Short T.A. Electric Power Distribution Handbook

Подождите немного. Документ загружается.

Voltage Sags and Momentary Interruptions 507

10.7.3 Voltage Class

Figure 10.18 shows that 5-kV systems have much lower numbers of voltage

sags and interruptions. Lower voltage systems have less feeder exposure,

and higher line impedance relative to the station transformer. Fault rates are

often lower on 5-kV systems. Somewhat surprisingly, the 25 and 35-kV

systems were not worse than the 15-kV systems.

10.7.4 Comparison and Ranking of Factors

We analyzed data available on the DPQ site characteristics to determine what

parameters most affected power quality events. Figure 10.19 shows the vari-

ations of SARFI

ITIC

with site characteristics.

The three most signiﬁcant predictors of excursions below the lower ITI

curve are:

1. Circuit exposure — The total exposure on the circuit including three-

phase and single-phase portions is a good predictor of voltage sags.

Any fault on the circuit sags the voltage.

2. Lightning — Lightning causes many faults on distribution systems,

and lightning strongly correlates with voltage sags (based on the 10-

year average, 1988–98, from the U.S. National Lightning Detection

Network). In addition, lightning predicts weather patterns — areas

with high lightning tend to have more storms and more wind and

tree-related faults.

3. Transformer impedance and number of feeders — The n

◊ kV

/MVA

xfrmr

term in Figure 10.19 contains the number of feeders off of the trans-

former bus along with an estimate of the transformer impedance.

The transformer impedance is Z

/MVA; but since the per-unit

FIGURE 10.18

Comparison of feeder sites by voltage class in the DPQ data.

5 kV

15 kV

25 & 35 kV

0 20 40 60

100

Sarfi

ITIC

Percent of locations

exceeding the x-axis value

1791_book.fm Page 507 Monday, August 4, 2003 3:20 PM

508 Electric Power Distribution Handbook

impedance of station transformers is roughly constant (7 to 10%),

we use kV

/MVA.

This last term requires a bit more explanation. The number of bus sags is

directly proportional to n

, the number of feeders off the bus and to Z

, the

source impedance (a lower station transformer impedance — a bigger trans-

former or lower percent impedance — improves voltage sags at the station

bus). We approximate these two terms as n

◊ kV

/MVA

xfrmr

FIGURE 10.19

Variations in the number of excursions below the lower ITI curve (which are mainly voltage

sags) vs. various site parameters. The correlation coefﬁcients (r) are given in the upper-left

corner of each plot.

Total feeder exposure, miles

0.33

SARF-IITIC

020406080

020406080 120

Total 3-phase exposure, miles

0.29

01020304050

Flash density, flashes km

year

0.24

024681012

base

MVA

xfmr

0.41

SARF-IITIC

020406080 120

Nominal line-to-line voltage, kV

0.2

51015202530

Station transformer base MVA

0.08

020406080100

Percentage tree coverage

0.08

SARF-IITIC

10 20 30 40 50 60 70

020406080 120

Number of feeders

0.05

1234567

1791_book.fm Page 508 Monday, August 4, 2003 3:20 PM

Voltage Sags and Momentary Interruptions 509

Other variables have much less impact on the number of voltage sags than

the three main parameters given.

10.8 Prediction of Quality Indicators Based on Site

Characteristics

We derive a formula for predicting the number of events for a quality indi-

cator based on a few of the characteristics of the site. If no measurement or

historical data is available, this is useful in estimating the utility-side quality.

Regression techniques are commonly used to ﬁnd a model prediction

formula. A generalized linear model is a least-squares ﬁt to an equation of

the following form:

The x’s are site characteristics (such as base voltage or lightning ﬂash

density), and the a’s are coefﬁcients ﬁtted to the model. The generalized

linear model is somewhat different from a standard linear model; we used

a generalization where the distribution of the error e is assumed to be a

gamma distribution rather than a normal distribution in a strictly linear

model. A gamma distribution skews to the right, like the log-normal

distribution.

A model for estimating SARFI

ITIC

where

ITIC

= predicted annual number of events which fall under the lower

ITI curve

l = total exposure (including three-phase and single-phase por-

tions) on the circuit, mi (multiply kilometers by 1.609)

= lightning ground ﬂash density, ﬂashes/km

/year

kV = base line-to-line voltage, kV

= total number of feeders off of the substation bus

MVA

xfmr

= station transformer base rating (open-air rating), MVA

If any of the circuit characteristics are unknown, we could use the follow-

ing medians from the DPQ data:

yax ax ax

=++++

11 2 2

L e

NlN

nkV

MVA

ITIC g

xfmr

=+ + +

◊

474 076 247 0192

.. . .

