RUN-TIME RECONFIGURATION OF AC DRIVE CONTROLLERS

1
RUN-TIME RECONFIGURATION OF AC DRIVE CONTROLLERS
University of Miskolc Department of Automation
Vásárhelyi József
E-mail vajo_at_mazsola.iit.uni-miskolc.hu
2
Details about the author

• 1983 - Graduated in Electronics and
  Telecommunication at Technical University of
  Cluj, Faculty of Electrical Engineering, Romania
• Since 1992 lecturer at the University of Miskolc
  Department of Automation, Hungary
• Presently PhD student at the Technical University
  of Cluj, Department of Electrical Drives and
  Robots, tutor Prof.Dr. Mária Imecs
• Research interests configurable architectures
  and their application to control devices

3
Summary

• INTRODUCTION
• Short introduction to reconfigurable systems and
  why are they used or should be used in AC drive
  control?
• BACKGROUND
• Presents the background of AC motor control.
• IMPELMENTATION STRUCTURE
• Give an answer of (a) possible hardware
  structure(s)
• RECONFIGURABLE CONTROLLER IMPLEMENTATION
• Presents the idea of reconfigurable controller
• TIME CONSTRAINTS OF THE RECONFIGURABLE CONTROLLER
• Time constrains of the reconfiguration process
  are presented
• Implementation issues and possible reconfigurable
  structures are presented
• CONCLUSION
• Conclusions and future work are presented

4
Introduction

• There are different approaches to define the
  reconfigurable systems Brebner, Hauck, Luk,
  Maciejowski, Shirazi, Vuillemin.
• Reconfigurable systems are usually considered
  those computing platforms whose architecture is
  modified by the software to suit the application
  at hand.
• Most of Re-configurable Computing Systems are
  plug-in boards made for standard computers and
  they act as a Co-processor attached to the main
  micro-processing unit.
• There was demonstrated significant potential for
  the acceleration of computing in general-purpose
  applications Hauck, Smith, Villasenor,
  Vuillemin.
• To treat the reconfiguration as a process one
  need a simple model for specifying and optimising
  designs, which contain elements that can be
  reconfigured at runtime.

5

• Comparing to the number of applications known in
  the reconfigurable filed just a few of them are
  concentrated in the study of vector control for
  AC drives.
• Vector control is a special field for digital
  signal processing.
• There are known dedicated DSP processors for
  digital motor control and successful
  implementations of vector control Beierke are
  referred. The DSP implementation of
  speed-sensorless induction motor drive using
  artificial intelligence is also known Vas.
• Up to now the studied literature by the author,
  only the research of Monmasson and his group is
  reported as direct application of reconfigurable
  structures in vector control for AC Drives
  Monmasson, Tazi. The most significant result
  introduced in reconfigurable control was the
  parallel-machine control architecture.

6

• The necessity of reconfiguration is based upon
  the practical observations that the performances
  of different types of vector controlled drives
  are different, depending primarily on the range
  of speed.
• It is known that the rotor flux oriented vector
  control is simpler to implement and therefore,
  widely used. One drawback of this method is the
  low efficiency at low ranges of speed.
• For lower speed range, the stator flux oriented
  vector control is preferred.

7
1. Background

• Complex industrial systems and robotics make use
  of electrical drives.
• Research efforts to find the optimal solution for
  AC motor control.
• Since the reconfiguration idea appeared by the
  introduction of Field Programmable Gate Arrays
  FPGA there is an increasing interest to find
  other solutions then DSPs for AC motor Control.
• Conclusion Find a solution for reconfigurable
  control instead of using adaptive control

8
Vector control structure for AC drive
is
Current feedback
Set parameters
Power
imR
Converter
PWM
Magnetising Flux

S
motor
-
Reference speed

S
Speed feedback
-
w
Source Texas Instruments
9

• Most of the motor control applications use
  asynchronous motors.
• The most often used method to control induction
  motors is the field oriented control method to
  achieve the best dynamic behaviour.
• Using the Parks direct and reverse
  transformations the AC drive can be controlled
  like a separately exited DC machine, whereby the
  direct (d) path is representing the flux building
  component and the quadrate (q) path sets the
  electrical torque.
• Best results are obtained when the magnetising
  current imR is kept constant, which is direct
  proportional to the rotor flux ?r under the
  assumption that the main inductance Lh is
  constant

