels per inverter rather than one flywheel per inverter. The inverters are the dominant expense in the flywheel system, so us-ing one larger inverter rather than several small ones may be economically feasible. This inverter would then multiplex the flywheels in these cases.
9
5. Simulation The simulation discussed the computer modeling of a specified train scenario. The pur-pose of such a model is investigate performance sensitivities to independent variables such as track speed profile, passenger load, size of prime mover, amount of storage capability, and losses. Individual components used in this simulation have been scrutinized in published work to validate the results of the system model. Although only the general details are presented in this interim report, a description of the simulation and the models used will be included in the final report.
5.1 Model Description A software simulation was developed in MATLAB to model the locomotive propulsion
system as it travels the length of the route. The objective of this simulation is to develop a basis for performing the design and control optimizations described in the previous section. Therefore, the models for the power converters and the rotating machines represent input/output energy flow. The system energy flow is driven by the power requirements of the traction inverters, which are driving the motors to run the locomotive at the desired speed. The simulation consists of several modules, as shown in Figure 3. Additional details of these models are provided in Figure 4 through Figure 6 and in the following discussion.
EfficiencyControlModel
SpeedControlModel
PropulsionModel
LocomotiveModel
Spe
ed
Speed
Position
Route Data
Flywheel Power
TripTimeRouteData
CommandedSpeed
TractiveEffort
Flywheel EnergyStorage Capacity Energy
Consumed
Speed
Figure 3. Block diagram of train simulation
The propulsion model shown in Figure 4 uses information energy conversion processes and the commanded torque to the motors to compute the energy efficiency of the locomotive. This propulsion module has power efficiency models for the alternator, rectifier, inverters, trac-tion motors and other components. Each block described in Figure 4 represents the losses asso-ciated with that particular conversion process. Many of these loss components have already been quantified in other areas, especially for diesel locomotives 4. The rotor angular velocity input is obtained from the Locomotive model and the commanded torque input from the Speed Control model, both of which are discussed below. The FOC (field oriented control) model is used to si-mulate the efficiency of the torque energy conversion to train mass - velocity energy of the in-
10
duction traction motors . Each traction motor is assumed to operate under identical conditions. The outputs are the power input of the generator required to maintain the train speed and the ef-ficiency of the system. The generator uses a load follower regulator, producing the necessary energy demanded by the system to maintain the DC Bus voltage constant.
IM(FOC) VSI DC
BUS Rectifier Generator
TractionMotor
Velocityfrom
LocomotiveSimulation
Model
Torquefrom Speed
ControlModel
InputPower &Efficiency
Figure 4. Propulsion Model
The locomotive model illustrated in Figure 5, takes the tractive effort (real power) com-
mand from the speed control module and combines it with route data to determine where the train is on the route and to output that position and speed and acceleration to the other modules. The acceleration of the train is computed from the tractive effort force operating against the mass of the train. The train velocity and the position along the track is obtained from the two integra-tors. The velocity is fed back by a function of train drag which calculates the drag resistance ac-cording to the Davis 5 equations. The position output is used to generate the resistance generated from the track grade and curvature. The grade and curvature are obtained from a lookup table in the track file. The input to this model is a power command generated by a proportional plus integral (P-I) controller common to classical linear systems control theory.
The output from the locomotive simulation model are train acceleration, train velocity, and position of the train along the track. The velocity is fed back to the Speed Control model for closed loop speed control as well as the propulsion model for efficiency calculations. The posi-tion is used to obtain the desired train speed from the track profile database, which will be also fed back to the Speed Control model. The curve and grade resistance is determined from this po-sition and the track profile database. Losses associated with the track and train resistance are included in efficiency computations. The track profile database does not include elevation changes and hence there is no energy storage as a function of train position.
11
1/Mass 1/s
Drag
TractiveEffort
Acceleration
From SpeedControl Model
+
To SpeedControl Model
1/s
Speed
To SpeedControl Model
andPropulsion
Model
Speed
PositionTo
EfficiencyControlModel
Curve
Grade
+
+-
Figure 5. Block diagram of the Locomotive Simulation Model.
The speed control model determines the tractive effort command needed for the train to run at the requested speed as specified by the train position and the track profile database. The control algorithm is a combination of rule-based control as well as linear proportional plus integral control. A block diagram for the speed control model is shown in Figure 6. The two li-miters control the trains acceleration for passenger comfort and safety.
P IDeltaPowerLimiter
da/dt
PowerLimiter Torque
TractiveEffort
Traction motorVelocity
To PropulsionModel
ToLocomativeSimulation
Model
FromEfficiencyControlModel
SpeedCommand
FromLocomotiveSimulation
Model
FromLocomotiveSimulation
Model
AccelerationSpeed
fromLocomotiveSimulation
Model
Figure 6. Block diagram of the Speed Control Model
12
5.2 Results The results of software simulation can be used to aid in the design of the locomotive it-
self. The required power rating for the prime mover can be determined, as well as both energy storage capacity for the flywheels and the power rating for the inverters transferring energy into and out of the flywheels. Once this task has been completed, the simulation is next used to final-ize the overall control scheme for the locomotive system. Figure 7 through Figure 9 show some initial results from the software simulation. Figure 7 shows the speed set point for the locomotive as compared to the speed limit. The speed setting needs to plan ahead to make sure the speed is always at or under the speed limit. Figure 8 compares the actual locomotive speed on a section of track to the speed limit on that section. The speed must always be less than the speed limit on a section of track, requiring the train to slow before it reaches it.
The fundamental objectives is to determine the most energy and time efficient profile for accelerating and decelerating. The contribution of the energy storage provides many additional degrees of freedom in this optimization. Figure 9 shows the total power delivered to the traction motors as the train moves down its route. Notice that there are a few high peaks and there are also areas where the net power is negative due to regenerative braking. The flywheel energy sto-rage could provide the power needed to meet those peaks, allowing a smaller prime mover to be used to meet the “base load.” More detailed results will be presented in a future paper after the completion of the project.
0
20
40
60
80
100
120
140
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150
Track Position (miles)
Velo
city
(mph
)
track limit
command
Figure 7. Commanded Speed Compared to Track Speed Limit
13
0
20
40
60
80
100
120
140
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150
Track Position (miles)
Velo
city
(mph
)
velocity
track limit
Figure 8. Actual Velocity Compared to Track Speed Limit
-10000
-8000
-6000
-4000
-2000
0
2000
4000
6000
8000
10000
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150
Track Position (miles)
Pow
er (h
p)
Figure 9. Net Power To Traction Motors
14
6. Demonstration At the time of this interim report, the demonstration system is still under construction at the University of Idaho, in Moscow Idaho. The final phase to be completed is the integration of the flywheels into the rest of the system. All other major components have been tested and inte-grated at this time.
