novel technique to reduce leakage current and commutation losses in electric drives

8/6/2019 Novel Technique to Reduce Leakage Current and Commutation Losses in Electric Drives

http://slidepdf.com/reader/full/novel-technique-to-reduce-leakage-current-and-commutation-losses-in-electric 1/5

8 t h I n t e r n a t i o n a l C o n f e r e n c e o n P o w e r E l e c t r o n i c s - E C C E A s i a

M a y 3 0 - J u n e 3 , 2 0 1 1 , T h e S h i l l a J e j u , K o r e a

9 7 8 - 1 - 6 1 2 8 4 - 9 5 7 - 7 / 1 1 / $ 2 6 . 0 0 © 2 0 1 1 I E E E

[ T u C 2 - 1 ]

Novel Technique to Reduce Leakage Currentand Commutation Losses in Electric Drives

H. Khan1,2

, E. Miliani1

, and K. Drissi2

1 IFP Energies Nouvelles, Reuil Malmaison, France2 UBP, LASMEA, Clermont-Ferrand, France

Abstract -- The continuous advancements in solid state

device engineering have considerably minimized the

switching transients for power switches. These fast switches

are gaining popularity because of lower switching losses for

hard switched systems. However this introduces high

leakage current in electric drives. A technique to

simultaneously reduce the leakage current and the switching

losses is presented in this paper.

Index Terms--Common-mode current, DPWM, Electric

Drives, Leakage current, IGBT (Insulated-Gate Bipolar

Transistor), SVM, Switching Losses.

I. I NTRODUCTION

The EMI (Electro-Magnetic Interference) aspect israrely taken into account at the conception of electricdrives which increases the cost and time to market of thedrivetrain due the tedious and tiresome process of EMIsuppression [1]. The problem of common mode current or leakage current arises due to the parasitic capacitivecoupling between the stator winding and the stator framealong with high rate of change of voltage 'dv/dt' in the

windings giving path to non negligible common-modecurrents (5A) for switching transients of 200ns for DClink voltages as low as 200V. The recent trend of usinghigh speed switches has only worsened the situation andhas created the need for an EM filter to eliminate thisnoise from the electric drive. In this paper, we present atechnique for reducing the leakage current as well as theswitching losses.

A. Leakage Current

A high frequency model of an electrical machineshows a parasitic capacitive coupling between the stator

end windings and the stator body [2]. Since the stator isconnected to the protective earth and the current leaksinto the earth, this is also called common mode current[4]. This is the main cause of the radiated noise [5]. Theexpression for the leakage current is given by (0.1)where Cws is about a few nano-farads for well constructedmachine while 'dv/dt' is about 5G V/s.

leakage ws

dv I c

dt =

(0.1)

The density of the leakage current depends on themodulation frequency f PWM, pulse positioning techniqueand the modulation technique used. The modulation

frequency determines how frequently the voltagecommutations occur, the pulse positioning gives the no.of voltage commutations for line voltage, for e.g. pulses

placed at the extremes of the modulation period the linevoltage have the same no. of voltage commutations as the

phase voltage while for pulses placed around the centre of the modulation period give twice as many voltagecommutations as the phase voltage whereas the type of modulation technique could alter the effective modulation

frequency such as in the case of DPWM techniques.These trends are demonstrated in Fig. 1 in which the

curves in green and navy blue are the phase voltage andcurrent respectively while the turquoise curve is theleakage current as observed on the test bench (Fig. 9).

Fig. 1: Leakage current (top), Ia and Va0

B. Switching Losses

In voltage source inverters, the phase voltage generatedhas only two levels Vdc/2 and –Vdc/2 with respect to theDC mid point and since they both have the same absolutevalue it becomes a constant while calculating theswitching losses. Hence the only variable is the phasecurrent [6]. Equation (0.2) gives the switching lossesover the fundamental period for a phase.

