PMSM Drives Using Artificial Neural Network For Reducing Torque Ripple

PMSM Drives Using Artificial Neural Network For Reducing Torque Ripple

Praveen Kumar

ME Scholar, Department of Electrical Engineering, RCET, Bhilai

Anurag Singh Tomer

Assoc. Prof., Department of Electrical Engineering, RCET, Bhilai

Abstract

This paper describes intelligent direct torque control (DTC) technique for Permanent Magnet Synchronous Motor (PMSM) drive based on Artificial Neural Network (ANN) Systems. The proposed system has proven successful in controlling the instantaneous torque so as not to depend only on the estimation flux, torque and position, but also the estimation of the lookup table and the generation of driver switching table. Experimental results prove the MATLAB simulation results for torque, speed and flux estimations.

1. Introduction

   Among the ac drives, permanent magnet synchronous machine (PMSM) drives have been increasingly applied in a wide variety of industrial applications. The reason comes from the advantages of PMSM: high power density and efficiency, high torque to inertia ratio, and high reliability. Recently, the continuous cost reduction of magnetic materials with high energy density and coercitivity (e.g., samarium cobalt and neodymium-boron iron) makes the ac drives based on PMSM more attractive and competitive. In the high performance applications, the PMSM drives are ready to meet sophisticated requirements such as fast dynamic response, high power factor and wide operating speed range. This has opened up new possibilities for large scale

   application of PMSM. Consequently, a continuous increase in the use of PMSM drives will surely be witnessed in the near future [1 – 3].

   A simplified variation of field orientation known as direct torque control (DTC) was developed by Takahashi and Depenbrock [4]. DTC has a relatively simple control structure yet performs at least as good as the FOC technique. It is also known that DTC drive is less sensitive to parameters de-tuning (only stator resistor is used to estimate the stator flux) and provides a high dynamic performances than the classical vector control (fastest response of torque and flux).

   This method allows a decoupled control of flux and torque without using speed or position sensors. This type of command involves nonlinear controller type hysteresis, for both stator flux magnitude and electromagnetic torque, which introduces limitations such as a high and uncontrollable switching frequency [5]. This controller produces a variable switching frequency and consequently large torque and flux ripples and high currents distortion

   In order to improve DTC performance by reducing torque and flux ripples, many control strategies have been presented since 1990s one of these methods is by using Artificial Intelligent (Artificial Neural network (ANNs), Fuzzy Logic Control (FLC) techniques [6-8].This is mainly due to their ability learning and generalization.

   For that, we developed an intelligent technique to improve the dynamic performances of the DTC control; this method consists in replacing the hysteresis comparators asynchronous machine DTC by a controller based on the artificial neurons networks in order to lead the flux and the torque

   towards their reference values during a fixed time

   u = Rs.i

   + sq .

   (2)

   period.

   sq sq

   sd

2. Direct torque control

   1. Structure

      Where the direct and quadrate axis flux linkages are,

      sd = Ld.isd + f (3)

      sq = Lq . isq (4)

      The electromagnetic torque of the motor can be evaluated as follows,

      T = 3 P [ . i

      + (L – L ) . i . i ]1/2 (5)

      e 2 f sq

      d q sd sq

      Figure 1. Direct Torque Control structure

      A direct-torque-controlled PM synchronous motor supplied by a voltage-source inverter is present and the stator flux linkage and the electromagnetic torque are controlled directly by applying optimum voltage switching vectors of the inverter. It is a principal goal to select those voltage switching vectors which yield the fastest electromagnetic torque response. The six active switching vectors (u1, u2, , u6) and the control scheme of the DTC PM synchronous motor supplied by a VSI.

