You are on page 1of 7

Evolutive Discontinuous SVM for Variable Speed Electric Drive

KHAN Hamid

Evolutive Discontinuous SVM for Variable Speed Electric Drive


H. Khan, E. Miliani, K. Drissi IFP Energies Nouvelles 1 avenue de bois prau, Rueil-Malmaison, France Tel.: +33 / (0) 147.52.57.83 Fax: +33 / (0) 147.52.70.12 E-Mail: hamid.khan@ifpen.fr URL: http://www.ifpen.com

Keywords
Pulse Width Modulation (PWM), Voltage Source Inverter (VSI), Switching losses, Electrical Drive, Modulation Strategy.

Abstract
Space Vector Modulation (SVM), with its many advantages over classical Pulse Width Modulation (PWM) techniques, is gaining popularity. Different Discontinuous-PWM with harmonic injections were developed during the past few years with the objective of reducing switching losses in power converters. However discontinuous modulation based on space vector calculations tends to be more complicated than injecting zero sequence voltage with the required phase angle. The proposed method extends the DPWM concept to tackle the temperature imbalance amongst the inverter legs and optimize switching losses for unbalanced loads. Switching losses also mean increased thermal stress on the power switches, which reduces the life of the power switches and requires an appropriate cooling system. A discontinuous auto-adjusting SVM technique intended for HEV oriented ElectricDrives is presented in this paper.

1 Introduction
The issue with Electric Vehicles is their weak autonomy, whereas for Hybrid Electric Vehicles (HEV) it is the limited space under the hood for another drive train; to improve the autonomy, the battery energy should be used efficiently, on the other hand lower losses would need a smaller and lighter cooling system. In hard switched power converters the switching losses are significantly high, and often represent the bulk of the inverter losses. The greater the switching losses, the greater the need for cooling the power switches. A new modulation technique is proposed in this paper to reduce the switching losses, which means efficient battery usage, reduced thermal stress and cooling effort that in turn means a lighter and compact power module with a higher life expectancy. The electric traction drive under consideration consists of a Permanent Magnet Synchronous Motor (PMSM) fed by a 3phase voltage source inverter. The Field Oriented Control (FOC) is used for torque regulation.

1.1 Switching losses


In VSI, the phase voltage generated has only two levels VDC/2 and VDC/2 with respect to the DC midpoint and since they both have the same absolute value it becomes a constant while calculating the switching losses; hence the only variable then is the phase current [1]. Equation 1 gives the switching losses over a fundamental period for a phase.

EPE 2011 - Birmingham

ISBN: 9789075815153

P.1

Evolutive Discontinuous SVM for Variable Speed Electric Drive

KHAN Hamid

on off Psw = Psw + Psw

where, f V t 1 P = sw dc rise 2 2
on sw off Psw = 2

i( ) d
0 2

(1)

f swVdct fall 1 2 2

i( ) d
0

and i ( ) = I m sin( + )

2 Discontinuous Space Vector Modulation


SVM is a digital Power Converter PWM technique where the duty cycle of inverter switches are calculated directly using mathematical transformations [2]. DSVM stands for Discontinuous SVM techniques [3]. For 3-wire 3-phase systems (isolated neutral) the phase current depends only on the line voltages, i.e. phase voltages don't directly dictate the current in the armature but their difference does. SVM is chosen over ordinary PWM techniques for the following reasons: better DC voltage utilization and hence lower conduction (copper) losses or a smaller machine for the same power rating and possibility of naturally lower switching losses. Reduced switching losses can be explained by the freedom that one has in generating the pulses when the duty cycles of active vectors are known beforehand by using only one of the two zero vectors for 3 phase 2 level inverters, V0 at the beginning and the end of the period or V7 in the middle of the period. Whereas PWM techniques are based on comparison of the reference signals and the carrier signal, one cannot avoid the insertion of the vector V7 at the middle of the period or V0 at the extremities. It can be seen as an intrinsically or naturally discontinuous-PWM technique i.e. without having to add a zero sequence voltage to the reference. An analogy can be drawn with the Generalized Discontinuous Modulator (GDPWM) [4] where the zero sequence voltage 'u0' is given by Equation 2:

u0 =

Vdc vmin 2

(2)