6.1 Discussion The demonstration system will be presented in three parts; the computer controls, the power conversion electronics and machinery, and the performance results. Discussions of the computer controls all include the analysis of networked controls as compared to conventional direct wired controls, a review of control network technologies and the criteria for selecting a control network technology.
6.2 Computer Controls
6.2.1 Philosophy of Control Communications The locomotion controls regulate the train speed. This is accomplished by transferring
power from the prime mover, whether that diesel, turbine or catenary to kinetic energy mani-fested as train mass multiplied by the square of train velocity. Classically, trains decrease speed by transferring the kinetic energy into heat either with friction brakes or dynamic braking. Re-newable energy system on board allow prime mover peaking or recovery of energy usually lost as heat when the train is decelerating.
Control wiring, generally speaking, connects the origin of decisions to the point of deci-sion implementation. Instrumentation wiring brings the measurement of state back to the point of analysis and decision generation. Hence, the network implementation would in the strictest sense merely replace point to point control and instrumentation wires. The control scheme as-sumes the following design philosophies:
1. High speed equipment and personnel protection controls are independent of the ener-gy management controls.
2. The level of system reliability for proposed systems must be equivalent to or better than conventional controls
3. All control equipment (conventional and proposed) has been pre-qualified for equal or better reliability specifications. Hence any new electronics is assumed to have comparable reliability in so much as the degree of complexity is comparable.
4. A bandwidth maximum of 1Khz is needed for any control and instrumentation sig-nals.
Using networked distributed control for the demonstration revealed two important con-clusions; considering the total amount of computer and control electronics needed for the system, the network technology contributes minimal additional electronics and hence no cost advantage is held by either direct control wiring or networked control. The first conclusion is based upon the available of high performance microprocessors with the network electronics packaged to-gether. The second conclusion arises from the minimal amount of control and instrumentation needed to achieve the necessary degree of control.
The incremental cost for equipment to control and monitor flywheels or some other re-newable energy source will be a very small percentage of the total installed cost of such a sys-tem. The exception to this claim is if many (more than 10) flywheels are used in a single train
15
system. One may then desire that each flywheel energy storage unit be individually controlled. This would then have some impact on the complexity of the control wiring. The UI demonstra-tion control scheme presently uses three or less signals per power device. For example, each traction motor and flywheel requires one control signal and two instrumentation signals for in-verter management for traction motors and flywheels. Controlling the power plant requires one control signal and two instrumentation signals. The number of controls required for dynamic braking is determined by the amount of brake control.
6.2.2 Analysis of Networked Control and Instrumentation This analysis focuses on the comparison of distributed network controls and conventional direct wired controls which use a single central computer. The comparison is based upon func-tional capability, reliability, and cost. The discussion is intended to provide a procedure for se-lecting control technologies. Based upon the importance of key issues, the same procedure may result in selecting different but still viable technologies. The network will be limited to support-ing only those controls suggested by the simulation and those actually needed for the speed con-trol and energy management. Fault protection controls and regulatory instrumentation are not included in this analysis as the investigators are convinced that those systems should be kept en-tirely separate and autonomous for reliability and security.
6.2.2.1 Communications Reliability Three aspects of component reliability are considered: intrinsic failure rate, operational
failure rate, factors that accelerate intrinsic failure rates. Intrinsic failure rates are manufactured into devices and systems and are independent of operations. Operational failure rates can be de-termined by decisions that place the equipment in abnormal operating conditions. But more of-ten, ordinary use only increases the probability that components will be exposed to a failure rate accelerator.
The three dominate accelerating factors are heat, electrical stress, and interference. Heat may be either self generated or ambient. Elevated ambient heat is generally caused by inade-quate engineering or the failure of another component or system. Electrical stress is either sus-tained or transient. Sustained electrical stress is either caused by poor engineering or a failure of another component or system which perpetuates a sustained over or under voltage situation. Voltage transients are generally initiated by a high rate of change of inductor current caused by switching contactors or other electro-magnetic circuit. Voltage transient are capacitatively coupled into neighboring circuits Current transients are generally caused by switch voltage into uncharged capacitors and are inductively coupled into neighboring circuit.
Either voltage or current transients can cause equipment failures, but more often these transients generate interference on instrumentation signals. Interference is usually thought of as an intrusion that degrades the quality of information. The results of which can be failure of a system not because of any component failure but because the processing of the corrupted infor-mation yielded an erroneous result perpetuating an incorrect action.
System engineering is challenged with two design goals; choose components and systems with low intrinsic failure rates and integrate them into a system with the lowest probability of exposure to a failure rate accelerator. Assuming that the first goal is already achieved, we look at networked controls to strive for improvement in the second. In order of ranked importance, each method of control communications is analyzed from the perspectives of; 1. reliability, 2. cost, and 3. adaptability to future modifications. Reliability assessment is performed on system
16
differences recognizing that either networked or direct point to point control wiring requires common computers, instrumentation and control equipment. This wiring to achieve the neces-sary communications is next discussed to illustrate where improvements are obtainable.
6.2.2.1.1 Reliability of Conductors and Terminating Hardware Conductors and connecting hardware are considered because these devices have low re-liability and high cost. They are required for both direct wired control as well as networked con-trol but the manner in which they are used can significantly affect reliability. These issues are discussed in the following paragraphs.
The chief cause of insulated conductor failures is insulation breakdown. The reliability of conductors is therefore inversely dependent on the electric potential stress caused by the cir-cuit voltage. The reliability of connectors that terminate the control circuits is inversely depen-dent upon heat. One source of this heat is the I2R losses in the terminating device. The reliability of both conductors and terminating devices have an inverse time dependency as well that is sepa-rate from the probability random failures. Circuits that operate continuously have higher failure rates than those operated intermittently. Being physical devices, the reliability of terminating devices are also inversely affected by the number of connecting and disconnecting operations due to mechanical wear.
Three classifications of control circuits are investigated: low power, medium power and high power. Low power circuits refer to signals that operate under 10 watts at voltages less than 100 volts AC or DC and currents less that 1 amp AC or DC. Medium power circuits operate be-tween 10 watts and 100 watts with voltages less than 500 volts AC or DC and less that 10 amps AC or DC. High power circuits operate above 100 watts at voltages greater than 500 volts AC or DC or currents in excess of 10 amps.
Low power circuits can use multi-conductor bundled cable and are within the capability of standard low and medium power semiconductor devices. Heat generated in the control of low power circuits is easily mitigated and has little effect on reliability. Electrical transients generat-ed during control operations can be easily managed by commercial and industrial grade semi-conductor devices. Control of low power controls are suitable for direct computer control with reasonable attention to engineering design.
Medium power circuits require special transient protection and heat management. Devic-es are more expensive and greater engineering expertise is needed or reliability will be signifi-cantly impaired. For direct computer control of medium power circuits, significant attention for engineering design is required.
High power control circuits are not suitable for direct computer control. In train control environments, these controls generate high energy RFI and EMI noises which are difficult to contain. The longer the control circuits the more severe the noise problems. Usually a low or medium power control circuit interfaces computers to high power circuits.