      With:

      : rotation's speed electric. P: Number of pairs of poles.

      f: flux produced by the permanent magnet. sd: d axis stator magnetic flux,

      sq: q axis stator magnetic flux,

      Ld: d axis stator leakage inductance, Lq: q axis stator leakage inductance, Rs: stator winding resistance,

      Te: electromagnetic torque,

   2. Flux and torque estimator

      The stator electric equations of the PMSM, are given by:

      =

      R . i

      (6)

      sd 0 sd

      s sd

      Figure 2. Phasor diagram of the PMSM

      In this paper PMSM(Permanent Magnet Synchronous Motor), which consists of three stator windings and a rotor magnet described by the following equation (Voltage, Flux, Torque),

      sq = 0 sd Rs . isq (7)

      s = 2 + 2 (8)

      = arc tan( ) (9)

      The electromagnetic torque is:

      T = 3 P [ . i

      – . i

      ]1/2 (10)

      usd

      = Rs.isd

      + sd .

      sq

      sq

      (1)

      e 2 sd sq sq sd

      P: Number of pairs of poles

      The electromagnetic torque in a PMSM machine can be regulated by controlling the magnitude and angle of the stator flux linkage or load angle . This can be performed by applying the proper output voltage vectors of an inverter to the machine

   3. Control switches of the inverter

      The switches of the converter of tension must be controlled so as to maintain the flux and the torque of the machine. The vector of the stator voltage can be written in the form:

      When flux is in zone I, the vectors Ui+1or Ui- 1are selected to increase the amplitude of flux, and U i+2 or Ui-2 to decrease it. What shows that the choice of the vector tension depends on the sign of the error of flux, independently of its amplitude [4]. This explains why the exit of the corrector of flux can be a Boolean variable. One adds a band of hysteresis around zero to avoid useless commutations when the error of flux is very small,[1] [6]. Indeed, with this type of corrector in spite of his simplicity, one can easily control and maintain the end of the vector flux, in a circular ring. The switching table proposed by Takahashi [4], as given by Table 1.

      Table 1. Switching scheme

      sector		1	2	3	4	5	6
      Flux	Torque
      e=1	Te = 1	U2	U3	U4	U5	U6	U1
      	Te = 0	U7	U0	U7	U0	U7	U0
      	Te = -1	U6	U1	U2	U3	U4	U5
      e=0	Te = 1	U3	U4	U5	U6	U1	U2
      	Te = 0	U0	U7	U0	U7	U0	U7
      	Te = -1	U5	U6	U1	U2	U3	U4

      sector		1	2	3	4	5	6
      Flux	Torque
      e=1	Te = 1	U2	U3	U4	U5	U6	U1
      	Te = 0	U7	U0	U7	U0	U7	U0
      	Te = -1	U6	U1	U2	U3	U4	U5
      e=0	Te = 1	U3	U4	U5	U6	U1	U2
      	Te = 0	U0	U7	U0	U7	U0	U7
      	Te = -1	U5	U6	U1	U2	U3	U4

      2 j2 j4

      U =

      s

      s

      3

      . U . (Sa + Sb . e 3 + Sb . e 3 ) (11)

      Where (Sa, Sb, Sc) represent the logical state of the 3 switches: Si = 1 means that the high switch is closed and the low switch is open and Si=0 mean that the high switch is opened and the low switch is closed.

      One will thus seek to control flow and the couple via the choice of the vector of voltage which will be done by a configuration of the switches. As we have 3 switches, there are thus 2^3=8 possibilities for the Ui vector. Two vectors correspond to the null vector (U0, U7):

      (Sa, Sb, Sc) = (0, 0, 0) & (Sa,Sb,Sc) = (1,1,1)

      Figure 3. Six active switching vectors

   4. Switching schemes

3. Structure of DTC System Based on ANN

   A neural network is a machine like human brain with properties of learning capability and generalization. They require a lot of training to understand the model of the plant. The basic property of this network is that it is capable of approximating complicated nonlinear functions.

   It constitutes an approach which gives more opportunities to approach the problems of perception, memory, learning and analysis under new angles. It is also a very promising alternative to avoid certain limitations of the classic numeric methods. Due to its parallel treatment of the information, it infers emergent properties able to resolve problems qualified in the past as complex.