10
(1)
(2)
(3) - (4)
Based on the mathematical model of the induction
machine in field co-ordinates, given in equation
(1-4), a controller was developed and a flux
model was derived.
11
Vector control system for voltage-source
inverter-fed induction machine
12
2. IMPELMENTATION STRUCTURES

• The control system presents modularity as shown
  in the previous figure. The main modules are
• System transformations direct and reverse
  Parks transforms.
• Orientation field computation
• Control Strategy
• Co-ordinate transformation
• There is need for an extra module, not presented
  on the figure, which used for the external A/D
  conversion control.
• This modularity allows exploiting of all the
  parallelism of the control algorithm.

13
Starting from the mentioned modularity a
reconfigurable controller structure it is
introduced.
14
The reconfigurable controller concept

• Implement different controllers for the same
  controlled process.
• Each controller structure can be seen as a
  distinct state of a state machine.
• Transition from one state to the other can be
  determined by the state parameters of the
  controlled system.
• If a transition condition occurs, i.e. the motor
  speed reference transits a limit value, the need
  for reconfiguration is fulfilled and the
  controller generates the self re-configuration
  process.

15
The desired re-configurable controller can be
implemented under the following conditions

• External memory is needed to store the several
  configurations (Configuration Store).
• Either software or hardware has to be capable to
  start a reconfiguration on need. (Configuration
  Starter).
• The evolution of the system must be predictable
  in order to pre-compute the possible
  configuration.
• The system control states have to be quantified
  and finite that is a condition imposed by the
  finite capacity of available external memory.
  (Configuration Memory)
• The existence of high-fidelity models and
  effective approximation-identification algorithms
  for multivariable systems..

16
Programmable logic structures considered for the
hardware support of the controller

• Triscends CSoC
• Configurable System Logic
• Incorporated processor core
• External and internal memory
• Ability to start self reconfiguration

• Xilinxs FPGA Virtex
• Abundant logic resources
• Internal memory
• Ability for partial reconfiguration
• High computing speed
• Relative high reconfiguration frequency

17
3. TIME CONSTRAINTS OF THE RECONFIGURABLE
CONTROLLER

• One can see that the controller structure is
  implemented in the configurable logic and the
  controller supervisor is a processor core.
  Depending on the implementation hardware the
  reconfiguration can be done as
• Partial reconfiguration reconfiguring each
  module step by step conform to the method
  introduced by Hauck. The method is called
  pipeline morphing, intended to reduce the latency
  involved in reconfiguring from one pipeline to
  another. (This is the case if Xilinx FPGA it is
  used.)
• Total reconfiguration reconfiguring the
  controller as a whole. (This is the case of
  Triscends CSoC.)

18
The reconfiguration times computed for partial or
total reconfiguration methods are

• For the partial reconfiguration (reconfiguration
  is done by pipeline morphing) the maximum
  reconfiguration frequency is 66 MHz if the best
  existing hardware support is the Virtex FPGA.
  Reconfiguration of each module can be done if
  SelectMAP mode is used and the time needed for
  reconfiguration is under 1 ms.
• The time needed for total reconfiguration of a
  CSoC by using the parallel mode initialisation,
  is 7.4 ms at 40MHz-reconfiguration frequency.
  This involves for implementation of the
  reconfigurable controller the use of two CSoC
  chips.

Q and P represent the two control structures, C
and C represent the reconfiguration control,
is and is are the observed current signal
and the current control signal, respectively.
19
Testing CSoC chip CSL resources used in the
implementation of the controller
20
4. CONCLUSIONS

• It was explained why reconfigurable structures
  are used in vector control.
• The controller modularity help reconfiguration
• The controller states are quantified.
• Reconfiguration have to be done between two
  sampling event.
• Reconfiguration time have to be less or equal to
  sampling period of the controller.
• Reconfiguration time may became critical and
  there is need for reconfigurable structures with
  faster reconfiguration time.

21
Conclusion II.

• Future work
• Finalise the CSoC implementation and test with an
  AC drive.
• Implementation of control structures in Xilinx
  Virtex FPGA
• Create standalone module library for the
  reconfigurable controller.