6.2.2.1.2 Signal Reliability For a given control system with a fixed number of points of control and instrumentation,
the number of conductors and terminating devices are only slightly more for networked control wired systems. However the length of medium power circuits can be significantly reduced by being replaced with low power circuits. This reduces the probability of EMI and RFI interfe-rence which inherently improves reliability.
17
Networked control wiring can offer significant advantages in validating information. If the network is engineered with the expectation of periodical communications, the lack of com-munication contains information of failure. Corrupted data is detected with a highly level of confidence using error detection codes.
Direct wiring offers little or no opportunity to qualify information particularly if the noise is indistinguishable from valid signals. For example, a digital signal can be represented by zero volts which indicates a logic zero, and the presence of a voltage above a fixed threshold. A receiver cannot distinguish the zero volts sent as a valid control signal from zero volts due to an open circuit. Analog signals have a degree of signal qualification if the zero point is excluded from the range of valid data. Such is the case for 4-20 mA constant current analog circuits. However, if interference added or subtracted current from the signal such that the resultant cur-rent is in the valid range, the signal error is indistinguishable from an uncorrupted signal.
Another means of discriminating information from noise is by frequency domain filter-ing. The information must be characterized to determine the signal’s center frequency and bandwidth as well as required signal to noise ratio. The circuit must then be characterized to de-termine the noise magnitude and bandwidth. The type and degree of filtering may be determined if sufficient frequency separation exists between the information signal and noise. This filtering may be analog, digital or a combination of both to achieve the desired signal to noise ratio. If filtering cannot provide the necessary signal to noise ratio, the circuit will have to be modified to reduce the amplitude of the interfering noise.
6.2.2.2 Cost Conventional train control wiring schemes will be used as the base case. Only the network data electronics will be considered since one of the assumptions is that the amount of control electronics will be approximately the same for either direct wire or networked controls. The in-cremental cost of the additional network management electronics will be considered.
6.2.2.2.1 Installation Cost
6.2.2.2.1.1 Cables No specialized proprietary equipment is necessary for terminating the network cable. Since no particular handling care is required to install network cable, the cost of doing so is strictly a function of the number of cables. To achieve proper impedance matching electrical termination, CAN networks use a linear configuration rather than a star or ring configuration. Termination connecting devices may be simple and inexpensive as quick disconnect connectors are not re-quired since the network connection would only be broken for repairs.
6.2.2.2.1.2 Data Communications Electronics The cost of CAN electronics is less than $50 per node. This electronics can be a separate IC such as the Intel 82527 CAN controller or the 80196CA processor which integrates the 82527 functionality on the silicon of an Intel 80C196KR type processor. The processor and data com-munications electronics will need to be packaged in conformance with conventional electronics. Usually all electronics are specified for industrial temperature range operation. ESD protection is also required on all interfaces to wires outside the computing environment.
18
6.2.2.2.2 Maintenance Cost Maintenance of networked controls will require sophisticated test equipment and very know-ledgeable technicians. This is in contrast to electronic digital multimeters (DMM) and analog oscilloscopes which may be the only test equipment required for direct wired controls. Such ad-ditional equipment to support networked controls may include network analyzers, digital storage oscilloscopes, and CAN protocol analyzers. The cost of this equipment may exceed $25,000. The larger problem that faces the rail industry is to identify technicians who have the knowledge and aptitude to use the sophisticated equipment and to interpret the results to locate failures quickly and economically.
6.2.2.2.2.1 Cables Since the signals used on the CAN twisted pair cable constitute them as low power circuits, an improved cable and connector reliability can be expected. If the data rates are pushed to 1M baud, specialized network analyzers as discussed above maybe required to characterize the cable necessary to pin point physically damaged cable.
6.2.2.2.2.2 Data Communications Electronics Field repair of electronics is usually at the board level replacement. Shop maintenance will be limited to the degree of easily removable parts. Placing IC’s on sockets reduces maintenance but conversely degrades reliability. Printed circuit boards using surface mount technology with VLSI (vary large scale integration) integrated circuits are not field repairable. Seldom is it eco-nomical to repair these printed circuit boards at any level and hence maintenance will be reduced to board level replacement. The objective then becomes to use sub assembly printed circuit boards that are sufficiently inexpensive and reliable to make the board level replacement eco-nomical.
6.2.2.3 Adaptability to Future Modifications Network controls offer a distinct advantage over direct wired controls when new equip-ment is to be added. The cost of adding control wire is minimized requiring only the routing from the point of instrumentation or control to the nearest point of network interface. There are two caveats to the previous claims; if the extension is sufficiently long to affect the impedance of the network cable and if the information traffic density is to the degree that the additional traffic will adversely affect the train system performance. In case of the long network extension, the impedance mismatch can be mitigated by use of a network repeater or bridge. For the case of the network traffic density being to high, this is actually a network band-width or bit rate problem. There are no low cost solutions but there are three relatively high cost solutions: increase the bit rate (assuming that there is sufficient bandwidth remaining), add paral-lel networks, and use data compression. All of these solutions involve adding hardware and-or changing firmware. The question that remains is whether the incremental cost of increasing the network bandwidth is more economical than adding a direct wire control or instrumentation cir-cuit. The answer may be unique to each case.
6.2.2.3.1 Life Cycle of Trains Trains that are expected to operate continuously for many years usually remain servicea-ble past many generations of computer advances. Certainly the life expectancy of the electronics far exceeds that of the train’s mechanical system. It is not recommended that the trains electron-
19
ics be upgraded before the rest of the train plant unless there is some extenuating regulatory, safety, or significant performance motivation.
6.2.2.3.2 Total Rebuild When economics dictates that a total train rebuild is required, the control equipment represents a small portion of the total cost. The control system should be designed to maximize the performance of the new power plant and include provisions as dictated by the FRA for safe-ty. Due to the rate of advances in 1997 of computer systems performance and reliability, it is recommended that controls systems to be put into new or train control upgraded system be no older than three years old.
6.2.2.3.3 Modular rebuild For controls designed with excessive bandwidth and processor capability, some
systems of the train such as the hotel power, breaking, or wayside control can be upgraded and incorporated into the existing computer controls. Designing unused bandwidth and computer capability results in an investment that may not be recovered. The problem is further compli-cated by the inability to accurately predict how much excessive capacity will eventually be used before economics dictates that the entire system be upgraded. The nature of networked control systems tend to be more distributed. than direct wired controls which lend itself to using a single central processor. Distributed controls place the intelligence in the new equipment and make smaller demands on the supervisory processor. This results in a lower requirement for excess capacity in the supervisory processor for networked controls. Certainly, it is easier and less ex-pensive to build slack bandwidth into networks controls than additional unused cables for direct wired controls.