** * So the new reference signal becomes: vx = vx + u0 . For a balanced 3 phase sinusoidal system the new reference signals would be clamped to the negative DC rail for one-third of the fundamental period. This technique is also known as DPWMMIN so we should refer to the space vector version as DSVMMIN. The switching function waveform can be seen in Fig.1. Similarly for DPWMMAX the zero sequence voltage is given by Equation 3.

Fig. 1. DSVMMIN: Switching Function leg A; mi=0.907

u0 =

Vdc vmax 2

(3)

EPE 2011 - Birmingham

ISBN: 9789075815153

P.2

Evolutive Discontinuous SVM for Variable Speed Electric Drive

KHAN Hamid

SVM increases the achievable amplitude of the phase voltages by 15.47% compared to ordinary PWM techniques. There are techniques such as harmonic injection (3rd and its multiple), to increase the linearity of PWM techniques [5]. However these techniques would not be accurate for dynamic systems where the fundamental component of the voltage varies with time in amplitude and frequency brusquely, hence vector control is better suited since instantaneous values are used and the wave form of the reference signal is of no significant importance. 2.1 Phase clamping
In spite of the many advantages of SVM, its drawback is the switching losses. DPWM is commonly used to attenuate this problem by adding zero sequence voltages to saturate the phase voltage reference to either VDC/2 and VDC/2 in positive or/and negative half cycles respectively. Equation 4 must be respected while doing so. Let the phase to be clamped be 'x', i.e. a phase can only be clamped if its instantaneous value is either the largest or the smallest.

v* = vmax vmin x
2.2 State of the art

(4)

There are many papers on DPWM and the most common of these techniques can be found under the names DPWM0, DPWM1, DPWM2 and DPWM3 [6]; the clamping is obtained by adding a zero sequence voltage to voltage references given by Equation 5.

u0 = sign(vmax ) where,

Vdc vmax 2
(5)

* * * vmax = max ( va (t + ) , vb (t + ) , vc (t + ) )

{30,0,30}
This technique works well for static loads and hence is optimal only while feeding a static load with a known load angle [7]. The only work on Discontinuous SVM close to technique proposed in this paper known to us is found in [8], however we obtain different results specially the switching loss curve, our result are shown in Fig. 4 and unlike the technique presented there, the modulation function presented in this paper does not depend on the Modulation Index (mi) as at low mi high common mode voltage is observed so the DC link is adjusted to maintain a high mi and moreover at relatively lower mi high armature current is drawn, e.g. starting torque requires high current but low voltage. The DSVM algorithms found in literature are complex compared to the one proposed in this paper which does not require sector , voltage and current phase calculation. The EDSVM algorithm, in spite of being simpler, takes into account the load unbalance to minimize the losses even further and even increase life expectancy of the inverter.

3 Evolutive Discontinuous Space Vector Modulation


EDSVM is a new solution proposed to minimize the switching losses for inverter fed dynamic loads or variable speed drives. It is a very generic method that makes full use the entire degree of freedom for phase clamping, i.e. depending on the load characteristics this method can lead to different types of clamping strategies, symmetrical or unsymmetrical about a phase or can have either equal or unequal

EPE 2011 - Birmingham

ISBN: 9789075815153

P.3

Evolutive Discontinuous SVM for Variable Speed Electric Drive

KHAN Hamid

clamping durations for different phases as expressed by Equation 6 and Equation 7 respectively for phase 'a', the expressions for phases 'b' and 'c' will be out of phase by 2/3 respectively.

