Five-Phase Induction Motor DTC-SVM Scheme with PI Controller and ANN Controller Academic research paper on "Electrical engineering, electronic engineering, information engineering"

Available online at www.sciencedirect.com

ScienceDirect

Procedia Technology 25 (2016) 816 - 823

Global Colloquium in Recent Advancement and Effectual Researches in Engineering, Science

and Technology (RAEREST 2016)

Five-Phase Induction Motor DTC-SVM scheme with PI Controller

and ANN controller

Subodh Kanta Barika*, Kiran Kumar Jaladib

aM.Tech Student, NITKurukshetra, Kurukshetra, India bAsst. Prof., NIT Kurukshetra, Kurukshetra, India

Abstract

This paper analyze design of space vector modulation based direct torque controlled (DTC-SVM) five-phase induction motor incorporated with ANN controller and has given a comparison with PI controller. This ANN controller is employed to improve the control performance parameters such as reducing torque & flux ripple, reducing the settling and rise time compare to that of PI controller. The modelling and simulation is performed in MATLAB, it shows that with intelligent controlled the performance parameters have been improved and drive system could be operated at low speed. The proposed model is less complex, require a single ANN controller for decoupled flux and torque controlled.

© 2016 The Authors. Published by ElsevierLtd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.Org/licenses/by-nc-nd/4.0/).

Peer-review under responsibility of the organizing committee of RAEREST 2016

Keywords: Modelling of Jive-phase motor;SVM ;SVM-DTC;ANN controller;PI controller;MATLAB/SIMULINK

1. Introduction

The utilization and demand of electrical energy is increasing very fastly day by day as population growth has been increasing rapidly. According to ministry of power, the generation capacity in India now a days is around 235 GW and the expected demand will rise up to 930GW in 2030[1]. Thus saving of this energy is one of the most challenging priority for research area now a days. Normally Electric Motor consume most of energy generated hence energy could be saved by devising efficient operation scheme of these electrical motors. One of the method could be special design of motor with high-energy efficiency and other method is proper speed control. It is found that efficiency of the motor could be improved significantly by adjusting the speed depending on load requirement or variation.

Corresponding author. Tel.: +917504727888 E-mail address: subodh.nitkkr@yahoo.com

2212-0173 © 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Peer-review under responsibility of the organizing committee of RAEREST 2016 doi:10.1016/j.protcy.2016.08.184

Nomenclature

Rs Stator Resistance

Rr Rotor Resistance

Lis Stator leakage Reactance

Llr Rotor leakage Reactance

Lm Magnetizing Reactance

Ls Stator Reactance

Lr Rotor Reactance

J Inertia

P Number of Poles

wr Speed in radian per second

Te Electromagnetic Torque

TL Load Torque

Synchronous speed

n Number of phases

s Subscript denotes to stator

r Subscript denotes to rotor

A Transformation matrix

Arbitrary stator voltages on respected d-q-x-y axis

^dr,Vqr,^xr yyr Arbitrary rotor voltages on respected d-q-x-y axis

^ds^qs ^xs^ys Arbitrary stator fluxes on respected d-q-x-y axis

^dr^qr iïxr^yr Arbitrary rotor fluxes on respected d-q-x-y axis

Arbitrary currents associated with stator on d-q-x-y axis

ldr,lqr ¿xr,lyr Arbitrary currents associated with rotor on d-q-x-y axis

v's Reference voltage vector

ts Sampling time

Val ,Vbl Large voltage vector of phase a & b

Medium voltage vector of phase a& b

Switching time of large and small vectors of corresponding phase a & b

to Switching time of zero vector

vDC Inverter DC voltage link

k Sector number

a Angle subtended by reference vector with respect to phase a axis

T A constant value

Stator flux speed

The variable speed drive is applied predominately in three-phase machine. Now a days, because advent and flexible use of high power semiconductor devices in inverter drives for supply purpose (inverter with DC link), there is a scope of increase of number of phases, by increasing the number of legs. This has led an interest of multiphase machine, since multiphase has certain advantage over three-phase machine such as lesser torque ripple, greater fault tolerance, higher torque density, reduction of switches been used in inverter legs, improved noise characteristic and flat torque [2]. This type of motor has some application in high power ship propulsion, more electric aircraft, electric hybrid vehicle and traction. In this paper five phase motor is taken as multi-phase motor [3]. Since 1980, DTC has been used in industrial application worldwide because of direct decoupled torque control and flux control with high dynamic performance of the machine. The only drawback in conventional DTC control is production of higher torque ripple [4]. Since then, many method in literature invented different type of techniques to reduce the torque ripple such as duty ratio control, use of multilevel inverter, applying model predicting algorithm and space vector modulation (SVM) technique. Among all, SVM technique is most accepted because of significantly reduction of torque ripple. Now a days with advanced used