6.2.3 Review of Network Technology Conventional information networks are usually concerned with only the magnitude of the data. Control network are designed to communicate information that has both magnitude as well as time attributes. The time quality of data makes information invalid if it is too “old” to gener-ate reliable control decisions. Control networks invariably are limited to small data packets to minimize the probability of data collisions on the network. Various technologies were investi-gated to determine their applicability to train controls. As discussed above, the network technol-ogy must be capable of data rates in excess of 300Kbps. Table I is a listing of the comparison of control networks technologies considered.
Table I.Comparison of control network technologies
Technology Data transfer rates
Message size - max.
Sources Physical Layer
Industry Focus
1 Token ring (Arc-net)
156Kbps - 5Mbps
2048 bytes SMC
Coax TW FO
Industrial control
2 CEBus 10 KBps 32 bytes Intellon PLC RS485 RF
Home and commer-cial bldg. automation
20
3 parvNET ? No limit Parvus RS485 PLC FO
Home automation
4 Echelon 5Kbps - 1.25Mbps
255 Bytes Motorola Toshiba
PLC Coax RF TW FO
Industrial control - Building and home automation
5 ATM 622MBps -2.4GBps
48 Bytes Various SONET Tele-comm.
6 CAN(J1939) 5Kbps - 1.0Mbps
8 Bytes Intel Phillips Motorola
RS485 TW FO
Automotive - Ground transportation
7 I2C Microwire SPI
1 Bps - 100KBps
256 Bytes Motorola Phillips National (Any uP)
TW Inter PC Board communications
8 Summit MIL-STD-1553
? ? UTC FO Coax
Military- aircraft space
9 SCRAMnet to 10GBps ? Systran FO Coax
Military - space
6.2.3.1 Network Technology Requirements
1. The network must have a very low probability of an undetected transmission error. 2. Nodes entering and leaving the network will have zero impact on the communications
traffic. 3. Minimum network management
4. Suitable speeds for real-time control for systems whose bandwidth is up to 1khz. 5. The cost of the network equipment must be less than $1K per node. 6. The vendors must exhibit or demonstrate a reasonable degree of customer support ca-
pability 7. The processor overhead to support the network communications must be less that 20%
of the total processor resources.
6.2.3.2 Evaluation of Network Technology
From the above criteria, Table II identifies those networks still under consideration.
Table II. Evaluation matrix for selecting network technology.
Technology Fails Network criteria
Fails Processor criteria
Eliminated Selected
1 Token ring (Arcnet) 1 X 2 CEBus 4 X 3 parvNET
21
4 Echelon 3,4,6 I.A - I.G III
X
5 ATM 6 CAN(J1939) X 7 I2C
Microwire SPI 1 X
8 Summit MIL-STD-1553 5 X 9 SCRAMnet 5 X
The final selection for the control layer network depended upon the speed requirements
to support the bandwidth of the process. Initially the UI group looked at the requirements to support vector control. This type of induction motor speed control requires sample rates on the order of 70K samples per second. Even the processing requires a high speed DSP type processor (either TI DSP32C5X or Motorola DSP5600). Although there exists networks that support the data transfer rates required for vector control, the expense far outweighs any advantages of net-working the shaft position encoder to the controller.
The next area of control is the transient stability of the DC bus and the interaction of the masses of the of the locomotive, rotor inertia of the traction motors, and the mechanical inertia of the flywheels. In addition to the mechanical dynamics, the transient dynamics of the DC-AC converters add additional complexities. These dynamics are currently being studied to determine the speed requirements to achieve the necessary control.
Network technologies supported by IC’s with a high degree integration offer low cost and minimal processor overhead. Hence when selecting processors for the energy management con-trol (not the power electronics control which has been already determined to be a high perfor-mance DSP). The selection criteria favors those processes which integrate networking proces-sors on the same silicon substrate as the numerical processor. The following list of processor cri-teria are added to the table above to visually assist in making the final determination. 1. High degree of on-chip resources for embedded applications 1.a. High speed I/O and EPA 1.b. Counters 1.c. Multiple serial I/O channels 1.d. On chip AD conversion 1.e. On chip PWM 1.f. Interrupt management 1.g. High computational performance index. 2. Low processor overhead 3. Low cost development tools
4. Reasonably supported by industry 5. Low cost (<$100)
6.2.3.3 Selection of Network Technology The 87C196CA processor which directly supports the CAN network technology was
chosen because it meets the above qualities as well as the following additional reasons: 1. CAN meets the data rate requirements 2. The development tools were relatively inexpensive 3. CAN is a mature technology in the automotive and industrial control areas.
22
4. The designers were familiar with the 8X196 family of processors and the develop-ment tools.
6.2.4 Description of CAN Technology The CAN protocol as specified by ISO/DIS 11898 is widely used in automobiles and
freight trucks. CAN is also used for the industrial control using proprietary applications layers. One such provider is Allen-Bradley control network called DEVICE-NET Of the seven level OSI-ISO network model, CAN directly supports levels zero, one and seven which is common for control networks design to meet stringent time requirements and have relatively small data packet.
Since the data efficiency for CAN is only 50%, the data rate must be 2Khz per control or instrumentation channel for the two methods to be comparable. Thus for a 50 channel system, the data rate would need to be equal to or greater than 300K baud. This represents a conserva-tive estimate based upon the discussion above for the number of instrumentation and control sig-nals for traction motors and flywheels.
If a CAN interface IC such as a Phillips PCAA82C250 is used, the digital network can operate up to 1M baud depending upon the length of the contiguous network wire. RS485 can also be used as the interface to the physical layer but is subject to more restrictive data speed - network length constraints with a maximum data rate of 250K baud. The physical layer for CAN is a shielded or unshielded twisted pair cable.
6.2.5 Energy Management Organization Figure 10 illustrates the computer information flow to implement the energy management scheme for distributed control of the train’s sources and uses of energy. Figure 11 illustrates the hardware organization use to implement the management scheme. For the train system under consideration, there are three sources of energy, the DC alternator powered by the prime mover, the energy stored in the fly wheels and the kinetic energy of the moving train. Given that the track route ignores elevation changes, energy can be absorbed or stored in two forms; accelerat-ing the train or accelerating the flywheels. Energy is always lost in the form of heat from this closed system either intentionally or unintentionally. Unintentional losses result because of inef-ficiencies in the power conversion equipment and mechanical friction. Intentional losses occur when brakes are used to reduce the train’s velocity by using either dynamic brakes or friction brakes. The DC bus in the electrical system is the energy conduit between these sources and uses of energy.
23
EnergySupervisor
KnowledgeRules
NetworkInterface
GraphicalUser
Interface
TrackDataBase
Operator
PowerSourceNetworkControl
TractionMotor
NetworkControl
FlywheelNetworkControl
GeneratorField
Control
DC BUSSensors
TractionMotorsDSP
RotorSpeedSensor
FlywheelDSP
FlywheelSpeedSensor
DynamicBrake
Figure 10. Information relation diagram for distributed networked control.
