Puesta a Tierra y Drives

7/31/2019 Puesta a Tierra y Drives

1/15

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 37, NO. 5, SEPTEMBER/OCTOBER 2001 1423

Effect of Adjustable-Speed Drives on the Operationof Low-Voltage Ground-Fault Indicators

Gary L. Skibinski, Barry M. Wood, Senior Member, IEEE, John J. Nichols, Member, IEEE, andLouis A. Barrios, Senior Member, IEEE

AbstractOver the years, the petroleum and chemical industryhas found increasing favor with 60-Hz low-voltage ( 600 Vac)power systems that utilize a high-resistance grounded (HRG)neutral philosophy. Historically, the older generation of ad-

justable-speed drives (ASDs) had little or no effect on the normaloperation of ground-fault indicators (GFIs) used with the installedHRG systems. This paper focuses its investigation into nuisanceGFI alarms that may occur when present-generation ASDs areretrofitted into the existing plant. The paper first reviews possibleneutral grounding systems, with emphasis on the types of HRGsystems possible and GFI alarm philosophy. The paper then dis-cusses how ASDs may generate zero-sequence high-frequency noisecurrents in the HRG neutral circuit, which may cause nuisanceground-fault alarms and potentially mask a legitimate groundfault. GFI noise current magnitude is defined for both presentand older ASD technologies (e.g., insulated gate bipolar transistorversus bipolar junction transistor drives). The effect this transientzero-sequence noise current magnitude has on GFI operation isdescribed. Mitigationmethodsused at thedriveto reduceASD noisecurrent magnitude to acceptable nonalarm levels is investigated.Filter solutionslocated at the HRG/GFI meter that reduce nuisancealarms are also investigated. The pros and cons of at the drive or atthe meter filter solutions are supported with laboratory and fieldtest data. Application guidelines are given to help avoid nuisanceproblems with a plant ground-fault protection scheme, whichneeds to successfully operate in the presence of multiple ASDs.

Index TermsAdjustable-speed drives, common-mode noise,electromagnetic interference filters, ground-fault indicators, high-

resistance neutral ground, pulsewidth modulation, zero-sequencecurrent.

I. INTRODUCTION

A. System Grounding Philosophy

SYSTEM grounding philosophy for process industry ap-

plications is based on providing safe and reliable power

distribution, while insuring maximum protection against

Paper PID 002, presented at the 1999 IEEE Petroleum and Chemical In-

dustry Technical Conference, San Diego, CA, September 1315, and approvedfor publication in the IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS by thePetroleum and Chemical Industry Committee of the IEEE Industry ApplicationsSociety. Manuscript submitted for review September 15, 1999 and released forpublication June 5, 2001.

G. L. Skibinski is with Standard Drives, Rockwell AutomationAllen-Bradley Company, Inc.,Mequon, WI 53092 USA (e-mail: [email protected]).

B. M. Wood is with Chevron Research and Technology Company, Richmond,CA 94802 USA (e-mail: [email protected]).

J. J. Nichols is with the Wood River Refinery, Tosco Refining Company,Roxana, IL 62084 USA (e-mail: [email protected]).

L. A. Barrios is with Equilon Technology, Houston, TX 77082-3101 USA(e-mail: [email protected]).

Publisher Item Identifier S 0093-9994(01)08316-5.

Fig. 1. Typical neutral grounding systems.

transient voltages and maintaining uptime availability during

ground faults [1].

An advantage of the ungrounded system of Fig. 1 is that a

single line-to-ground fault does not require immediate inter-

ruption of power flow. A disadvantage is that primary line-to-

groundvoltage transients arepassed to the secondarywithout at-

tenuation. Impulse testing of a transformer with an ungrounded

neutral shows a 2000-V-peak 1.2 s/50 s impulse test applied

line to ground on the transformer primary is virtually unatten-

uated line to ground on the secondary side. Another disadvan-

tage is that neutral point is capacitively coupled to ground,

allowing voltage to float toward the line voltage during tran-

sients and overstress the line-to-neutral insulation system. The

principal problem with an ungrounded neutral system is that an

arcing ground fault can cause escalation of the system line-to-

ground voltages to several times normal line-to-ground voltage.

This occurs as the arcing fault alternatively extinguishes and

reestablishes itself, successively trapping a higher charge on thesystem shunt capacitance each time. This causes displacement

of thesystem neutral, which is only connected to ground through

the shunt capacitance, to a voltage that is several times normal

line-to-ground voltage. Voltages to ground which are sufficient

to cause insulation failures, especially in motors, are very pos-

sible [2].

An advantage of grounded wye systems in Fig. 1 is a

5 : 1 10 : 1 attenuation of transformer primary line-to-ground

voltage transients. Impulse testing of a transformer with a

grounded neutral shows a 2000-V-peak 1.2 s/50 s impulse

00939994/01$10.00 2001 IEEE
http://-/?-http://-/?-http://-/?-http://-/?-


2/15

1424 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 37, NO. 5, SEPTEMBER/OCTOBER 2001

test applied line-to-ground on the transformer primary had only

a 200-V-peak line-to-ground voltage on the secondary side.

Also, an neutral solidly connected to earth ground potential

reduces the possibility of overstressing line to neutral insula-

tion. A disadvantage is that a line-to-ground fault will interrupt

power to the process line. With a solidly grounded neutral

system, the problem with line-to-ground voltage escalation due

to an arcing ground fault is eliminated, since the neutral is heldto ground potential. However, the disadvantage is that a ground

fault must now be immediately interrupted due to the higher

ground fault current. The load being supplied by this circuit is

immediately shut down without warning. The higher ground

fault current also has the possibility to inflict more damage due

to the higher energy to be dissipated. Other circuits may also be

disrupted due to the severe voltage drop caused by the higher

fault current.