of PC with high speed DSP processor, it is possible to employed computational intelligence to this SVM-DTC. Normally plant is a dynamic and non-linear control system, and parameters varies continuously time to time rapidly. Due to fixed gain, PI may not perform to control required parameters properly and as result appliances might be malfunctioning [5-6]. To address these issue, artificial neural network (ANN) has been employed here taken as substitute of PI. This neural network is simple in structure, easy to train and control, ability to approximated the non-linear response function and insensitivity to saturation & temperature variation, hence it is superior to a DSP based controller in execution time and hardware. Neural network in computer world works on human brain's mechanism of problem solving strategies. Like in human brain, several linkage are provided through network of axons and synapses to be computing called neurons. The placement of neurons can be modified in different layer from input to output. They communicate to each other where electrical signal are generated among them to pass information.

This paper organize as follows. Section 2 gives a brief discussion regarding modeling of five phase induction machine. Section 3 describes SVM techniques for 5-phase system, section 4 & 5 presents a brief discussion on PI & ANN controller with complete drive system. Simulation results and Conclusion are given in section 6 & 7.

2. Mathematical Modelling of five-phase Motor

The induction motor can be represented mathematically with arbitrary no. phases on both stator and rotor sides [7]. In this type modelling original phase variable transform into new phase variable in arbitrary d-q plane rotating with synchronous speed o>e. Number of phases before and after transformation on both sides must be same and spatial displacement between two phases is 2n"/n. Correlation between these two variable in different plane as follows.

Vs. = AVS .. isH =A is , . Vs . =A Vs . .

aq s abcae aq s abcae ' aq s ' abcae ,1.

Vr. =A Vr . . irH =A ir, . yrH =A^

aq r abcae aq r abcae ' aq rT abcae

For five phase machine, the mathematical model represents in d-q-x-y-0 arbitrary plane. However first two components (d-q) are responsible for fluxes developing, power generation, torque production and remaining components (x-y) generates losses in the system. The only reason for zero component being used to show power invariance in the system. The decoupling transformation matrix for five-phase machine represent as follows.

A = J-

cos(-) cos(-)

sin(—)

sin(—) 1

sin(—)

cos(-) cos(-)

sin(—) 1

4^ cosf-)

sin(—)

cosf-)

-sm(—) 1

,2K cosf-)

Sin(—)

cosf-)

■ r4^ -sin^—)

The necessary machine model equations in stationary arbitrary reference plane represent as follows.

For Stator

Vds = Rsids + Pfds Vqs = Rsiqs + P^qs *¥xs = Llsixs ys = Lis iys

For Rotor

Vdr = Rridr + Wqr + PVdr (3) Vqr = Rriqr-Wdr + PVqr X¥xr = L]riXr Vyr = Llriyr

Flux Linkages equation as follows: For Stator side

^ds = (L]S+ Lm)ids + Lmidr ~¥qs = (Lis + lm)iqs + Imiqr

Vxs^lsixs (5)

^ys = L]si ys

For Rotor side

Vdr = (Llr + Lw)idr + Lmids ^qr = (Llr + lm)iqr + Imiqs

X?xr = Llrixr (6)

^ yr = Lfri yr

The torque can be expressed as: Te = pLmUdriqs ~ 'ds'dr) (7) {fTe-TLi (8)

where Ls = Lls + Lm (9)

Lp = L lr + ¿m

3. Space vector Modulation Scheme for Five-Phase VSI

The supply of multiphase machine always provided by voltage source inverter (VSI) because of flexibility in its increase number of legs, thus number of phase increases. A number of PWM technique are available to control the THD (%) output of VSI. Among all, SVPWM is most popular and widely adopted, as very easily implemented by digitally and better DC bus utilization, when compared to sinusoidal PWM (SPWM). The space vector representation is valid for both transient and steady state purpose while the phasor representation is valid only for steady state [8]. The concept of space vector is being used to synthesize the reference voltage vector by switching between neighboring vectors of a sector, such that volt-second principle employed. A five-phase vector offer total of 32 space vector, of which 30 are active vector. The thirty vector spanning over 360 degree in 2-D plane, forming a decagon with 10 sectors each with 36 degree. Five-phase SVM can be applied by using 2-vector method or 4-vector method. In 2-vector method, for synthesis only large vectors and zero vectors are used corresponds to their sector, where as both medium and large vectors along with zero vectors are used in 4-vector method. Compare to 2- vector, in 4-vector application dominant harmonic reduces significantly, and therefore is used for practical application. The reason is, the restriction of free flow of x-y components [9]. The expression for switching time using four-vector method can be find out by volt-second principle as follows.

