Paper-1990 01 NASA Field Oriented Control of induction motors

8/7/2019 Paper-1990 01 NASA Field Oriented Control of induction motors

1/10

T _

NASA Technical Memorandum 103154 .............................

.... Field Oriented Controlof Induction Motors :i:_ _ i : :_ : :

Linda M: Burrows .................Lewis Research Center : _..........._ :Cleveland, Ohio

Don S. ZingerUniversity of AkronAkron, OhioandMary Ellen RothLewis Research CenterCleveland, Ohio

Prepared for the__25th Intersociety Energy Conversion Engineering Conferencecosponsored by the AIChE, SAE, ACS, AIAA, ASME, and IEEE .._Reno, Nevada, AUgust 12-17, 1990

(NAqA-TM-1031_4)i _JDUCT I,qN MOT r)'3, S

FIELD O_KIFNTtO CuNT_t]L OF(_!ASA) 8 p CSCL 09C

G 31 _3

N90-22 73i


2/10


3/10

FIELD ORIENTED CONTROL OF INDUCTION MOTORS

Linda M. BurrowsNASA Lewis Research Center

Cleveland, Ohio 44135

Prof. Don S. ZingerUniversity of Akron

Dept. of Electrical Engr.Akron, Ohio 44325

Mary Ellen RothNASA Lewis Research Center

Cleveland, Ohio 44135

ABSTRACTInduction motors have always been known for their

simple rugged construction, but until lately were not suitablefor variable speed or servo drives due to the inherentconaplexity of the controls. With the advent of fieldoriented control (FOC), however, the induction motor hasbecome an attractive option for these types of drive systems.At NASA Lewis Research Center, we are currently workingwith an FOC system which utilizes the pulse populationmodulation method to synthesize the motor drive frequencies.This system allows for a variable voltage to frequency ratioand enables the user to have independent control of both thespeed and torque of an induction motor. A secondgeneration of the control boards have been developed andtested at NASA with the next point of focus being theminimization of the size and complexity of these controls .Many options have been considered with the best approachbeing the use of a Digital Signal Processor (DSP) due to itsinherent ability to quickly evaluate control algorithms. Thispaper will discuss the present test results of the system andthe status of the optimization process using a DSP.

1. INTRODUCTIONThis paper will briefly discuss the state-of-the-art (SOA)

in Electromechanical Actuation (EMA) technology and itslimitations. It is these limitations that have lead us to selectresonant power processing and induction motors as thebuilding blocks for our EMA system. A detailed descriptionof the 5 Hp induction motor controller (resident at NASALewis) will be presented. Included in this description wil lbe an explanation of an optimized field oriented control ofthe motor. This leads to a discussion on how to optimizeand minimize the control circuitry. The effort in progress atLewis proposes the use of a Digital Signal Processor toimplement the cohtrol algorithms for servo control of themotor.

2. BACKGROUNDThe SOA in Electromechanical Actuation CEMA)

technology has been the dc brushless motor (ac permanent

electronics do not scale up readily. As a result, the dcbrushless EMA technology is inadequate for thrust vectorcontrol (TVC) and other large vehicle applications with highpower requirements. In fact, when compared withhydraulics, this technology has failed to compete. In aswitched mode power processor, a pulse width modulation(PWM) technique is used to synthesize machine frequencywaveforms from a de link. All operations are performed atthis low (machine) frequency. If the frequency is raised, theturn off switching losses grow proportionately. Also, theenergy to be switched tends to increase as the squa_ of thecurrent. At the high power levels required for TVCapplications, this results in a driver with unacceptablethermal loads and a reduced efficiency characteristic. Sucha driver will usually be many times larger (and heavier) thanthe motor, making it an undesirable technology choice for aTVC application where weight and size are critical factors.

Other drawbacks of this technology include the inherentlimitations of adc brushless motor. It has a l imi ted torqueper ampere capability when compared to other motors. Inaddition, the permanent magnetic material limits the thermalcapability of the motor to below 1500 C. The desired motorfor a TV application, however, should have a high torqueper ampere capability as well as being rugged in design,capable of high temperature (> 200 C) operation.It is these inherent limitations in the dc brushless EMAtechnology that have lead us to pursue induction motors andresonant power processing. We believe there are markedadvantages in both the motor and remnant processing. Withresonant power processing, a pulse population modulat ion(PPM) technique is used to synthesize the machine frequencywaveforms from a high frequency link. All the switching isdone at the voltage zero-crossing, thereby reducing theswitching losses. The losses in this driver are thenprincipally the result of the voltage drop across theswitching element. In addition, all switching is performed atthe carrier (20 kHz) frequency which remits in small filterelements. With a resonant circuit topology, the driver canbe scaled up to achieve the higher power levels required forTVC and other control surface applications. By selectinghigh current semiconductor switches such as the newlydeveloped MOSFET Controlled Thyristor (MCT), which hasa low forward voltage drop, the size of the motor driver willbe comparable to the size of the motor.