The high-resistance grounded (HRG) system of Fig. 1 adds

sufficient neutral-to-ground resistance to limit the ground-fault

current to a value greater than or equal to the system capacitive

charging current, which is normally in the range of 15 A.

Although the neutral is not held at earth ground potential as

with the solidly grounded system, this value of resistance will

limit the voltages to ground during an arcing ground fault to

levels sufficiently low to avoid insulation failures. Secondary

attenuation of primary line-to-ground voltage transients de-

pends on the resistor value chosen, which is outside the scope

of this paper. The largest advantage of the high-resistance

neutral grounding method is that a line-to-ground fault does

not require immediate shutdown of the affected equipment.

When the ground-fault alarm is initiated, an orderly shutdown

of the affected equipment can be organized within a reasonable

amount of time period, usually 24 h. The ground fault still must

be given attention, to avoid escalation into a higher magnitudephase-to-phase fault if another ground fault should occur on a

different phase.

A detriment of HRG systems is additional equipment cost

over a solidly grounded or ungrounded system. The fault on an

HRG system may be more difficult to locate than on a solidly

grounded system because there is usually minimal damage and

the individual affected circuit is not automatically isolated by

circuit breaker or fuse action. The HRG system usually has a

current pulsing scheme to enable tracking down the fault with

a clamp-on ammeter. However, due to the benefits of providing

for continuous operation, minimum ground-fault damage, and

avoiding voltage escalation during ground faults, the HRG

system, used in conjunction with a ground-fault indicator (GFI)meter, has become the standard practice in the petrochemical

process industry.

B. GFI Alarm Philosophy

Voltage-sensing and current-sensing GFI schemes are shown

in Fig. 2. In voltage-sensing GFIs, voltage across -to-ground

is sensed with a high-impedance voltmeter. The high input

impedance of the GFI meter insures the desired ground-fault

current trip point is not desensitized by a parallel current path

through the meter.

(a) (b) (c)

Fig. 2. Ground-fault detection schemes with HRG/GFI systems. (a) Voltagesensing. (b) Current sensing. (c) GFI sensing ungrounded system.

In the Fig. 2 current-sensing GFI scheme, current through the

grounding resistor passes through a current transformer (CT)

and is sensed by a low-impedance ammeter on the CT sec-

ondary. The low input impedance of the GFI ammeter insures

essentially allof -to-groundvoltage is across the resistor and

not the meter, so that the desired ground fault current trip pointis not desensitized.

Various GFI alarm philosophies exist for both GFI schemes.

Maximum rms fault current ( ) is typically sized for 15 A

and is defined by ( ). is the neutral grounding re-

sistor and is rated rms line-to-neutral voltage. Maximum

voltage across to groundis , which is 277 V on a 480-V

system. In general, it is desirable to set an alarm trip point as

low as possible without nuisance tripping.

In a balanced three-phase system withoutASDs, neutral

voltage is theoretically near zero potential, so a GFI voltmeter

would normally read 0 V and a GFI ammeter scheme would

read 0 A. A relatively low GFI alarm level ( 10% of the max-

imum voltage or current value) is desirable to provide sensi-tive ground-fault detection, especially for wye-connected loads

where a failure near the neutral point results in relatively little

fault current. In practice, line-to-ground 60-Hz capacitive cable

charging current, 60-Hz nonlinear loads such as transformer

magnetizing current or fluorescent lighting ballasts wired line

to ground, and unbalanced loads may exist. These parasitics are

discernible on most voltmeters. As a result, a minimum prac-

tical voltage trip point is usually in the 10-V range. GFI volt-

meters usually contain high-frequency filters or time delays to

prevent false alarms during normal system transients that occur,

e.g., across the line motor starting. A GFI ammeter approach

is somewhat less sensitive in that neutral current for a typicaland 10-V -to-ground voltage is barely seen

on the 05-A scale of the GFI. Historically, a minimum pickup

setting of 10% of maximum voltage or current has proven sat-

isfactory for most applications.

In a balanced three-phase system with ASDs, neutral

voltage is not close to zero potential, so nuisance alarms and

masking of legitimate faults on existing standard HRG and GFI

system is possible. This paper discusses how ASDs generate

transient zero-sequence current, which, in turn, creates a

nonzero voltage at the GFI. The effect the ASD transient

noise current has on GFI and methods to mitigate the effect of

this noise source on existing HRG systems are described.


3/15
http://-/?-


4/15


Fig. 6. Expanded voltage and charging current at kHz (0at offset 3 div; 50 s/div; 2 A/div; 100 V/div).

current waveform is the combined zero-sequence or CM line-to-

ground current from all ASD phases. The 8-Apk current

spike of Fig. 6 is typical of systems using solidly grounded

transformers. Exact calculation of magnitude is discussed

in [4]. Although kHz in Fig. 6, the current spike repe-

tition rate into the ground circuit is closer to 20 kHz, because

every phase switching contributes line-to-ground current tran-

sients. Thus, the repetition rate is estimated as .

B. Zero-Sequence Current of Past-Generation BJT Drives

Previous-generation ASDs used BJT semiconductors withrise/fall times of 12 s. For the same system capacitance,magnitude is lower than in IGBT drives because of thelower output . Also, BJT drives had a maximum of 750Hz1.5 kHz. Thus, reduced current and lower explainswhy there were fewer problems of nuisance GFI alarms withpast-generation drives.