2

0

2

0

,

1( )

2 2

1 ( )2 2

on off

sw sw sw

on sw dc rise sw

sw dc fall off sw

P P P

where

f V t P i

f V t P i

π

π

π

π

= +

=

=

∫

∫

t dt

t dt

(0.2)



A simple yet very effective algorithm shown in Fig. 5 isdeveloped to reduce the switching losses as much as

possible using the boundary conditions represented by(0.5) to clamp a given inverter leg under balancedconditions. The algorithm takes into account the loadimbalance as well. If there is a load imbalance, one of the

phases will have currents of higher amplitude flowingthrough it than the other phases and hence higher switching losses and higher thermal stress on the switchesof that particular inverter leg which could lead to afailure, EDSVM can avoid such failures by non uniformclamping which would mean an inverter leg will rest idlefor a longer duration than the other legs.

_

2, ,

6 3 6

2 2,

3 6 3 6

20,3

a clamping

x for t

T o

x for t

where x

π π π

ω

π π π π

r

π ω π

π

⎧≤ < +⎪

⎪= ⎨

⎪⎪ − + ≤ < +⎩

⎡ ⎤∈⎢ ⎥⎣ ⎦

+

(0.5)

Fig. 5: Algorithm DSVM

The Fig. 6 shows the phase current and modulationfunction, it can be noticed how the modulation functionadapts itself to avoid switching where it would matter themost i.e. at high currents as explained above. A differentdead time correction algorithm is used with this techniquesince there are fewer commutations and different turn ON

and OFF times. The duty cycles are re-adjusted tocompensate for the slower commutations. For extremevalues (around 0 or 100) of duty cycle there is less marginfor the compensation hence lesser liberty to slow down theswitching without deforming the output voltage.

Auto-adjustment

Fig. 6: DSVM- Ia (green) & Modulation function for phase a

Fig. 7 shows theoretical results, reduction in theswitching losses, the leakage current magnitude and itsdensity for a modulation index of 0.907. The drop in theIl reduction around 60° (θ, not φ) is due to the presence of the third harmonic component in the phase voltage. Onlythe first quadrant is shown as the curve is symmetricabout φ=90°.

Get *abcv

Compute: dabc (DSVMMIN)

Get Îabc

Fig. 7 : Switching losses, leakage current magnitude and densityreduction

C. Dead Time

For DSVM, an inverter leg stays in the same state for several modulation periods and therefore no dead timeeffect would be observed for that leg and no correction isrequired. Moreover the inverter nonlinearities for extremevalues of duty cycles are also reduced, for e.g. for amodulation frequency of 20KHz and a dead time of 1µsduty cycles smaller than 2% and greater than 98% are asgood as 0% and 100% respectively. Thus, for the clamped

phase no compensation is required so standardcompensation techniques that adjust the voltage of the

phase according to the sign of the phase currents throughout the period cannot be used. The low frequencydistortion of the phase voltage is alleviated to some extent

for discontinuous modulation techniques.

i = = k

* * *max( , , )

ˆ ˆ ˆmax( , , )

i a b

k a b c

v v v

i i i i

=

=

cv

1 1 ( )

2o k i

o

d i i

o

x x d

sign vd

d d

d d

+ ×=

Δ = −

= + Δ

False

True

o

x xd d =



D. Drawbacks

DSVM has an indirect impact on the quality of the linevoltages. The effective line voltage modulation frequencyis twice that for standard SVM with centered pulse. It can

be seen in Fig. 8 that the line voltages have 2 pulses per PWM period for standard SVM and only one pulse for two of the line voltages for DSVM, where only one zerovector (V0) is used. It is clear that the volt-seconds for theline voltages is equal in both the cases. This makes theeffective modulation frequency smaller by 33.3% for DSVM. This increases the current ripple and in turn thetorque ripple.

Fig. 8 Line voltage wave form

The ability to slow down the switching speed using thisalgorithm without significantly increasing the switchinglosses depends on the load angle so the higher the loadangle the higher the liberty to reduce the leakage current.Hence this technique is dependent on the load angle.