V t =V ,t +Fh.t +V t +F, t

— s s —al al —bl bl —am ¡¡m —bm bm

2cosj &)

^L = II bl

t am bm kJ

\v \ = \vh =F = V =-V

| — am I I—bm\ \—m\ m ^ ac

The equation of switching time will be found out by solving equation (10), (11) & (12). These are as follows.

si T n

t al =-— (--)fsSin(-i-a)

tu = ,V sinta-tk - 1) ")

Vmsin,

\V s\ 1 n

t =-(-Vcsin(—¿-a)

am . * 2 5 ;

tbm= -J^ (-L)fss.„(a-(A -1) f)

Fmsin(-) L+T

0 s al bl am bm

Equations (13) & (14) are switching time of large and medium vector, used to construct the reference vector by two neighboring vectors. The placement of switching vector are such that, the large vector in d-q plane will be small vector in x-y plane. The concept is behind these switching to get sinusoidal voltage at output with less harmonic contains, as d-q components are responsible for main power and torque and x-y components leads losses to system. The placement of five-phases in both d-q & x-y plane are shown in fig. 1. Some constraints have to consider while application, (i) time of each vector can't be less than zero (ii) sum of time taken by both active and zero vector can't be exceeds the switching time. It is found that with these constraints, the maximum output voltage by inverter is 0.5257KDC which is approximately 16% less than 2-vector (large and zero vector) method. Like normal inverter operation, in order to keep switching frequency constant, each switching state should alter twice (once 'ON' to 'OFF' and again 'OFF' to 'ON' or vice-versa). This is called as symmetrical switching

(a) SWTCHING VECTORS ON d-q PLANE (b) SWTCHING VECTORS ON x-y PLANE

Fig. 1. Five-phase on both d-q & x-y plane

4. SVM-DTC with PI controller

The proposed DTC-SVM structure depicts in fig.2. The structure contains three PI regulator and a SVM unit. Among PI, one for speed and remaining are for flux and torque purpose, which are used to developed v^ & vq . PI controller are used for error correction and proper gain to the system of SVM. The equations of d-q components fed to the unit are as follows [10].

v<=(K~+K'Js )(<-*.) (15)

k= tor+^/'X7:*- (16)

Stator flux speed ws is calculated by taking two successful estimation of stator flux and Vs(k+i) as

= (VWJVWd - 1)) / (^nt.vfs ) (17)

Normally vd & vq are called flux and torque components of voltage vector, means flux of the system is controlled by vd where as vq is responsible towards torque of the unit. The SVM unit is used to control the inverter voltage smoothly. It receives d-q axis vector in stator flux reference plane and position of stator flux as input. The switching techniques have been explained in the above section. The reference voltage vector v*s is defined by its magnitude (|v*s|) and angle (P) represent by two adjacent active vectors and zero vectors. The expression of magnitude and angle subtended with respect to stator flux axis are as follows.

V . =, V

P = arctan

Taking practical constraints, there may be chance of getting higher input voltage at inverter terminal. Because of that, the errors flux error and torque error) fed to PI controller would be high, which might cause PI controller nonfunctional or saturates. At that time, the system will operates as classical DTC which is called as feed-forward DTC strategy. During this strategy, a single vector is applied throughout switching period and target will be reached quickly and minimizes the error.

Fig.2. Complete block diagram of 5-phase drive system with conventional PI controller 5. SVM-DTC with ANN controller

The complete block diagram SVM-DTC induction motor drive with ANN controller as shown in fig. 3. The practical difficulty with PI controllers have been addressed in previous section. The PI controllers are being replaced by ANN controller to get better response in speed, torque & flux. Torque error, flux error & position of stator flux are taken as input to ANN. These data are fed to 3 neuron of hidden layer. A 2-layer neural network been used in this system, where layer 1 is called hidden layer & layer 2 is called output layer, each with 20 no. of neurons. The three input connects with each of twenty neurons of hidden layer repeated till the output layer. A neuron consist of weight (product operator), bias (constant), sum (adder) and a linear or non-linear transfer function [11]. To maximize the performance, all weight to be set properly. In this paper back-propagation technique has been used as a training algorithm.

Fig.3. Complete block diagram of 5-phase drive system with ANN controller

6. Simulation Results and Discussion 15 10

O -5 -10 -15.