SyncGen
InductionMotor
LoadControl
LoadControl
TractionMotor
Fly
Whe
el
Fly
Whe
el
240VAC
DC BusInstrumentation
DynamicBrake
andField Control
Load SimulationComputer
EnergyManagement
Computer
Energy Management Demonstration Hardware Orginization
Train Physics Simulation
Tach
omet
erIn
vert
erC
ontr
olle
rDC/AC
Tach
omet
erIn
vert
erC
ontr
olle
rDC/AC
Tach
omet
erIn
vert
erC
ontr
olle
rDC/AC
Tach
omet
erIn
vert
erC
ontr
olle
rDC/AC3 PhaseRectifier
Direct Wire Control
Control Network
Direct Wire Control
ControlNetwork
ControlNetwork
Figure 11. Demonstration train system control and instrumentation block diagram.
24
The energy supervisory computer (PC) manages the knowledge rules stored in database form and processes user requests. The user controls can be initiated either by a local operator using a user interface similar to Figure 12 or a remote telemeter control signal from wayside sen-sors and dispatchers. This information is communicated to the distributed generators and con-sumers of energy via the control network. Each network control processor communicates with a DSP (digital signal processor) controller via a point to point 38.2Kbaud RS232 serial data com-munications channel. The DSP processors are needed to implement the field oriented control (FOC) used to control the rotational velocity of for the induction traction motors and the flyw-heels. Other network processors directly control the DC bus voltage and the dynamic brakes us-ing amplifiers and power switches.
Figure 12. Energy management control LabVIEW display panel
6.2.5.1 General Network Hardware Overview As illustrated in Figure 11, eight processor boards make up the Controller Area Network
(CAN); six of which are based on the i80C196CA processor and two on the i87C196KD proces-sor. Software on one i87C196KD processor board controls the DC bus voltage while the other simulates the track load by controlling the load motors. All processor boards have an RS232C serial interface operating at 38.2KB and a CAN interface using the RS485 physical layer.
The i87C196CA processor boards are produced by Dearborn Group in Dearborn, MI. The systems are supplied with Intel RISM (Reduced Instruction Set Monitor) ROMS and were used while the application programs were under development. The i87C196KD processor board is produced by Intel and also supplied with a RISM ROM base monitor. These processor boards
25
interface with the CAN using an i87527 CAN evaluation board. After development is completed, the RISM Monitor ROMS are replaced with application specific programmed ROMS as well as programmable logic devices. The applications code was developed using IAR ICC196 version 5.1 and CodeVew IDE (integrated development environment) from Chip Tools,. Inc.
The program contained on the MCS96 - 87196CA processor board supports three classes of applications programs; the Master Controller, Inverter Controller, and the Dynamic Brake Controller. The particular applications running on the i87C196CA processor boards are selected by settings on the controller boards. LED lights on the controller boards provide processor and network status information.
6.2.5.2 CAN Message Structure The CAN uses a message based communications scheme that requires no network man-
agement resources or time when nodes enter or exit the network. The primary information iden-tification comes from a message ID and does not identify either the source or the destination of the message except by specific design using fields allocated inside the message data area. Only the information itself is identified by a 28 bit word unique to each message type. The format of each message is shown in Figure 13. The message types specified in SAE standard J1939 assign variables to each data byte in the message data filed.
Figure 13. CAN message format
Message data lengths are from zero to eight bytes. There are three instances that initiate network traffic: unsolicited, request for data, and response to request for data. The unsolicited messages are initiated for system exceptions and by the master controller as control messages. The master unit requests new data on timed intervals by sending an abbreviated message that contains the message identifier and a single bit that signals other units on the network that this is a request for data. Units with this information immediately respond with the appropriate data message.
All network arbitration and error management is handled by the CAN controller IC’s ei-ther the Intel 82527 or the 87C196CA internal CAN controller. Each CAN transmission uses bit stuffing for synchronization, 16 bit CRC for error checking, and message prioritization for colli-sion avoidance.
6.2.6 Information Communications The decision center for the networked control and instrumentation is a PC that runs the Windows 95 based MATLAB and LabVIEW applications programs. The real time operating system works as illustrated in Figure 14. Most transactions are implemented by flow of informa-tion described by the right most column. The LabVIEW program serves as the presentation in-terface with the operator. The MATLAB program functions as the dynamic control program and has access to the track profile database. Information is exchanged between LabVIEW and the
26
network master processor using a text data stream over a 38.2KB RS232 serial port. Information is exchanged with the MATLAB program using a data file serving as a mail box. LabVIEW posts system state information in the mailbox file before invoking the MATLAB program. While the MATLAB program is running, the LabVIEW program is suspended. MATLAB reads the mailbox data file, performs the appropriate computations, incorporates data from the track profile database, and posts the resulting new commands back to LabVIEW in another mailbox file. Upon termination of the MATLAB program, control is returned to the LabVIEW program for display and serial data transmission to the network supervisory controller.
Real-Time OSEvent
Processor
UnsolicitedCAN
MessageUnsolicitedSupervisory
Message
TImeEvent
Poll CANNodes
Communicateupdates toSupervisor
UpdateCAN
Nodes
Communicateupdates toSupervisor
Poll CANNodes
Communicateupdates toSupervisor
LabVIEWupdatesdisplay
LabVIEW"mails"Matlab
system state
Matlab "mails"LabVIEW new
controls
LabVIEWupdatesdisplay
Communicatecontrols
via network
Figure 14. RTOS flow diagram
Although CAN accommodates peer-to-peer network communications, this implementa-tion currently has the master unit managing the network information. The master node gathers status information from the network and communicates the system state with the supervisory PC over a 38.4KB RS232 serial channel. Although using ASCII text characters requires greater communications time than sending the information as binary data, ASCII allows data to be quali-fied quite easily and is very beneficial during development phases. The PC is expected to oper-
27
ate on this information and generate new commands in a timely manner. When the new com-mands are received by the master unit, the information is disseminated over the network as spe-cific messages to be implemented at remote locations.
Each second, the master network node polls the slave nodes for updated status informa-tion passes this information to the supervisory PC and the master waits for new commands form the supervisory PC. If there is no response from the supervisory PC within one second, no commands are sent out on the network but the master again polls the network slaves and updates the supervisory PC. If no commands are received by the master network node for a period of five seconds, the master node initializes all control variables and communicates the reset condi-tion to all slave nodes. In effect, if the master network looses communications with the supervi-sory PC, the train is, in a manner of speaking, stopped dead in its tracks.
Any i80C196CA board can be set to be the master network node by setting switches on on the board. While running in the master mode, the serial communications to the supervisory PC contains 14 space delimited values using the format shown in Table III. The serial string is always 69 characters long including the terminating CR-LF sequence. All values are four de-cimal digits with leading zero suppression, except for the eleventh value which is a single decim-al digit. The serial communications operate at 38.4KB, eight data bits, no parity, and one stop bit.