III. EXPERIMENTAL SETUP AND R ESULTS

The test bench consists of two electric motorsmechanically coupled fed by two different inverterscontrolled independently, shown in Fig. 9. One electricdrive emulates a variable load to perform differentdriving cycles and the other drive is the system under observation which consists of a 15 kW, 3-phase 2-levelIGBT inverter fed by an adjustable DC link and a 3 kWPMSM with an incremental encoder (4096 points), anECU for rapid prototyping, and a buffer board for gateignal conditioning with incorporated dead time.s

Fig. 9: Test Bench

Fig. 10, in which the green curve is the voltage, navy

blue curve the current and the turquoise curve is theleakage current, demonstrates that for high currents thevoltage is clamped to either the positive or the negativeterminal. The leakage current density is now very sparsecompared to the results obtained using a standardtechnique shown in Fig. 1 where the leakage current

envelope is consistent and denser.

Fig. 10: DSVM Leakage current (top), Ia and Va0

Global inverter 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 thestandard technique. The losses in the IGBT drivers are nottaken into account. Table 1 shows results using EDSVM,as expected at higher switching frequencies we observean increase in the contrast between the two techniques interms of losses. The global inverter losses are further reduced which means that while the conduction lossesremain constant the switching losses increase withincrease in frequency. However, it should be noted thatthe absolute inverter losses increase with an increase inthe switching frequency.

TABLE IEXPERIMENTAL R ESULTS

Frequency KHz % Loss reduction15 12.9825 20.7535 22.85

IV. CONCLUSIONS

The experimental results show a decrease in globalinverter losses from 7.18% to 5.69% which translates toan overall reduction of inverter losses by 20.75% for aswitching frequency of 25kHz. Leakage current densitywas reduced by 33% with smaller peaks around phasecurrent zero crossing. Since the electric motor is adynamic inductive load, the load angle 'φ' varies with theoutput frequency or the machine speed and determinesthe extent to which the switching transients could beslowed to reduce the leakage current without increasingthe switching losses. However discontinuous modulationtechniques have some drawbacks too, so in the end acompromise has to be made between torque ripple andswitching losses.



R EFERENCES

[1] Firuz Zare, "EMI Issues in Modern Power Electronicsystems", IEEE EMC Society Newsletters, 2009; pp. 66-70;

[2] H. Akagi, S. Tamura, "A Passive EMI Filter for Eliminating Both Bearing Current and Ground LeakageCurrent From an Inverter-Driven Motor", IEEE

Transactions on Power Electronics, 2006.[3] N. Idir, Y. Weens, M. Moreau, J. J. Franchaud, “High-

Frequency Behaviour Models of AC Motors” IEEE

Transactions on Magnetics, Volume 45, Issue 1, Part 1, pp.133 - 138, Jan. 2009.

[4] Firuz Zare, "EMI in modern AC motor drive systems", IEEE EMC Society Newsletters, summer 2009; pp. 53-58.

[5] Sabine Marksell, "EMC Aspects of PWM ControlledLoads in Vehicles" PhD thesis. Department of IndustrialElectrical Engineering and Automation, LUND University,2004.

[6] J. W. Kolar, H. Ertl, and F. C. Zach “Influence of themodulation method on the conduction and switching losses

of a PWM converter system", IEEE Transactions On Industry Applications, 27, 1063–1075, 1991.

[7] G. Li, Z.Q. Liu, A. Golland, F. Wakeman, "New extra fastsoft recovery diodes and their applications", EuropeanConference on Power Electronics and Applications, Dresden, 2005.

[8] John W. Palmour, "Energy Efficient Wide BandgapDevices", Compound Semiconductor Integrated Circuit Symposium (CSIC), San Antonio, 2006.

[9] Ahmet M. Hava, Russel J. Kerkman, Thomas A. Lipo “AHigh-Performance Generalized Discontinuous PWMAlgorithm", IEEE Transactions On Industry Applications,VOL. 34, NO. 5, September/October 1998.

[10] Hamid Khan, Youssef Touzani, Khalil El KhamlichiDrissi, "Random Space Vector Modulation for ElectricDrives: A Digital Approach" 14th International Power Electronics and Motion Control Conference, EPE-PEMC 2010.

novel technique to reduce leakage current and commutation losses in electric drives

Documents