III. EFFECT OF ASD ZERO-SEQUENCE CURRENT ON GFIs

System grounding philosophy affects the ASD tran-sient zero-sequence current magnitude and circuit path. Theungrounded system of Fig. 1 breaks the return path backto the ASD in Fig. 3. Thus, and CM noise voltage(e.g., ) developed across the ground grid of Fig. 3 doesnot exist. The grounded wye system of Fig. 3 detrimentallycompletes a transient zero-sequence noise current return pathas previously discussed. current is highest with groundedsystems. However, the current is contained totally withinthe transformers solidly grounded neutral ( ) and secondarycircuit. zero-sequence current cannot flow into the trans-former primary circuit of Fig. 3. Thus, the benefit of a solidly

grounded transformer of Fig. 3 is that it controls the returnpath to the ASD and insures that other sensitive equipment orplant HRG unit connected on the primary side are not affected.

An HRG system has 55277 (15 A) in series with thezero-sequence noise current return path which significantly re-duces peak current. Thus, CM voltage differences acrossthe ground grid are smaller than with solidly grounded systems.The bottom trace of Fig. 7 shows that in a balanced three-phasesystem with the ASDs off, the neutral point voltage at isclose to zero potential. The top trace of Fig. 7 shows that in a bal-anced three-phase system with ASDs operating, the neutralpoint voltage is not close to zero potential, which is due to theASD zero-sequence current developing a transient voltage

Fig. 7. Voltage across with ASD operating at kHz and with ASDoff. Top trace: with ASD operating (0 level at offset 4, 50 V/div, 1 ms/div).Bottom trace: with ASD off (0 level at offset 1, 50 V/div, 1 ms/div).

across the high resistance in to ground. The GFI meter usu-ally has insufficient filter action to prevent false alarms duringthe voltage spike intervals induced by ASD transients. As aresult, the initial desired GFI alarm/trip point must be substan-tially increased to prevent false or nuisance trips. Consider the

single ASD operating waveform of the top trace of Fig. 7 whichcontains a high-frequency rms voltage of 45 Vrms and voltagespikes up to 175 Vpk across . It is unclear where the GFIalarm must be set. The problem is that the GFI set point mayhave to be so high that it cannot reliably detect a ground faultwhen theASDs areoperating. Also, unlessprecautionsare takenin rating the neutral grounding resistor, the power dissipation ofthe ASD high-frequency component in the resistor may exceedthe power rating, especially when combined with the added fun-damental frequency component during a ground fault.

IV. METHODS TO REDUCE CURRENT IN THE NEUTRAL

There are a few methods that minimize, but may not elim-inate, ASD zero-sequence current and its effect on the ASD-generated voltage developed across the neutral grounding re-sistor.

Lower the carrier frequency . Reduce the number of ASDs on a system transformer. Operate the ASD closer to 60 Hz. Add CM chokes to drive output. Add output line reactor. Add an electromagnetic interference (EMI) filter to ASD

input. Add CM capacitors to each ASD dc bus. Investigate lighting surge capacitors in plant.

Use shielded or armor cable. Use a drive isolation transformer.

Lowering the carrier frequency to the minimum allowablevalue (typically 2 kHz) will increase the time between tran-sient current spikes in Figs. 6 and the top trace of 7, sothat the GFI meter responds to a lower effective RMS noisevoltage. Section VII-D test results of a single ASD drive showthe ASD-generated voltage read on the GFI meter is 1/41/6lower at kHz as compared to 6 kHz.

Decreasing the number of ASDs on a transformer is imprac-tical and may not reduce the -to-ground voltage to accept-able GFI levels. The GFI meter readout obtained under singledrive operation may not necessarily increase as the number of
http://-/?-http://-/?-


5/15


6/15


Fig. 11. Input impedance and phase angle versus frequency of GFI voltmeter.

The above methods attempted to reduce the ASD zero-se-quence current in the -to-ground circuit, so that groundresistor voltage is reduced and lower GFI alarm set points areallowed. Each method has a performance versus cost expense

associated with it that must be evaluated. The following sec-tions investigate allowing system ASD current to flow, butsolving the GFI readout problem with filtering at the GFI meter.

V. CHARACTERIZATION OF VOLTAGE AND CURRENT GFIS

Prior to designing a high-frequency zero-sequence noise filterbetween the HRG and GFI, it is necessary to characterize thefrequency response of the respective GFIs.

A. Characterization of Voltage-Sensing GFI

Fig. 11 plots input impedance versus. frequency of a typ-ical GFI voltmeter shown in Fig. 2(a). The meter looks like asimple lag filter with 60-Hz impedance of 620 k and cornercutoff frequency of 20 kHz. Meter input impedance at 60 Hz ishigh, as compared to , and avoids desensitizing the 60-Hzground-fault alarm set point. The 20-kHz corner frequency im-plies an 8- s filter time constant. Thus, some filtering of the

rise-time component exists, but the ASD zero-sequencevoltage at willoverloadthe meter, causing GFI readoutproblems.

B. Characterization of Current-Sensing GFI

Frequency response of a typical current-sensing GFI systemof Fig. 2(b) was evaluated at a field site. System voltage was

480 Vac. was rated at 130 so that 60-Hz ground-faultcurrent was limited to 2 Arms. The CT had a unity primary-to-secondary ratio (5 A : 5 A) and fed a digital GFI with a 1-A alarmsetting and 3-s delay. The GFI digital meter read 0 A, evenwith 15 PWM IGBT ac drives powered on the feed transformer.

Fig. 12 shows an ASD-induced 75-Vpk voltage developedacross the neutral grounding resistor (0.5 Apk in CT primary),while the CT secondary response read zero current. The CT ironlamination thickness, chosen for 60-Hz frequency, could not re-spond to the high-frequency ASD transient current.

Thus, some GFI current-sensing schemes may inadver-tently solve the GFI readout problem under ASD operation bytheir inherent high-frequency rejection characteristics.