Table IV describes the messages from the supervisory PC to the master controller. Six control variables that are passed manage the energy production, use and storage for the train. To reduce communications overhead, only positive variables are passed. Present convention uses a range of zero to 2000 as nominal control parameters. This represents -100.0% to +999.9% for all converter controls. The DC bus voltage has a range of 0 to 1000 although it will be nominally set for 200VDC.
Table III. Master to PC data format.
Digit Parameter Range Min Max
1 Traction Motor #1 Torque 0 2000 2 Traction Motor #1 Speed 0 2000 3 Traction Motor #2 Torque 0 2000 4 Traction Motor #2 Speed 0 2000 5 Flywheel #1 Torque 0 2000 6 Flywheel #1 Speed 0 2000 7 Flywheel #2 Torque 0 2000 8 Flywheel #2 Speed 0 2000 9 DC BUS Volts 0 2000
10 DC BUS Amps 0 2000 11 Dynamic Brake Control 0 1 12 Train Speed 0 255 13 Train Distance modulo 10K feet 0 9999 14 Train Distance / 10K feet 0 9999
Table IV. PC to master data format
Digit Parameter Range Min Max
1 Traction Motor #1 Power 0 2000 2 Traction Motor #1 Speed 0 2000 3 Traction Motor #2 Power 0 2000 4 Traction Motor #2 Speed 0 2000
28
5 DC Bus Voltage 0 1000 6 Brake control 0 1
The master unit is configured to receive eight different CAN messages with each mes-
sage containing eight bytes of data. Figure 15 illustrates the network configuration and the mes-sage routing. Messages from the slave to I/O nodes are either “request for status data” or “com-mand” messages. Messages from the I/O units are all “status” messages in response to the “re-quest for status data”. Six units receive “commands” from the master unit; the two Traction Converter control units, the two Flywheel Converter control units, the Dynamic Brake control unit, and the DC bus voltage control unit. The track position and train velocity data is provided by the Track Simulator unit.
CAN Network
IBM Compatible
ServerUnit
Track Simulator
Traction Motor #1Controller
Flywheel #1Controller
Flywheel #2Controller
Traction Motor #2Controller
DC BusController
Dynamic BrakeController
3276
9
3277
0
3277
1
3277
2
65256
65248
6527
465
271
Figure 15. Demonstration train network with J1939 message routing.
Table V describes the message content for each of the message type used in the system. As the bidirectional arrows in Figure 15 imply, control messages and status messages between a given unit use the same message number. For example, if the master unit needs an update from the Traction Motor #1 unit, a “request for data” is made on the network for message number 32769. Traction motor #1 unit then responds with it’s data using message number 32769. When the master unit issues a command to Traction Motor #1 unit, it does so using the same 32769 message number. Another unit monitoring network traffic that is set to receive message 32769 would be able to distinguish the command data from the status data since the different informa-tion occupies different fields in the message structure as Table V describes. The CMD- STATUS column indicates the information source. As stated above, the source of command information (C) is the master unit while status information (S) is generated by I/O units. Message data with unidentified sources are not used in this implementation. Although the master unit controls the network traffic in a master-slave fashion, peer-to-peer and unsolicited slave to master communi-cations is not precluded by hardware or software.
29
Table V. Network message data formats.
Number Message name Bit Position
Message data CMD Status
Data Type
32769 Inverter Bit 0 Destination Unit Byte -72 Torque Bit 1 Source Unit Byte
Control Bits 2-3 Unit speed S Word Bits 4-5 Commanded Power C Word Bits 6-7 Unit Torque S Word
65248 Train Distance Bits 0-3 Trip distance - feet S Long Bits 4-7 Total Distance - feet Long
65256 Train Speed Bits 0-1 Direction Word Bits 2-3 Ground Speed S Word Bits 4-5 Track Pitch Word Bits 6-7 Track Altitude Word
65271 Train Electrical Bits 0-1 DC Bus Amps S Word Bits 2-3 Alternator Current Word Bits 4-5 Alternator Volts C Word Bits 6-7 Bus Volts S Word
65274 Dynamic Brakes Bit 0 Brake Application C S Byte Bit 1 Brake Pressure #1 Byte Bit 2 Brake Pressure #2 Byte Bits 3-7 Unused NA
6.2.7 Observations on the Energy Management Code
6.2.7.1 LAbVIEW Development Issues The interface between LabVIEW and MATLAB proved to be poorly documented and cumbersome. After discussions with technicians at both LabVIEW and MATLAB, it became apparent that, although the sales literature claimed interface was possible, no one had done it previously. The interface implemented with the mailbox system discussed above does not start easily. First the MATLAB control program is launched which results in many errors reporting unidentified variables. These errors are to be ignored as the variables are supplied by the Lab-VIEW applications program. The LabVIEW program is launched and this then waits for a serial stream from the CAN processor. Normally, the entire system timing is managed by this CAN controller. That is to say, until the CAN network supervisory controller initiates a status update communications, both the LabVIEW and MATLAB programs are idle. No information gets sent on the network until invited by the CAN network supervisory controller. The serial interface was manageable but the LabVIEW code is neither as elegant nor as simple as most asynchronous point to point communications tend to be. Not withstanding the complexity of implementing simple text based bi-directional serial communications with Lab-VIEW, this interface seemed suitably stable as long as the number of characters in the data stream remained fixed. The serial interfaces on the CAN controllers makes use of code available
30
on the Intel FTP site for embedded controllers. This required little effort to implement and proved robust. The iconic nature of the LabVIEW symbols does little to improve documentation. The level of detail for needed for process interconnection is inconsistent with information conveyed by enigmatic icons. Comments that usually aid in software documentation are difficult to incor-porate. Programs are difficult to write and even more difficult to analyze and manage unless one is extremely well versed in using National Instruments “G” graphics language. Displays are easily generated with LabVIEW. The control display panel shown in Figure 12 is entirely comprised from the LabVIEW library instrumentation and controls.
6.2.7.2 LAbVIEW Operational Issues When using Windows 95 for real time control, the operations systems determines the
priority of service and to management other computer resources such as disk and RAM data sto-rage. While the MATLAB and LabVIEW applications were running, it was observed that the time between the status update from the CAN master and the response of new commands to be distributed to the slave controllers varied from five to fifty ms. The level of process security can only be guaranteed if the
6.2.8 Route Simulator and Track Load Simulator Processor The route simulator controls the load motor-generator (M-G) set to simulate the train curve and grade resistance as a function of position. The track simulator and its associated route profile data are implemented on the Intel 18MHz 80196KD microprocessor. This unit receives no commands from either the master network node nor the supervisory PC although the position and train ground speed is communicated supervisory PC via the CAN network. It is representative of information obtainable from a GPS (global positioning satellite) system . The only input to this controller is traction motor angular velocity. The functional block diagram for the track load simulator shown in Figure 16 consists of the Intel 80196KD microprocessor (which includes the Route Data Table in memory), the DC Drive Unit, the Scaling and PWM Conditioning Circuits, and the M-G unit.