The selection of the motor is equally as important asmagnet motor) driven by of a switched-mode power selecting the correct circuit topology for the motor drive.processor. Most drivers that have been built for de The motor chosen should be sensitive to the requirements ofbrushless motors are for low power applications and the the particular application. For TVC applications, the


4/10

demands on the power system are moderate for the durationof most of the mission. Peak demands, on the order of afew seconds, are required for the balance of the missiontime. The motor best suited for this application is one thatis capable of meeting the peak demands, yet not beoverdesigned (oversized) for the moderate demands whichmake up greater than 95% of the duty cycle. For this, aninduction motor is the best choice. It has a greater torqueper amp capabi_ty than other motors and can deliver torquesin excess of five times its nominal rating for short periodsof time (several seconds). It can be operated at highertemperatures (> 200 C), since there are no magnets. Theinduction motor is more rugged, smaller and lighter than adc brushless motor for a given peak power output [1]. Thedrawback of this motor has always been the difficulty incontrolling it. This, however, is no longer the case. Withthe electronics available today servo control of an inductionmotor is easily accompl ished through FOC techniques.

3. 20 kHz MOTOR DRIVE SYSTEMAt NASA we are currently developing a 20 kHz

induction motor drive system (built in part by GeneralDynamics Space Systems Division), which was firstdemonstrated at the University of Wisconsin and thentranaferred to NASA [2]. This motor driver integrates ahigh frequency link, PPM, and an induction motor with afield oriented control scheme, shown in figure 1. The pulse-population modulation converts a 20 kHz single phase inputinto a 3-phase low frequency output to drive inductionmotors. The PPM method selects individual pulses of the20 kHz voltage to produce the machine frequency waveform,as shown in figure 2 [3]. The amplitude (voltage) isdetermined by the density (population) of the pulses whilethe frequency is determined by the actual waveform pattern.This type of modulation enables the independent control ofboth voltage and frequency. Our system has been tested to2000 Hz and currently is controlling a 400 Hz 208 volt off-the-shelf induction motor.

The induction motor can be controlled in either avoltage or current regulation mode. The voltage mode isdirectly controlled by an external programmable controllerwhile the current mode is accomplished with the FOCboards. The current mode uses rotor speed feedbackinformation to create a closed loop control system. In thismode, the speed control board allows control of the torquecommand within microseconds. A maximum torque changecommand can be sent causing the motor to reverse directionover the full speed range very rapidly. This is important forthe TVC applications. For TVC, quick direction reversals

4. oFrIMIZATION TECT-[NIQUESOptimization of the motor drive system is being

attempted by two separate means in order to improve thesystem operation. An advanced motor control technique isbeing evaluated to improve efficiency. For increased systemreliability, the complexity of the control circuitry is beingaddressed. A detailed discussion of both techniques follows.

Optimized control of the motor is essential in order todevelop an EMA system that is as efficient as possible. Thelimit is (and probably always will be) the electronics. Evenusing resonant power processing, the capability of the driverelectronics is still the limiting factor for these systems atthese power levels. It is, therefore, advantageous to alwaysoperate the motor at its most efficient point. The advancedmotor control technique being investigated is the varying ofthe v/f ratio to obtain the maximum efficiency for any loadcondition. Figure 4 shows plots of typical motor data.With the v/f control, the curves can be moved left, right, upor down to operate the motor at its most efficient point forany load at any speed. Initial tests were run at theUniversity of Wisconsin and the results can be seen in figure5 [2]. This shows that by decreasing the v/f ratio (or flux),higher efficiencies can be obtained at lower torques. We arein the process of running tests to completely map efficiencyversus torque and speed curves for various loads. Thesecurves will help determine the optimum operating conditionfor any load applied. Control programs can then bedeveloped to operate the motor at its highest efficiency at alltimes.