(a)

(b)

Fig. 12. Neutral grounding resistor voltage and CT secondary current ofcurrent-sensing GFI system measured at field site with multiple PWM IGBTdrives operating between 3060 Hz with kHz. (a) Voltage acrossneutral grounding resistor (0 at offset 5, 25 V/div). Time scale: 200 s/div.(b) CT secondary current (0 at offset 2, 0.2 A/div). Time scale: 2 ms/div.

Fig. 13. GFI meter plus high-frequency bypass capacitor.

VI. METHODS TO REDUCE ZERO-SEQUENCE VOLTAGE AT GFI

The advantages and disadvantages of the following solutions

used at the GFI meter location to reduce the effect of zero-se-

quence voltage are discussed:

high-frequency bypass capacitor across grounding resistor

and GFI meter;

low-pass filter between grounding resistor and GFI

voltmeter;

impedance-buffered filter between grounding re-

sistor and GFI meter.

Fig. 13 shows a high-frequency bypass capacitor placed in

parallel with the grounding resistor and GFI meter. There is a


7/15


8/15


(a)

(b)

Fig. 16. (a) Top trace: applied 60Hz voltage across HILO filter input

terminals; middle trace: voltage across terminals ; bottom trace:voltage across terminals . (b) Applied 60-Hz voltage across filterinput terminals HILO and filter output terminals HILO.

VII. IMPEDANCE-BUFFERED LOW-PASS FILTER DESIGN

A. Conceptual Design of the Filter

The input section of Fig. 15 consists of a voltage step-downautotransformer . A standard 480-V/120-V control trans-former is reconnected as an autotransformer with a 5 : 1 ratio.Fig. 16(a) shows 60-Hz voltage applied to the filter inputterminals, with 80% of the input voltage across ofand 20% across of .

Voltage across is applied to a low-pass filter sectionconsisting of , , and . Low-voltage capacitor and inductorcomponents rated for 600 Vac may handle much higher tran-sient voltages across the grounding resistor because of 1 : 5step-down ratio. Transformer provides impedance bufferingand is critical in determining the low-pass . Low-pass sectioncomponents along with the parameters set a cutoff frequency

Hz. Filter inputoutput attenuation follows the stan-dard 40 dB per decade past .

Capacitor voltage is applied to voltage step-up auto-transformer , which is a standard 120-V/480-V controltransformer reconnected as an autotransformer with a 1 : 5ratio. Transformer also acts as an impedance buffer, in case

the output terminals are shorted together.Fig. 16(b) shows there is no input-to-output attenuation ofthe 60-Hz component through the filter. Voltage applied to theHILO filter input terminals is stepped downed, passed throughthe low-pass section, and stepped up back to the original inputvoltage magnitude at the filter output terminals HILO. Thesealed filter box physical dimension is 9 in 6 in 4 in andmay be placed at the HRG location.

B. Filter InputOutput Attenuation Versus Frequency

Fig. 17 is a plot of filter input-to-outputattenuation versus fre-quency. Fundamental frequencies between 0100 Hz are not at-tenuated. Attenuation of ( ) 40 dB per decade exists beyond the

Fig. 17. Measured inputoutput attenuation of GFI filter versus frequency.

Fig. 18. Measured input impedance magnitude and phase angle versusfrequency of the impedance-buffered low-pass GFI filter.

95-Hz cutoff frequency up to 1000 Hz. Third harmonic (180 Hz)voltage in the transformer neutral is attenuated by ( ) 14 dB, ifit exists.

Attenuation greater than ( ) 40 dB/decade occurs forfrequencies 1000 Hz due to high-frequency skin effects,substantially increasing the ac resistance of the inductor andtransformer coils. The ambient noise floor [( ) 85 dB] isreached at 620 kHz. At 20 kHz and greater frequencies,the inductance of the iron components is reduced to air corevalues. A high-frequency magnetic skin-effect phenomenonreduces the effective permeability of the iron lamination to thepermeability of air. Beyond 20 kHz, the gradual decrease inattenuation versus frequency is due to parasitic winding ca-pacitance of the magnetic components becoming predominantalong with the coil skin-effect ac resistance. The followingdiscussion identifies the critical ASD noise frequencies, which


9/15

SKIBINSKI et al.: EFFECT OF ASDs ON THE OPERATION OF LOW-VOLTAGE GFIs 1431

TABLE IGFI METER READINGS WHEN USED ALONE, WITH A BYPASS CAPACITOR FILTER CIRCUIT, AND WITH AN IMPEDANCE-BUFFERED LOW-PASS FILTER CIRCUIT

(a)

(b)

Fig. 19. (a) Single 2-hp ASD operating at kHz, Hz, 30-ftcable. Top trace: -to-ground voltage (0 level at offset 4, 50 V/div, 1ms/div); bottom trace: filter output (0 level at offset 1, 50 V/div, 1 ms/div).(b) Time scale expansion of (a) but at 50 s/div.

need to be attenuated by the filter before entering the GFImeter. Corresponding filter attenuation at these frequencies is

obtained from Fig. 17. Conversion of decibel-to-attenuationratio may be done using .

The predominant frequency of ASD-induced voltage acrossthe grounding resistor that corrupts the GFI readout is the carrierfrequency and frequency components of Fig. 6. Re-sults ofFig. 17show a 80-dB to 85-dB reduction (1: 17800)and total removal of these components prior to the GFI meterconnection.