E 1000

E 1000

0.5 time (sec)

0.5 time (sec)

(a) (b)

Fig. 4. Comparison of flux variation (a) with PI (b) with ANN

0 0.5 1 0 0.5 1

time(sec) time(sec)

(a) (b)

Fig. 5. Comparison of Torque variation (a) with PI (b) with ANN

0 2 -2 0 2 flux-d flux-d

(a) (b)

Fig. 6. Comparison of Stator current variation (a) with PI (b) with ANN

0.5 10 0.5 1

time (sec) time (sec)

(a) fb)

Fig. 7. Comparison of Steady State Speed variation (a) with PI (b) with ANN

(a) (b)

F ig. 8. Comparison of Torque at zero speed and Full load (16 Nm) variation (a) with PI (b) with ANN

time(sec)

(a) fb)

Fig. 9. Comparison of Speed at zero speed and Full load (16 Nm) variation (a) with PI (b) with ANN

(a) (b)

Fig. 10. Response of Torque at + and - torque variation (a) with PI (b) with ANN

(a) (lb)

Fig. 11. Response of Speed at + and - torque variation (a) with PI (b) with ANN

torque ripple in ANN controller is improved by 17 % with respect to PI control and settling time of speed improved by 5.5%-6.5%. The steady state speed error has been reduced by 2-3 rpm with ANN controller. Different type of combination of cases are taken such as reference speed as zero with full load torque and positive & negative load torque variation. In above cases with ANN controller shows less ripple along with fast response compare to PI counterpart.

7. Conclusion

A new ANN based five-phase SVM-DTC is proposed for unifying flux and torque control to simplify overall drive system. The practical difficulty such as insensitive to non-linear dynamic response can be overcome with the implementation of ANN controller instead of conventional PI controller. It is seen that torque ripple reduced significantly along with less current ripple, more sinusoidal flux, and fast speed response. Future work can be extended by considering sensor less operation.

Steady State Torque Ripple (ANN compare to PI ) Improved by 17 %

Settling time of Speed Improved by 5.5 % to 6.5 %

Steady State Speed Error Reduced by 2-3 rpm (At rated load speed charges from 1479 to 1482 rpm )

8. Appendix. Parameters of 5-phase induction motor used in the simulation

Stator Resistance ion

Rotor Resistance 6.3 n

Stator leakage Reactance 0.04 H

Rotor Leakage Reactance 0.04 H

Magnetizing Reactance 0.42 H

Inertia 0.03 kgm2

Rated Torque 8.33 Nm

Number of Pole 4

Rated Phase Currents 2.1 A

Rated Phase Voltage 220 V

References

[1] Indian Power sector: [online]. Available: http://indianpowersector.com/home/about/.

[2] Emil Levi. Multi-phase machine for variable speed application. IEEE Trans. Ind. Elect. 2008:55:1893-1909.

[3] G.K Singh. Multi-phase induction machine drive research: A survey. Elect. Power Syst. Res. 2002:61:139-147

[4] Domenico Casadei, Francesco Profumo, Giovanni Serra and Angelo Tani. FOC and DTC: Two Viable Schemes for Induction Motors Torque Control. IEEE Tr. on power electronics. 2002: 17: 779-7.

[5] Haitham Abu-Rub, M. Rizwan Khan, Atif Iqbal and SK. Moin Ahmed. MRAS-Based Sensorless Control of a Five-Phase Induction Motor Drive With a Predictive Adaptive Model. IEEE International symposium on Industrial Electronics. 2010.

[6] Jose L., Azcue P. and Ernesto Ruppert. Three-phase Induction Motor DTC-SVM Scheme with Self-Tuning PI-Type Fuzzy Controller. International conference on FSKD: 2010.

[7] Haitham Abu-Rub, Atif Iqbal and Jaroslaw Guzinski. High performance control of ac drives with Matlab/Simulink models. UK : Wiley2012.

[8] A. Iqbal and E.Levi. Space Vector Modulation Schemes for a Five-Phase Voltage Source Inverter. European Conference on Power Electronics and Application. 2005.

[9] Hyung-Min Ryu, Jang-Hwan Kim and Seung-Ki Sul. Analysis of Multiphase Space Vector Pulse-Width Modulation Based on Multiple d-q Spaces Concept. IEEE Tr. on Power Electronics. 2005: 20: 1364-1371.

[10] Cristian Lascu, Ion Boldea and Frede Blaabjerg. A Modified Direct Torque Control for Induction Motor Sensorless Drive. IEEE Tr. on Industry Application. 2000: 36: 122-130.

[11] S. V. Jadhav, J. Kirankumar and B. N. Chaudhari. ANN based Intelligent Control of Induction Motor drive with Space Vector Modulated DTC. IEEE International Conference on Power Electronics, Drives and Energy Systems. 2012.