The minimization of the complexity and size of thecontrols to increase system reliability has also been studied.After evaluation of a number of minimizat ion techniques, aDigital Signal Processor (DSP) has been chosen due to thebuilt-in capability to quickly evaluate control algorithms.DSP's have been specially designed to process the largeamount of data generated by a system in a relatively shortperiod of time. Due to their high speeds, DSP's have madetheir way into many control applications. The DSP can bethought of as a microprocessor designed to do high speedcalculations. Based on a highly pipelined architecture, thecontrol algorithms are able to exploit the parallelism inherentin the DSP [4]. DSP's are designed to quickly evaluateaddition and multiplication functions which can furtherincrease the speed of the tasks they are designed to perform.

Many of the algorithms used in the control of electricmachines are well suited for this specialized architecture.The algori thms generally consist of repetitive calculationswith Little branching. They require numerous multiplicationsto be done quickly for real time control. Because of the

are imperative which are possible with the FOC design.The block diagram for the entire FOC system is shown

in figure 3. The speed control board takes the speed androtor feedback information and generates the torquecommand current, iq,. The iq, and manually commanded fluxcurrent, h,, are processed by the remaining FOC boards togenerate the desired phase current reference commands.These commands are then sent to the regulator whichcompares the reference commands with feedback currents toproduce an error signal. This error is used to generate thegate drive signals for the PPM converter.

coqo_ match between the algorithms used in FOC and thecalculat ion abi lity of the DSP, these integrated circuits havequickly made their way into field orientation applications.DSP's have been shown to perform the field orientationcalculations in the range of 30 to 35 microseconds [5,6].

Although FOC systems are capable of rapid torquecontrol, most are not complete motor controllers. They takein torque and flux commands and output either currentcommands or switching states to a separate circuit whichgenerates gate drive signals. Generally motor controllersalso require a speed or position input. With mostimplementations, including NASA's, this is taken care of byexternal devices such as an encoder. To include the entiremotor control in a DSP would considerably increase the


5/10

execution time. One system which included a total controlfrom speed input to gate signal outputs, shown inside thedashed line in figure 6, was capable of operat ing at amaximum of 4 Khz switching speed 1"7]. For our system, adecision on switching pulses occurs every 25 microseconds.Since this time is close to the calculation period for typicalfield oriented calculations, there would be little timeremaining for other operations. With the presently availableprocessors, it may not be feasible to implement the entiremotor control.

multiplications. An example of a high level pseudo code forthis implementation would be:SPEED_ERR = SPEED - SPEED_COMMANDINT_ERR = (SPEED_ERR + OLD_ERR) *TIME_STEP/2+ INT_ERRIQS_COMMAND = KP * SPEED E_ + KI * INT ERROLDERR = INT_ERR

The slip calculator determines the desired slip for aWhile waiting for the inevitable faster process_-, ihe g_ven torque and flux command. This

speed requirements of the processor can be greatly reducedby allowing a separate circuit to generate the gate signals.In such a system, the processor will only be required toperform calculations quickly enough to maintain a reasonablecurrent waveform. In our three phase system, this requires aminimum of 6 updates at the highest frequency of operation[8]. For this 400 hz motor operating at twice its ratedspeed, this will require an update speed of at least 4800 hz.Such speeds are obtainable with present DSP technology.

A block diagram of the proposed DSP system is shownin figure 6. The system input will be a digital speedreference from the user. The DSP will also re_qui__ anencoder input to determine the actual speed of the motor.The DSP will output the desired current commands whichare converted to analog signals and compared with actualcurrent feedbacks. This will produce the error signals neededto generate correct gate drive signals.

The field orientation scheme being used is an indirecttype shown in figure 7. The control requires thecalculations for speed regulation, coordinate transformation,and additional multiplications and divisions. Assuming atypical proporfional-integnd control, the calculationrequirement will include several multiplications and anintegration. The coordinate transformation calculations willrequire lookup tables and multiplications. Many of theseoperations can be performed in one instruction cycle(typically less than 200 ns for present day DSPs). Estimatesfor some longer calculation times are summarized in table I.These estimates are based on the TM5320C14 cycle time of160 ns and standard routines for similar operations. It isestimated that the total processing time will be less than 30microseconds. This is considerably less than the maximumallowed time between current updates ( approximately 200us ). The DSP, therefore, is expected to perform adequatelyand minimize the additional hardware required with theoriginal FOC implementation.

is performed insoftware by using the relationship between the torqueproducing current, flux producing current, and the sliprequired for FOC. Pseudo code for this calculation is:SLIP_COM = (IQS_COMMANDffDS_COMMAND)/TR

where TR is the equivalent rotor time constant.The angle calculator finds the present angle requited for

transforming from the synchronous reference frame to thestationary. This is accomplished by adding the slip speedand rotor speed onto the present angle. The scaled value ofslip speed can be added every time it is calculated. Therotor speed is determined by the number of pulses receivedfrom the encoder. The rotor speed can be added to thepresent angle by having each encoder pulse interrupt theprocessor to perform the calculation. Special code will keeptrack of clockwise or counter-clockwise calculation.