A second ASD frequency component developed across thegrounding resistor is that due to the IGBT 50200-ns rise timesshown in Fig. 4. Equivalent frequency ( ) relative to rise timeis obtained from Fourier analysis of the PWM pulse and definedby . The 6.41.2 MHz frequencies are

Fig. 20. Single ASD operating in laboratory at kHz, Hzwith 0.05 F added line to ground on drive outputs. Top trace: -to-groundvoltage (0level at offset 4, 250 V/div, 200 s/div); bottom trace: Filter output(0 level at offset 1, 50 V/div, 200 s/div).

TABLE IIGFI Vrms OUTPUT IN FIELD VERSUS LABORATORY WITH NO SOLUTION

APPLIED

also greatly attenuated by the filter at ( ) 40 dB to ( ) 50 dBfrom Fig. 17 and verified Section VII-D.

A third ASD frequency component developed across the

grounding resistor is that due to a 100400-kHz oscillationfollowing the IGBT rise time of Fig. 8. These components arealso greatly attenuated by the filter at ( ) 60 dB to ( ) 70 dBfrom Fig. 17 and verified in Section VII-D.

Thus, the impedance buffered low pass filter is effective inremoving ASD frequency components that might corrupt andfalsely trigger a GFI alarm. It also protects the GFI meter fromtransient voltages that might occur in the system.

C. Input Impedance of the GFI Filter

The filter 0100-Hz input impedance value is critical whenaccurate ground fault detection is required. Fig. 18 is a plot ofmeasured input impedance and phase angle with the filter output


10/15


(a) (b)

(c) (d)

Fig. 21. Waveforms at Field Refinery Site #1a with No Filter Solution and a High-Frequency Bypass Capacitor Solution across the HRG for an ASD operatingat kHz and Hz. (a) -to-ground voltage with No Filter Solution. (b) -to-ground voltage using 13- F Bypass Filter Cap Solution acrossHRG. (c) Spectrum of No Filter Solution (a) showing 52 Vrms at a fundamental frequency content of 2 kHz and 2-kHz harmonics. (d) Spectrum of Bypass FilterCap Solution (b) showing 5 Vrms at a fundamental frequency content of 2 kHz and 2-kHz harmonics.

terminals open circuited and short circuited. The open-circuitcase simulates operation with the GFI meters 620-k inputimpedance. The filter 60-Hz input impedance with its outputopen circuited is 4296 . The 60-Hz ground-fault currentcomponent, when it exists, will divide between the groundingresistor and filter input impedance. The GFI trip level is onlydesensitized by 6% and is an improvement as compared to otherfilters discussed. The 20-kHz resonant point in Fig. 18 is where

the filter-input impedance is maximum. Beyond 20 kHz, inputimpedancedropsand input phase angle becomes negative, duetocapacitive effects in the windings of the magnetic components.

The filter input impedance does not change much with theoutput terminalsshort circuited, so that there are no large currentinrushes into the filter when there is an accidental short circuiton the GFI meter terminals.

D. Laboratory Test Results of Impedance-Buffered

Low-Pass Filter

A single 480-V 2-hp ASD was connected to an input HRGtransformer with and to a motor with 30 ft of unshielded cable. The GFI voltmeter reading was recorded for

ASD operation at 5, 30, and 60 Hz and with two carrierfrequencies settings of kHz and kHz. Tests wererepeated with an 8- F Bypass Capacitor Filter of Fig. 13 andImpedance-Buffered Low-Pass Filterof Fig. 15 connected to theGFI meter.

Table I results show that with no filter solution, the false GFIreadings increase with higher carrier frequency and tend to de-crease with increasing ASD output frequency. GFI readings aregenerally 20 V for this single drive application with a low-ca-pacitance cable of short length. Both filter circuits of Fig. 13and Fig. 15 functioned properly when connected, in that theyallowed the GFI meter to accurately read the desired 60-Hzzero-voltage value.

Fig. 7 shows the input waveform to Fig. 15 filter for the cor-responding test conditions described and the ASD operating at

kHz, and Hz. Filter output voltage recordedwas 0 V. Fig. 19(a) shows the Fig. 15 filter input and filteroutput voltage waveforms for the ASD now operating atkHz, and Hz. Fig. 19 shows that the grounding resistorvoltage peaks are lower at the higher fundamental output fre-quencies. Filter output voltage at the desired zero-voltage value

is still achieved under this operating condition. Fig. 19(b) is atime-scale expansion of Fig. 19(a) showing the step-like modu-lation voltage across the neutral grounding resistor.

The next test condition attempted to simulate the appreciablecable line-to-ground capacitance of a long shielded or armorcable or a motor with a large line-to-ground capacitance. Thiswas accomplished by adding 0.05- F CM capacitors on eachASD output between line and ground. The results of Table Ishow large erroneous readings when the GFI meter is used alonewith no filter solution. Table I shows that ASD operation at lowfundamental output frequency is the worst case, as well as ASDoperation at higher carrier frequency.

The 8- F Bypass Capacitor Filter circuit reduces the worst

case 220-V GFI reading down to 20 V in Table I, but did notreduce it to zero. This suggests that this circuit is probably bestsuited for single drive operation with short cables and is subjectto system analysis for every drive application condition.