2 Ca _ +ve , for 6 t < 3 + 6 , Ta _ clamping = or (6) 2 Ca _ ve , for + t < + + 6 3 6 2 2 where Ca _ +ve , Ca _ ve 0, & Ca _ +ve + Ca _ ve = 3 3 Normally Ca _ +ve = Ca _ ve = for symmetrical clamping about a phase as in techniques mentioned by 3 Equation 4.

4 2 Ca _ +ve , Ca _ ve 0, & Ca _ +ve + Ca _ ve 3 3

(7)

Equation 6 shows the boundary conditions for a phase, for both positive and negative half cycles over a complete fundamental period. The maximum clamping duration for a three phase system is Ctotal = 2 , hence the effective switching frequency would be reduced to two-thirds. The EDSVM algorithm allows optimizing either the switching losses or can be used to regulate the temperatures of the inverter legs if there is a difference higher than a certain specified value. The importance of this method can be appreciated in cases where there is a slight imbalance in the load or machine phase winding are short e.g. turn-turn shorts that cause an imbalance and hence different switching losses in the inverter phases, this leads to a rapid destruction of the inverter, the proposed method can extend the life of the inverter before the fault could be repaired by allocating the concerned phase the appropriate clamping to avoid complete failure of the corresponding inverter leg. An exaggerated case can be seen in Fig. 2; it can be noticed that all the phases have different clamping durations, phase 'a' being allocated the most and phase 'c' the least.

Fig. 2 EDSVM: unsymmetrical clamping We've used the most simple discontinuous SVM technique, referred to above as DSVMMIN, as the basic technique to complement the generalized modulator in terms of the load angle . EDSVM

EPE 2011 - Birmingham

ISBN: 9789075815153

P.4

Evolutive Discontinuous SVM for Variable Speed Electric Drive

KHAN Hamid

Algorithm is very simple and intelligent with almost no extra burden on the processor, first of all the duty cycles are calculated for the given reference voltages (generated by the torque control algorithm) using the algorithm DSVMMIN. Now to optimally clamp the inverter legs we use the measured phase currents (this makes it a modulation technique with current feed-back). DSVMMIN was chosen over other DSVM techniques because in our case (the inverter is fed by a battery) the zero vector 'V0' does not generate any common mode voltage whereas vector 'V7' generates the maximum common mode voltage equal to VDC.

Fig. 3 Intelligent Switching Function - >30 Fig. 3 shows how EDSVM modifies the modulation function of a phase for power factor angle greater than 30. It repartitions the clamping zone of DSVMMIN when ever there is a possibility of minimizing the losses.

3.1 EDSVM switching losses


The switching losses for the modulation technique proposed in this paper are calculated using Equation 1, which assumes linear relationship between switching losses and the switched currents. Fig. 4 shows the theoretical switching loss reduction while feeding a load with a variable power factor angle with the proposed method compared to the standard continuous SVM. It should be noted that the curve is symmetric about 0 and 90 and hence could be extended to -180 to 180.

Fig. 4. ADSVM: Switching loss reduction

3.2 Simulation Results


All the simulations are done on MATLAB/SIMULINK. The inverter, the Permanent Magnet Synchronous Machine (PMSM) and the DC source are taken from the SimPowerSystems library. A torque drive is simulated using a PMSM supplied by a battery via a 3 phase inverter. The control strategy used is Field Oriented Control (FOC), with the proposed modulation technique EDSVM. The simulation models are completely discretized, with the simulation step being 100s, the proposed technique was compared against the standard technique for different speed-torque cycles. The results perfectly superimposed for both the techniques.

EPE 2011 - Birmingham

ISBN: 9789075815153

P.5

Evolutive Discontinuous SVM for Variable Speed Electric Drive

KHAN Hamid

Fig. 5 EDSVM: Stator Current Evolution Fig. 5 shows the phase currents in the alpha-beta coordinates, the outer circles represent the starting of the machine, high torque hence high currents with the inner circle representing high speed and low torque. Nothing unusual can be noticed, the torque regulation was comparable to continuous SVM. The gating signal and the phase current in shown in the Fig. 6, it can be seen that the voltage clamping occurs at high currents.