The calculated angle is used to transform the currentcommands from the synchronous coordinates used forcalculation to the stationary coordinates needed by thecurrent regulator. This is done by using the values found ina lookup table for the sine and cosine of this angle andapplying them to standard reference frame transformations.The pseudo code for this operation is:

sINx = SIN(ANG)COSX = COS(ANG)IQSS = IQS COMMAND * coax + IDS_COMMAND SINXIDSS = IQS_COMMAND * SINX + IDS_COMMAND *

COSXOnce these are calculated, the two phase Q and D

current commands need to be converted to three phase ABCcurrent commands. This is done using standard conversion.The pseudo code for these conversions is:

5. PROPOSED IMPLEMENTATIONThe current hardware in use at NASA performs 5 major

functions. They are the speed control, the slip calculation,the angle calculation, the coordinate transformation, and thetwo to three pha_ conversion. With the present hardware

/AS__COMMAND = IQSSmS_COMMAND = -IQSS/2 - SQRT(3) * IDSS/2ICS_COMMAND = - (IAS COMMAND+ IBS_COMMAND )

Some steps can be saved when doing the calculations bycombining the coordinate reference transformation and theimplementationperformed on a separate board, although some functionsoverlap. With the DSP, each function is capable of beingperforoaed by relatively simple algorithms in a few lines ofcode.

The speed control of the present system performs aproportional-integral control upon the error between thecommanded speed and the actual speed. The present analogboard could easily be replaced with a few lines of codeperforming a simple trapezoidal integration and a few

each of these functions is " essentially - two to three phase transformation.It should be noted that each of these lines of pseudocode will take several lines of assembly language. The finalresult, however, will be a program of a size that is easilymanaged. The DSP system will allow our high performanceFOC to be implemented with less hardware than the presentconstruction. This will compact the system and thusincrease its reliability.


6/10

6. CONCLUSIONWhile ongoing effort is being made to optimize

NASA's 20 kHz motor drive system, it has already beenshown that the system will exceed performance ofconventional PWM motor drive technologies. Oncecompleted the system will be advantageous in areas such assize, weight, and efficiency, which ave critical parameters forEMA applications.

7. REFERENCES[1] Dr. T. Kume, "Wide Constant Power Range Vector-Controlled AC Motor Drive Improves Spindle Drives,"Power Convers. l .ntel. Motion, pp 59-66: April 1990.[2] Thomas A. Lipo, et. al., "Testing and Modification ofthe General Dynamics Converter", NASA Contractor Report,No. NAG3-786, pp 51-52: Sept. 1988.

[3] P. A. Sood, T. A. Lipo, and L G. Hansen, "A Versati lePower Converter for High Frequency Link Systems," IEEETrans. Power Electronics, vol. 3 no. 4, pp 383-390: Oct.1988.

[4] A. Jonathan, "Computer Architecture of Digital SignalProcessors," Proceedings of IEEE, pp 853-873: May 1985.[5] R. W. DeDonker, and D. W. Novotny, "The UniversalField Oriented Controller." Conference Record of the IEEEIAS Annual Meeting, pp 450-456: 1988.[6] T. G. Habetler and D. W. Novotny, "The Universal FieldOriented ControLler," Conference record of the IEEE IASAnnual Meeting, pp 514-522: 1989.[7] X. Xu, and D. W. Novomy, "Bus Utilization of DiscreteCRPWM Inverters for Field Oriented Drives," IEEE I.ASAnnual Meeting, pp 362-367: 1989.[8] R. D. Lorenz, "Controller Considerations for HighPerformance AC Drives," Introduction to Field Oriented andHigh Performance AC Drives, D. W, Novotny and R. D.Lorenz eds., IEEE, New York: 1985.

]flClt FREQUEI',_CY LI;NK

ii!