The Impedance-Buffered Low-Pass Filter circuit reduces theworst case 220-V GFI reading down to the desired 0-V valuefor any ASD operating condition and provides both a technicaland cost-effective filter solution. Fig. 20 shows input and outputfilter waveforms with the CM output capacitors added. Thehigher ASD output capacitance changes the -to-groundvoltage waveform as compared to Fig. 19(b). It is seen that the

-to-groundvoltage in Fig. 20 does not fully decay to zero be-fore the next switching instant. Peak-to-peak voltage excursions


11/15


(a) (b)

(c) (d)

(e)

Fig. 22. Waveforms at Field Refinery Site #2 with no solution, a GFI filter solution, and a solidly grounded drive isolation transformer solution applied. (a)Line-to-line output voltage of 200-hp ASD (500 V/div, 2 ms/div). Zero-sequence current at ASD output (10 A/div, 2 ms/div). (b) -to-ground voltage with

No Solution showing 150-Vpk spikes and step-like modulation (50 V/div, 200 s/div). (c) GFI voltage with Impedance-Buffered Low-Pass Filter Solution (10V/div, 10 ms/div). (d) -to-ground current in the ASD isolation transformer when a Drive Isolation Transformer Solution is added to the plant (10 A/div, 50

s/div). (e) GFI voltage across the HRG unit of the main plant transformer when a Drive Isolation Transformer Solution is applied (50 V/div, 200 s/div).

onthe transformer neutral -to-groundvoltage can swingfrom( ) 750 Vpk to ( ) 750 Vpk depending on system conditions.

The GFI filter of Fig. 15 attenuates these repetitive high-fre-quency high-magnitude spike voltages and correctly outputs azero-voltage value to the GFI meter. The filter output waveformin Fig. 20, with the added CM capacitors, show a small 5-V peakspike that decayed to zero in 10 s. The 8- s filter time constantintheGFImetercaneffectivelyhandlethis5-Vspikeandallowedthe GFI meter to record the desired zero-voltagevalue.

VIII. FIELD TEST OF GFI FILTER SOLUTIONS

The GFI voltmeter results taken before and after filter solu-tions were installedare presented. Results were taken at separaterefinery sites during running of the standard process line.

Field Site #1a): This is a 480-V HRG unit of 300 installedon a 750-kVA transformer which supplied a single 200-hp ASD

and other fixed-speed loads. The ASD was connected to themotor with 600 feet of 600 V, nonshielded continuous weldedarmor cable. ASD operation was at kHz and5060 Hz.

Field Site #1b): This is a 480-V HRG unit of 304 installedon a 1-MVA transformer with a total bus loading of 400 A. Two25-hp ASDs with output line reactors were installed and con-nected to a motor with 700 ft of continuous welded armor cable.ASD operation was at kHz and Hz.

Field Site #2): This is a 480-V HRG unit of 250 installedon a 1-MVA main transformer which supplied a single 200-hpASD and other plant fixed-speed motor loads. The ASD wasconnected to a motor with 600 ft of three single conductors plus


12/15


a ground wire which are unshielded and installed in rigid steelconduit. ASD operation was at kHz and 4950 Hz.

Field Site #3): This is a 480-V HRG unit of 130 installedon a 1.5-MVA transformer with a total bus loading of 40% con-tribution from 15 installed ASDs and 60% from other plantfixed-speed motor loads. The 15 ASDs each had 3% input linereactors and ranged in horsepower from 7.5 to 150 hp. Contin-

uous welded armor cable was installed on the drive input andoutput with lengths ranging from 100 to 700 ft. ASD operationwas at kHz and 3060 Hz.

A. Test Results With No Filter Solution

Field Site #4) had a single 4-kHz IGBT ASD, connected toa motor with armor cable, and operating on a transformer withan HRG unit using No Filter Solution at the GFI meter. Table IIfield results verify the previous laboratory data taken from Sec-tion VII-D and show erroneous GFI voltmeter readings decreasewhen ASD output frequency is increased. Field Site #1) dataalso show a similar trend for and frequency .

Field Site #1a) -to-ground voltage across the groundingresistor with No Filter Solution applied is shown in Fig. 21(a).The waveform is similar in appearance to the drive outputvoltage waveform of Fig. 6. The ring up voltage on every driveoutput switching step is due to reflected wave interaction ofthe long output cable and motor line-to-ground capacitance.The HRG waveform has a peak voltage of 150 Vpk and a cyclethat repeats at the expected 2-kHz (500 s) carrier frequency.Fig. 21(c) shows the voltage harmonic spectrum taken acrossthe HRG/GFI unit that corresponds to the time window ofFig. 21(a). The 2-kHz 150-Vpk waveform has a fundamentalcomponent of 52 Vrms. This voltage gives false GFI readings.Due to the step-like modulated waveform, the appearance of

(8 kHz) (12 kHz) harmonics in Fig. 21(c) are also seenas previously discussed in Section II. The long armor cableadds significant line-to-ground capacitance, so that the GFI52-Vrms readout is close to the Table I 60-Hz laborarory dataentry labeled GFI meter alone with CM output capacitors.

Field Site #2) -to-groundvoltage across the grounding re-sistor with No Filter Solution applied is shown in Fig. 22(b).The waveform is also similar in appearance to the driveoutput voltage waveform of Fig. 6. The HRG waveform alsohas a peak voltage of 150 Vpk and a cycle that repeats at the ex-pected 2-kHz (500 s) carrier frequency. In this case, the largeline-to-ground capacitance of the 200-hp motor is predominantover the smaller cable-to-conduit capacitance in determining the

neutral voltage waveform.Field Site #3) -to-groundvoltage across the grounding re-

sistor with No Filter Solution applied is shown in Fig. 12. Thewaveform is notat allsimilarin appearance to previous Field Sitecases.The HRGwaveform, which hadvoltagepeaksto 150Vpk,now has an occasional peak voltage of only 75 Vpk with 25 Vpkbeingmorecommon.Thisisduetotheinstantaneous( ) and ( )peaks of each ASD waveform adding and subtracting in the

neutral circuit. The instantaneous sum ofall 15 ASD cur-rents tends to reduce the high peak values and create more of awhite noise average waveform at a lower rms value. This is in-teresting effect, since many of the 15 drives investigated hadcurrentsof 25 Apk toground onthe ASD output.