Fig. 6 EDSVM - Ia, ga

4 Test bench
The test bench consists of two electric motors mechanically coupled fed by two different inverters controlled independently Fig. 7. One electric drive emulates a variable load to perform different driving cycles and the other drive under observation consists of a 15 kW, 3-phase 2-level IGBT inverter fed by 350V DC and a 3 kW PMSM with an incremental encoder (4096 points), an ECU for rapid prototyping, a buffer board for gate signal conditioning with incorporated dead time.

Fig. 7 Test Bench

EPE 2011 - Birmingham

ISBN: 9789075815153

P.6

Evolutive Discontinuous SVM for Variable Speed Electric Drive

KHAN Hamid

5 Experimental Results
Global losses are calculated measuring the power supplied by the DC link and comparing it with the power at the inverter legs for the proposed and the standard techniques. The analysis is done over two complete fundamental electrical period with, high frequency data acquisition is done through an oscilloscope to get accurate. The analysis takes into account the DC link capacitor losses as well. The experimental results show a decrease in losses from 7.18% to 5.69% which translates to an overall reduction of 20.75%. However comparing only the switching losses accurately is practically impossible for fast switching transients (dv/dt ~ 300M v/s). However no tests under unbalanced conditions could be undertaken. The purpose of this paper was to present the switching losses as the most prominent source of loss which could be reduced with cheap and easy to implement techniques with no extra hardware needs. Table I: EDSVM global inverter loss reduction
Fswitching (kHz) 15 25 35 Loss reduction (%) 12.98 20.75 22.82

Conclusion
A digital discontinuous modulation technique is presented in this paper that is more flexible than conventional discontinuous methods. It is an extension of the DPWM strategies, which would not only minimize the inverter switching losses under standard conditions but can also regulate the temperature imbalance amongst the inverter legs and further optimize switching losses for unbalanced loads. Moreover it can also increase the life expectancy of the system by avoiding the weakest link failure by letting the most vulnerable inverter leg rest longer than the other legs.

References
[1] J. W. Kolar, H. Ertl, and F. C. Zach Influence of the modulation method on the conduction and switching losses of a PWM converter system", IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, 27, 1063 1075, 1991. [2] H. W. van der Broeck, H. Ch. Skudelny, and G. Stanke, Analysis and realization of a pulse width modulator based on voltage space vectors", IEEE TRANSACTIONS ON POWER ELECTRONICS, . 24, 142150, 1988. [3] Luca Dalessandro, Simon D. Round, Uwe Drofenik, Johann W. Kolar, Discontinuous Space-Vector Modulation for Three-Level PWM Rectifiers", IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 23, NO. 2, MARCH 2008. [4] H van der Broeck, Analysis of the harmonics in voltage fed inverter drives caused by PWM schemes with discontinuous switching operation", Proc. EPE Conf., pp. 261266, 1991. [5] Keliang Zhou, Danwei Wang, Relationship between space-vector modulation and three-phase carrier-based PWM: a comprehensive analysis ", Industrial Electronics, IEEE Transactions, Sch. of Electr. & Electron. Eng., Nanyang Technol. Univ., vol. 49 pp. 186-196, Feb 2002. [6] Ahmet M. Hava, Russel J. Kerkman, Thomas A. Lipo A High-Performance Generalized Discontinuous PWM Algorithm", IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 34, NO. 5, SEPTEMBER/OCTOBER 1998. [7] Andrzej M. Trzynadlowski, Stanislaw Legowski Minimum-Loss Vector PWM Strategy for Three-phase Inverters ", Electronics Letters 1997, IEEE, vol. A4 pp.108-117, September 2008. [8] Marian P. Kazmierkowski, R. Krishnan, Frede Blaabjerg Control In Power Electronics Selected Problems, Academic Press 2002 ISBN-0-12-402772-5 pg. 104-110.

EPE 2011 - Birmingham

ISBN: 9789075815153

P.7