6.2.8.1 Track Simulator Basic Operation: As shown in Figure 16, The shaft of the traction motor is mechanically coupled to the ro-
tor shaft of the DC motor load. The torque produced by the motor load is proportional to the field current that is regulated by the load simulator microcontroller. Rarely, if ever, does the mo-tor load ever turn in the direction of the torque applied by the field current. The motor load simply opposes the rotation of the simulated trains traction motors.
The torque meter on the shaft the couples the traction motor to the load motor measures the algebraic sum of the torque from the DC motor and the traction motor torque. The processor controls the field of the DC motor to generate a counter torque that is generated by the resis-tance loses due to curve and grade as well as drag losses. If the torque generated by DC motor that represents the losses does not equal the traction produced by the traction motor, the net tor-que produces a force that accelerates or decelerates the train.
31
TractionMotor
TorqueXdcr
InertialMass
15 hpDC
Motor
DCMotorDrive
PWMsIgnal
conditioningcircuit
LocomotiveSimulator
i87C196KD-20Microprocessor
Signalamplifier
Curveand
GradeTable
Section LengthDrag and Grade
Index
PWMOutputP2.5
Torque0-5Vdc
Speed0-5Vdc
AnalogControl0-5Vdc
Speed
ACfrom
Inverters
Field Armature
Figure 16. Track load simulator functional block diagram
6.2.8.2 Track Simulator Results (Section to be completed for final report)
6.2.9 Network Communications Results Figure 17 shows the order of communications for the network traffic as well as typical in-
formation timing. This timing is specifically for the system designed for RS232C serial commu-nication between the master and the supervisory PC operating at 38.4KB and the CAN operating at 74KB. The delay between the status update to the supervisory PC and new command data from the PC may be as long as 150ms depending upon PC activities and complexity of the MATLAB algorithm running at that time. Figure 18 shows the effects of an I/O processor miss-ing from the network. The master waits up to four milliseconds after a request for message up-date before continuing on to the next update request. Figure 19 and Figure 20 shows the CAN message timing during request for updates in greater detail. From these figures, the overhead associated with requesting message updates can be determined.
32
Figure 17. Message timing for normal operations
Figure 18. Message timing for no response from Flywheel #2.
33
Figure 19. CAN status update message.
Figure 20. CAN status message update with no response from Flywheel #2 unit.
6.3 Power Control
6.3.1 AC Drives The AC drive converts a DC voltage into a 3-phase AC voltage, using pulse-width mod-
ulation (PWM), which is used to control a 3-phase induction machine. The AC drive is com-prised of a voltage-source inverter (VSI) 6and control hardware. A schematic of the VSI is shown in Figure 21. The switches being used in the VSI are IGBTs. The capacitor in Figure 21 helps stabilize the DC bus voltage.
dc
A+ B+ C+
A- B- C-
Figure 21. VSI Schematic
34
Figure 22 shows a more detailed view of phase “A” of the VSI presented in Fig-ure 21. Phases “B” and “C” are identical in design to phase “A”. The “fly-back” diodes, DA+ and DA-, across each IGBT switch are necessary to allow current to flow in both directions since IGBTs only conduct in one direction. The “fly-back” diodes come as part of the IGBT package. The inductor, LA, is the turn on snubber, it limits the turn-on shoot-through current of the IGBTs, and the metal-oxide varistor (MOV) protects the IGBTs against voltage spikes.
MOV
A+
A-
A+
A-
dc
Figure 22. Schematic of phase “A” of a 3-phase VSI
The control hardware in the AC drive is composed of a digital signal processor (DSP) evaluation board, an A/D converter board, a couple of boards that deal with feeding back the ro-tor position, a delay and fiber-optic transmitter board, and six gate driver boards. The DSP is the brains of the AC drive, it controls the drive. The A/D converter board provides an interface be-tween two current and two voltage sensors and the DSP. The rotor position boards filter out noise and increase the resolution of the position encoder signal being fed back to the DSP. The delay and fiber-optic transmitter board provides some protection against allowing the inverter to short out the DC bus by delaying the turn-on signal of an IGBT. The delay and fiber-optic transmitter board also transmits the gate pulses from the DSP to the gate drivers over fiber-optic cables. The fiber-optic cables are used for noise immunity. The gate drivers provide the neces-sary current and voltage to turn the IGBTs on and off.
The control algorithm used to control the AC drive is indirect field-oriented control (FOC). The use of the FOC algorithm allows for control of the torque produced by the induction machine independent of the induction machine’s speed.
6.3.2 Power Train In addition to the software simulation, a scale model of the locomotive propulsion system
was built in the lab. The basic setup is shown in Figure 23. The gas turbine was modeled with an electric machine operated at constant speed, similar to the case with a gas turbine. A solid state exciter was built for the synchronous generator to match those generally present on locomotives.
Two traction drives were implemented, rather than four as would be the case with an ac-tual locomotive. The shafts of the traction motors were also connected to DC motors with an added inertia on the shaft to simulate the train's inertia. The DC motors were run by DC drives to simulate the varying loads that would be seen over the course of a route, with all of the route in-formation programmed into the drive controls.
35
6.3.2.1 Traction Motors (Section to be completed for final report)
6.3.2.2 Flywheels Two flywheels are implemented with induction machines run with an inertial mass on the
shaft. The machines are run relatively slowly, to reduce the rotational losses. The primary pur-pose of the hardware simulation is to verify the control system, so these added losses can be tole-rated, especially since DC motors, driven by DC drives are used to supply the flywheel losses.
Two flywheels are used to explore the issues related to sharing loading between multiple flywheels, versus the option of one large flywheel.
6.3.2.3 Generator and Rectifier The power source for a train is provided by an on-board engine or track side energy sys-
tem such as a catenary or hot rail. The demonstration system simulates the case where the pri-mary energy source is an on-board gas turbine. This power supply consists of three elements; the engine, the alternator and the rectifier as illustrated in Figure 23. For the demonstration sys-tem, the engine power is provided by an induction motor connected to the local utility service. The rotor shaft of the induction motor is mechanically linked to the rotor shaft of a synchronous generator which serves as the alternator. The three phase AC output from the synchronous gene-rator is rectified and the DC power is then distributed to the inverters which power the traction motors and flywheels. Additionally, a dynamic brake can remove energy directly from the DC bus, where it is dissipated as heat in a resistor bank. The DC bus is the energy distribution high-way allowing power to flow between the various sources and uses of energy.
TractionMotor
#1
TractionMotor
#2
FlywheelMotor
#1
FlywheelMotor
#2
DynamicBrake
RectifierAlternatorEngine
Figure 23. Train energy control block diagram.