(a)

(b)

Fig. 23. Waveform at Field Refinery Site #1b with ASD Output Line ReactorSolution and Impedance-Buffered Low-Pass Filter Solution for a single ASDoperating at kHz and Hz. (a) -to-ground voltage withASD Output Line Reactor Solution installed. (b) Voltage across GFI (output of

Impedance-Buffered Low-Pass Filter) with ASD Output Line Reactor Solutionand Impedance-Buffered Low-Pass Filter Solution installed.

B. Solution at GFI Using a High-Frequency Bypass Capacitor

Field Site #1a) -to-ground voltage across the groundingresistor with a parallel 13- FHigh-FrequencyBypass CapacitorSolution applied is shown in Fig. 21(b). The HRG peak voltageis now reduced by a factor of ten from 150 Vpk withNo Solutionto less than 10 Vpk, with a waveform cycle that repeats at the

expected 2-kHz carrier frequency. The capacitor addition causesincreased high-frequency noise ringing on every step edgedue to interaction with transformer leakage inductance, cableinductance, or possibly some amount of motor stator leakageinductance. Fig. 21(d) shows the voltage harmonic spectrumtaken across the HRG/GFI unit that corresponds to Fig. 21(b).The 2-kHz 10-Vpk waveform has a fundamental componentof a 5 Vrms, which is an acceptable GFI reading The step-likewaveform with the increased high-frequency ringing leads to the

to harmonics being higherin magnitude.

C. Solution at the ASD Using Output Line Reactor

Field Site #1b) -to-ground voltage across the groundingresistor is shown in Fig. 23(a) when output line reactors wereinstalled and one of the 25-hp ASDs shut off. The waveformis not similar in appearance to the previous No Filter Solutionvoltage waveforms of Fig. 21(a). The line reactor interacts withline-to-ground capacitance to smooth out the waveform. TheHRG waveform has a peak voltage of 20 Vpk. In this case, it isseen that the line reactor can reduce, but not eliminate the ASDinduced -to-ground voltage across the grounding resistor.

D. Solution at GFI Using Impedance-Buffered Low-Pass Filter

Field Site #1b) conditions of Fig. 23(a) were modified to in-

sert an Impedance Buffered Low Pass Filter Solution between


13/15


14/15


has a 180-Hzcomponent voltage that is further amplitude modu-

lated by a 10-Vpk15-Vpk ( 10 Vrms) component of unknown

25-Hz30-Hz origin. The corresponding GFI meter reading was

an acceptable 8 V. Due to the high value of neutral grounding

resistance, a relatively small amount (50 ma) of 180- and 25-Hz

harmonic current develops a significant voltage (8 Vrms) across

the resistor and indicating voltmeter with GFI voltage sensing

schemes.

E. Solution at the ASD Using Drive Isolation Transformers

Field Site #2) was modified per Fig. 10 to insert a solidly

grounded drive isolation transformer between the 200-hp ASD

and main plant transformer where the HRG unit is located.

Fig. 22(a) shows the line-to-line output voltage of the 200-hp

ASD and the resulting zero-sequence current at every ASD

output switching instant after the drive isolation transformer

was installed. Peak currents up to 30 Apk were recorded.

Fig. 22(d) shows the transient zero-sequence current is

totally contained within the -to-ground neutral circuit of

the ASD isolation transformer. Fig. 22(e) shows the resulting

GFI voltage across the HRG unit of the main plant transformer

is completely clean. The advantage is it effectively solves

the problem and keeps ASD zero-sequence high-frequency

current from the rest of the plant. Also, several drives can share

one isolation transformer. A disadvantage is cost. Another

disadvantage is that for the solidly grounded neutral the benefit

of a high-resistance grounded system is eliminated on the

transformer output, that is, you have to trip immediately for a

ground fault.

IX. GFI APPLICATION GUIDELINES WHEN USING ASDS

The GFI is designed to provide safe and reliable protection

by sensing and detecting the 60-Hz low-level ground-fault or

leakage current present in a system. This paper has shown a

methodology whereby the erroneous GFI readings and nuisance

false trips/alarms caused by PWM ASDs can be eliminated.

This was done either by: 1) reducing the high-frequency ASD

zero-sequence current that occurs at the drive switching fre-

quency rate; 2) rerouting the high-frequency current away from

the GFI; 3) desensitizing the GFI to the effect of ASD zero-se-

quence current with local filters; or 4) changing the GFI to cur-

rent sensing scheme, which on some models, may have some

inherent high-frequency filter mechanism. Table III provides asummary guideline of the pros and cons of using no filter and

applications that require a GFI solution, be it at the drive or at

the GFI meter location.

X. CONCLUSION

The method and conditions by which an ASD produces a

high-frequency zero-sequence current which interferes with

HRG-GFI systems has been described.

Several corrective measures including different filter arrange-

ments and an isolation transformer have been investigated. The

results of these investigations are supported by laboratory and

field measurements.

It is always desirable to know when to apply corrective mea-

sures to avoid a GFI problem up front before the equipment is

installed, so that the system can be properly designed from the

beginning rather than finding the problem in the ASD startup

phase. There is a wide variety of voltage-sensing GFI vendors,

current-sensing GFI vendors, and ASD vendors. It is difficultto predict when a GFI problem will occur since vendor vari-

ation must also be considered along with system application

variations in line-to-ground capacitance with horsepower, cable

length, cable type, HRG value, pulse rise-time variation, carrier

frequency, output frequency, line reactor usage, and use of single

versus multidrive systems. However, from observations made, it

appears that lower hp drives ( 10 hp) with short output cables

may not create enough ASD-induced zero-sequence voltage to

cause GFI problems.