$oures

PPM CONVERTER

i: iI,i tC.OgFUTER

T GAT1NG SIGNALS

VaN VO.r.V_CU_EK( leVca P,EGULATOA

T(_tOUE

[A

h _ I_;DUCTIONy MOTOR>

t,_EEO _f_;_EE 0C_NT_U.ER E_._Q(

Figure 1. Diagram of 20 KHz Motor Drive System

4ORIGINAL PAGE ISOF POOR QUALITY


7/10

VHF

VCF

VLF

, 201_HzWavelorrn

IAAAAAAAAAAA_A_ZVVVVVVVVVVVVV _

Mac_meFrequencyWavefarmA AA AA AAAA 6A AA AV VW _'Figure 2. Modulation Technique to Produce Machine Frequency Waveform

k ._..._1

Figure 3. Block Diagram of FOC

2CO

150>-

i ICO

50

02000 4000 $000 8000 10000 .I 2000

Figure 4. Typical Induction Motor Curves

0.7-

0.6-

0.S o

,,_ 0.4 -uJ

0.3-

0.2-

0.1,

0.0,0.0

Rgure 5.

! , io.2 o14 oi_ o.'8 ,.oTorque In Per Unit

Plot of Efficiency vs. Torque for Varying V/F(200 Hz and 100 Hz Operation)

5 ORIGINAL PAGE ISOF POOR QUALITY


8/10

Figure 6.

!I

IA i Gate Siqnalsco.,.o, i

_ [ EncoderCouniet I

|_

Block Diagram of Speed Control System (Dashed Block Represents A DSPWhich Would Contain an Entire Motor Control System)

CuW_ SPEED I

COORDINATE

_,TRANSFORMATION

Figure 7. Detailed Diagram of FOC Scheme. Table 1.

FUNCTION

PICONTROL

COORDINATETRANSFORMATION

DIVISION

TIME(_)

1.4

1.9

5.8

Estimates of Function Processing "rime


9/10

NationalAeronautics andSpaceAdministration1. Report No.NASA TM-103154

Report Documentation Page2. Government Accession No. 3. Recipient's Catalog No.

5. Report Date. Title and SubtitleField Oriented Control of Induction Motors

7. Author(s)Linda M. Burrows, Don S. Zinger, and Mary Ellen Roth

9, Performing Organization Name and AddressNational Aeronautics and Space AdministrationLewis Research CenterCleveland, Ohio 44135-3191

6. Performing Organization Code

8. Performing Organization Report No.E-5516

10. Work Unit No.946-02-20

11. Contract or Grant No.

13. Type of Report and Period Covered

Technical Memorandum14. Sponsoring Agency Code

12. Sponsoring Agency Name and AddressNational Aeronautics and Space AdministrationWashington, D.C. 20546-0001

15. Supplementary NotesPrepared for the 25th Intersociety Energy Conversion Engineering Conference cosponsored by the AIChE, SAE,ACS, AIAA, ASME, and IEEE, Reno, Nevada, August 12-17, 1990. Linda M. Burrows and Mary Ellen Roth,NASA Lewis Research Center. Don S. Zinger, University of Akron, Department of Electrical Engineering,Akron, Ohio 44325. , _ _ ___. :

16. AbstractInduction motors have always been known for their simple rugged construction, but until lately were not suitablefor variable speed or servo drives due to the inherent complexity of the controls. With the advent of fieldoriented control (FOC), however, the induction motor has become an attractive option for these types of drivesystems. At: NASA Lewis Research Center, we are currently working with an FOC system which utilizes thepulse population modulation method to synthesize the motor drive frequencies/This system allows for a variablevoltage to frequency ratio and enables the user to have independent control of both the speed and torque of aninduction motor. A second generation of the control boards havd been developed and tested at NASA with thenext point of focus being the minimization of the size and complexity of these controls. Many options have beenconsidered with the best approach being the use of a Digital Signal Processor (DSP) due to its inherent ability toquickly evaluate control algorithms. This paper-will disct_ the present test results of the system and the status ofthe optimization process using a DSP. A_ _ ':J i _ .,, _ ,_._

17. Key Words (Suggested by Author(s))Induction motorsField oriented controlDigital signal processors

19. Security Classif . (of this report)Unclassified

18. Distr ibut ion Sta tementUnclassified- UnlimitedSubject Category 33

20. Security Classif. (of this page)Unclassified

21, No, Of pages 22. Price"A02

NASAFORMS26OCT86 *For sale by the National Technical Information Service, Springfield, Virginia 22161


10/10

Documents

Paper-1990 01 NASA Field Oriented Control of induction motors