Key to effective energy management is maintaining nearly constant voltage on the DC bus. Poor regulation of the DC bus voltage can lead to equipment damage from overvoltages or degrade inverter performances due to undervoltages. There are two ways of regulating the DC bus voltage; using a controlled SCR bridge rectifier or varying the generator voltage output with an uncontrolled bridge rectifier. The SCR rectifier has advantages of faster response to sudden load changes. A microprocessor implementation of this type of rectifier was presented by R. Wall and H. Hess7. For the SCR rectifier, the DC voltage is described by Equation 1. In this ex-pression, α is the firing angel as referenced to the voltage zero crossing. IDC is the DC bus load current and VLL is the alternator three phase AC voltage.
36
( )V VL
IDC LLs
DC= −3 2 3π
αωπ
cos
Equation 1
The DC bus voltage for the uncontrolled rectifier method is expressed by Equation 2. With this method, the only way of controlling VDC is by regulating VLL. The alternator AC out-put is usually regulated by controlling the armature field current in some closed loop fashion. Although response time for this type of control is usually much slower than the SCR rectifier, it can be made acceptable by limiting the rate of change the load.
V VL
IDC LLs
DC= −3 2 3π
ωπ
Equation 2
Regulating the alternator field current is the primary means of controlling the output vol-tage. Using the assumption that load changes will be much slower than the transient response of the generator, a simplified model is sufficient. Equation 3 expresses the nonlinear relationship between the open circuit output voltage and the field current. Constants K1 and σ are determined from open circuit test. The results of these tests are recorded in Table VI and Figure 24.
E K Ia F= ⋅1σ
Equation 3
Ea represents the open circuit or no load voltage. The generator impedance is determined from short circuit tests. Equation 4 through Equation 6 follow directly from Figure ? which represents a simplifies steady state electrical model for a synchronous generator 8 a dynamic model for the synchronous generator. Recall that the problem is to maintain constant voltage on the DC bus. Equation 6 shows that this voltage is dependent on the field current as well as the load impedance hence the control system must regulate the field current in accordance to the nonlinear expression of Equation 7. Figure 25 represents the MATLAB model used for comput-er verification and simulation. A discussion of the gains and constants are provided in the report by Price Lockard to be included in the final report. The constants used in the PI algorithm block were systematically determined from the MATLAB model to provide the most responsive stable operation.
V E I Z EVZ
ZLL a L G aLL
LG= − ⋅ = − ⋅
Equation 4
V EZ
Z ZLL aL
L G= ⋅
+⎛⎝⎜
⎞⎠⎟
Equation 5
37
V K IZ
Z ZLL fL
L G= ⋅ ⋅
+⎛
⎝⎜
⎞
⎠⎟1
σ
Equation 6
IV Z
ZKf
LLG
L=⋅ +⎛⎝⎜
⎞⎠⎟
⎡
⎣
⎢⎢⎢⎢
⎤
⎦
⎥⎥⎥⎥
1
1
σ
Equation 7
Table VI. Synchronous generator open circuit test results.
If(dc amps) VOC (AC volts)0.34 33 0.68 59 1.02 84 1.36 110 1.72 136 2.05 158 2.36 177 2.73 199 3.07 217 3.425 233 3.75 249 4.08 262 4.43 274 4.79 286 5.12 296
38
Open Circuit Voltage Characteristics
0
50
100
150
200
250
300
0.34
1.02
1.72
2.36
3.07
3.75
4.43
5.12
Field Current - Amps
Ope
n C
ircui
t V A
C
Figure 24. Open circuit terminal voltage as a function of field current.
k1-------------k2s + 1
k5-------------k6s + 1
y= k3*xk4 y= k7*xk8 k9
+
k10
k11 + kp +ki/s +k12 -k12
BUS VoltageCommand
Value
PI Algorithm
C1 C2
DC LoadCurrent
Feedback
Rectifier
VoltageInst.
GateDrive #1
GateDriver #2
DC BusVoltage
Figure 25. DC bus voltage control MATLAB simulation model.
6.3.2.4 Dynamic Brakes (Section to be completed for final report)
39
6.3.3 Load Simulator (Section to be completed for final report)
6.3.3.1 Mechanical Dynamics (Section to be completed for final report)
6.3.3.2 Electrical Dynamics (Section to be completed for final report)
6.4 Demonstration Results (Section to be completed for final report)
7. Conclusion A system for improving the efficiency of high speed rail locomotives has been presented.
High speed rail transportation is gaining favor in the United States with several systems under consideration. An overview of high speed rail locomotives was presented, along with a brief re-view of flywheel energy storage. The addition of energy storage to the locomotive allows the possibility to optimize the design and operation of the locomotive in new ways, saving both capi-tal and operating costs.
This investigation looked at energy management schemes for trains carrying flywheels. The energy storage capacity that a train is equipped with both route and schedule dependent. The distribution of that storage capacity over a few or many flywheels is mainly dependent upon flywheel mechanical design issues. Some secondary effect optimizations are possible if numer-ous flywheels are used, however, the choice of few versus many flywheels will be affected by economy of scale and operating limitations. Network communications has some advantages over direct wire controls in signal relia-bility as well as hardware reliability. Conversion of command signals to high power controls can be located physically close to actuators thus reducing radio frequency and electromagnetic emis-sions. The network provides a low power interface to controls and instrumentation. The cost of networked control systems is relatively insignificant as compared to the total cost of train con-trols. Generally, inverter / induction motor traction motors drives systems receive identical commands since induction motors are self correcting for wheel slip. Hence, there are a minimal number of control wires for traction motors and a proliferation of instrumentation per drive axle. Some economic advantage may be gained over direct wired controls for systems controlling sto-rage systems with numerous flywheels. (Section to be completed for final report)
8. Recommendations (Section to be completed for final report)
40
9. Appendices (Section to be completed for final report)
10. Bibliography 1 Dooling, “Technology 1997 Analysis & Forecast: Transportation,” IEEE Spectrum. Vol. 34, No. 1, pp. 84-88, January 1997. 2 A. Ter-Gazarian, Energy Storage for Power Systems. Peter Peregrinus Ltd., London, 1994. 3 J.R. Hull, “Flywheels on a Roll”, IEEE Spectrum. Vol. 34, No. 7, pp. 20-25, July 1997. 4 Jonathan M. Meyer, Efficiency Optimization of Heavy Haul Rail Locomotive Traction Motor Drives. M.S. Thesis, University of Idaho 5 Diesel-electric Locomotives Transportation Formulae, General Electric 6 Mohan N., Undeland T.M., and Robbins W.P., Power Electronics, Converters, Applications, and Design, Second Edition, John Wiley and Sons, 1995 7 Richard W. Wall and Herbert L. Hess, “Design and Microcontroller Implementation of a Three Phase SCR Power Converter”, Journal of Circuits, Systems, and Computers, Vol. 6, No. 6 (1996), pp. 619-633. 8 Irving L Kosow, Electric Machinery and Transformers, Second Edition, Prentice Hall, 1991