Having determined when we need to apply a corrective mea-

sure, we can conclude that a filter can be applied at the GFI

and the Impedance-Buffered Low-Pass Filter is the most effec-tive filter for voltage-sensing-type GFIs. See Table III for limi-

tations. We can also conclude that some types of current-sensing

GFIs may have CTs that act as inherent filter mechanisms to the

high-frequency components of the zero-sequence current.

However, if it is desired to avoid ASD-generated zero-se-

quence currents and voltages from entering other portions of

the plant electrical system, an isolation transformer connected

delta-wye grounded can be used and is recommended. See

Table III for limitations.

ACKNOWLEDGMENT

The authors would like to thank J. Sands, J. Ulloa, and P.

OBrien of Chevron, J. McQuacker and F. Shewchuk of Candor

Engineering, and J. Mistry, D. Dahl, and H. Jelinek of RA for

their valuable assistance in arranging for and conducting the

field measurements, which provided essential data for the de-

velopment of this paper. Additionally, the authors would like to

thank L. Berg for his insightful comments and J. Propst for his

support of the paper.

REFERENCES

[1] J. Nelsonand P. Sen, Highresistance grounding oflow voltage systems:

A standard for the petroleum and chemical industry, presented at theIEEE PCIC, Philadelphia, PA, Sept. 26, 1996.

[2] D. Beeman, Industrial Power Systems Handbook. New York: Mc-Graw-Hill, 1955, pp. 286289.

[3] J. Schaefer, Rectifier Circuits: Theory and Design. New York: Wiley,

1965, pp. 2730.[4] G. Skibinski, D. Dahl, K. Pierce, R. Freed, and D. Gilbert, Installation

considerations for multi motor IGBT AC drives & filters used in metalsindustry applications, presented at the IEEE-IAS Annu. Meeting, St.Louis, MO, Oct. 1998.

[5] E. Bulington and G. Skibinski, Cable alternatives for PWM AC driveapplications, presented at the IEEE PCIC, San Diego, CA, Sept. 26,1999.

[6] D. Anderson, R. Kerkman, L. Saunders, D. Schlegel, and G. Skibinski,Modern drives application issues and solutions tutorial, presented atthe IEEE PCIC, Philadelphia, PA, Sept. 26, 1996.


15/15


Gary L. Skibinski received the B.S.E.E. andM.S.E.E. degrees from the University of Wisconsin,Milwaukee, and the Ph.D.degree fromthe Universityof Wisconsin, Madison, in 1976, 1980, and 1992,respectively.

From 1976 to 1980, he was an Electrical Engineerworking on naval nuclear power at Eaton Corpora-tion. From 1981 to 1985, his work as a Senior ProjectEngineer at Allen-Bradley Company concerned

servo controllers. During the Ph.D. program, he wasa Consultant for UPS and switch-mode power supplyproducts at R.T.E. Corporation. He is currently a Principal Research Engineerwith Rockwell AutomationAllen-Bradley Company, Inc., Mequon, WI. Hiscurrent interests include power semiconductors, power electronic applications,and high-frequency high-power converter circuits for ac drives.

Barry M. Wood (M73SM87) received theB.S.E.E. degree from Virginia Polytechnic Instituteand State University, Blacksburg, and the M.S.E.E.degree from the University of Pittsburgh, Pittsburgh,PA, in 1972 and 1978, respectively.

From 1972 through 1977, he was with West-inghouse Electric Corporation, Pittsburgh, PA,as a Power Systems Engineer for the IndustryServices Division. In 1978, he joined McGrawEdison Company, Canonsburg, PA, as a SeniorPower Systems Engineer and, in 1981, he joined

Electro-Test, Inc., San Ramon, CA, where he held positions of Senior ElectricalEngineer and Supervisory Electrical Engineer. Since 1987, he has been withChevron Corporation, where he is currently a Staff Electrical Engineer withChevron Research and Technology Company, Richmond, CA. His primaryresponsibilities include consulting in the areas of electrical power systems,adjustable-speed drives, motors, and generators.

Mr. Wood is a Registered Electrical Engineer in the States of California andPennsylvania.

John J. Nichols (S85M88) received the B.S.E.E.degree from the University of Missouri, Columbia,and the M.S.E.E. degree from the University of Mis-souri, Rolla, in 1988 and 1994, respectively.

He is currently with the Wood River Refinery,Tosco Refining Company, Roxana, IL, as theElectrical SupervisorTechnical.

Mr. Nichols is a member of the IEEE Industry Ap-plications Society and a Licensed Professional Engi-

neer in the State of Illinois.

Louis A. Barrios (S84M86SM00) received theB.S.E.E. and M.S.E.E. degrees from Louisiana TechUniversity, Ruston, in 1987 and 1989, respectively.

From 1989 to 1998, he was with Shell OilCompany, where he was an Electrical Engineerin petrochemical plants in Louisiana and Illinois.In 1998, he joined Equilon Enterprises, a jointventure between affiliates of Shell Oil Companyand Texaco Inc. Since 1999, he has been withEquilon Technology, Houston, TX, where he iscurrently a Staff Electrical Engineer providing

electrical consulting services to the petrochemical industry. He is a memberof the American Petroleum Institute Subcommittee on Electrical Equipment,an alternate member on Code Making Panel #1 of the National ElectricalCode, and an alternate member on NFPA 70EStandard for Electrical SafetyRequirements for Employee Workplaces.

Mr. Barrios is a member of theExecutive Subcommittee of thePetroleum andChemical Industry Committee of the IEEE Industry Applications Society and aRegistered Electrical Engineer in the State of Louisiana.

Puesta a Tierra y Drives

Documents

Transcript of Puesta a Tierra y Drives