   The structure of the direct neuronal torque control of a permanent magnet synchronous machine is illustrated bellow Figure 4.

   Figure 4. DTC neural networks controller

   scheme

   The Neural Network is based on layers of neurons. Fundamental processing element of a neural network is a neuron. Initially, the network has to be structured for a particular application, then the network can be trained be choosing the initial weights randomly.

   In this paper artificial neural network torque are used on torque comparator and flux comparator. The designed ANN controller is a feed forward neural network and it has three layers called input layer, hidden layer and the output layer. The ANN is trained by a learning algorithm in which the weights of the network is adapted iteratively until the error between target vectors and the output of the ANN is less than the error goal. Back propagation algorithm is used for training in this paper. In the back propagation method, the error signal is propagated through the network in backward direction and adjusts the weights of the free parameters and reduces the error. The back propagation training algorithm is shown in the flow chart Figure 5.

   Figure 5. ANN training algorithm

4. MATLAB/SIMULINK models and Results:

   Figure 6. Matlab/Simulink model of PMSM Drive for DTC method

   Figure 7. Results of PMSM Drive for DTC

   method

   Figure 8. Matlab/Simulink model of PMSM Drive for DTC method using Neural Network

   Figure 9. Results of PMSM Drive for DTC method using Neural Network

5. Conclusion

   This ANN controller is used successfully in Direct Torque Control system for PMSM. The design of ANN controller and its performance shows that the error is minimized. The overshoot in the speed curve is maintained very less and the ripples in the torque curve are maintained minimum.

6. References

1. B. K. Bose, Power electronics and motion control Technology status and recent trends, IEEE Trans. Ind. Applicat., vol. 29, pp. 902909, Sept. /Oct., 1993.

2. W. Leonhard, Adjustable speed ac drives, Proceedings of IEEE, vol. 76, pp. 455- 471, April, 1988.

3. B. K. Bose, Variable frequency drives-technology and applications, Proc. ISIE 93(Budapest), June, 1993, pp. 1- 18.

4. Takahashi and T. Noguchi, A new quick-response and high-efficiency control strategy of an induction motor, IEEE Trans. Ind. Applicat., vol. 22, no. 5, pp. 820827, Sept./Oct. 1986.

5. B.BOSSOUFI, M.KARIM, S.IONITA, A.LAGRIOUI,

   Performance Analysis of Direct Torque Control (DTC) for Synchronous Machine Permanent Magnet (PMSM) in Proc. IEEE-SIITME2010,pp. 275-280, 23-26 Sep 2010,

   Pitesti, Romania.

6. R.Kumar, R. A.Gupta, S.V.Bhangale, H.Gothwal,

   ARTIFICIAL NEURAL NETWORK BASED DIRECT TORQUE CONTROL OF INDUCTION MOTOR

   DRIVES, IETECH Journal of Electrical Analysis, Vol: 2,

   No: 3, 159 165, IETECH Publications, 2008

7. R.Toufouti S.Meziane, H. Benalla, DIRECT TORQUE CONTROL FOR INDUCTION MOTOR USING INTELLIGENT TECHNIQUES, Journal of Theoretical and Applied Information Technology, pp 35- 44, JATIT, 2007.

8. Mohammed Khalaf Masood and Essam M. Abdul- baki, Improvement of direct torque control system using radial basis function neural Network fuzzy control techniques, First Engineering Scientific Conferencce College of Engineering, University of Diyala 22-23, pp. 473-484,December 2010.

9. B. Bose, "Power Electronics and Motor Drives, Advances and Trends", Academic Press is an imprint of Elsevier, published 2006, 935 pages.

10. T. J. Vyncke, R. K. Boel and J .A .A. Melkebeek,

Direct torque control of permanent magnet synchronous motors an overview, 3rd IEEE Benelux young researchers symposium in electrical power engineering, Apr 2006.