. if moderate to heavy tree coverage

1791_book.fm Page 509 Monday, August 4, 2003 3:20 PM

510 Electric Power Distribution Handbook

where

l = 14.5 mi (23.4 km)

= 2.57 ﬂashes/km

/year

= 25

All three variable terms in the linear regression are signiﬁcant to at least

99% (there is less than a 1% chance that the terms of the model do not

inﬂuence the prediction). The tree coverage term is less certain; there is a 9%

chance that the term is not signiﬁcant. We based the tree coverage term on

the University of Maryland’s Global Land Cover Facility data from the

Advanced Very High Resolution Radiometer (AVHRR). Half of the DPQ sites

had more than 19% of the land area covered by trees, which we deﬁned as

“moderate to heavy tree cover.”

How good is the model? It is decent given all the factors that affect sags

and momentary interruptions and inherent variability. Given the variability

of power quality events, it is surprising that the model is this good: 34% of

the values are within 25% of the prediction, and 60% of the values are within

50% of the prediction. See Figure 10.20 for the prediction scatter.

For an example 12.47-kV case with three feeders, a 25-MVA transformer,

a ﬂash density of 4 ﬂashes/km

/year, moderate tree coverage, and a total

exposure of 32 km, the model predicts 29.8 events per year. For this case, the

data shows a prediction interval with a 50% conﬁdence level of between 15.6

and 34.2 events per year (the 90% conﬁdence prediction interval is between

0 and 68.3). The data is so dispersed that the model is not good enough to

use for precision estimates (such as in a contract for premium power).

FIGURE 10.20

Actual values vs. predicted values for the model predicting the annual average number of

events below the lower ITI curve.

nkV

MVA

xfmr

◊

0102030405060

020406080

Predicted value

Actual value

0102030405060

020406080

1791_book.fm Page 510 Monday, August 4, 2003 3:20 PM

Voltage Sags and Momentary Interruptions 511

The site characteristics most affecting sags but not included in this model

(because no information was available) are (1) subtransmission exposure and

characteristics and (2) percentage of the circuit that was underground.

A reasonable model for predicting momentary interruptions is

where

= the predicted annual number of events with voltage less than 10% of

nominal for more than 0.4 sec

= the three-phase circuit exposure, mi

The parameters differ somewhat from SARFI

ITIC

predictors. Two of the

strongest indicators of momentary interruptions are load density and three-

phase circuit exposure. Other signiﬁcant parameters are the lightning activity

and a term with voltage, number of feeders, and transformer MVA. The

model is not as good as the ITI model, but all parameters have more than a

95% probability of affecting the result. The site characteristic most affecting

momentaries that is not included in the model for lack of information is

whether fuse saving is used.

10.9 Equipment Sensitivities

10.9.1 Computers and Electronic Power Supplies

Computers and other equipment with electronic power supplies are the most

widely found equipment that is sensitive to power quality disturbances. The

power supply is typically a switched mode power supply as shown in Figure

10.21. Computers have a wide range of sensitivities. The ride-through capa-

bility for interruptions of several computers is summarized in Figure 10.22.

Many of the computers had ride through of more than 0.1 sec (0.28 sec was

the best of this set of studies), and some could not even ride through a 0.01-

sec interruption.

The ride-through capability of computers is close to rectangular. Two

points describe the characteristic on a volt-time curve: the interruption ride

through time and the steady-state ride-through point. There is usually a

steep transition between the interruption ride-through point and the

steady-state ride-through point. Other characteristics of computer ride

through are:

NlN

nkV

MVA

xfmr

10 3

552

029

161

030 027 124=

++ +

◊

.. .

if Rural

if Suburban

if Urban

1791_book.fm Page 511 Monday, August 4, 2003 3:20 PM

512 Electric Power Distribution Handbook

• There is little difference between the performance when the com-

puter is processing or accessing disk and when the computer is idle.

• The point on the waveform when the disruption occurs does not

matter.

Figure 10.23 shows volt-time sensitivities of computers from several stud-

ies. The most sensitive units, those that violate the ITI curve, were made and

tested before the ITI curve was created. That is not much of an excuse as the

most sensitive computers also violated the CBEMA curve which was avail-

able to the manufacturers at that time (remember, there is no standard that

requires testing computers to meet the ITI/CBEMA curve).

FIGURE 10.21

Switch-mode power supply used in most computers.

FIGURE 10.22

Capability of computers to ride through an interruption (n = 27). (Data from [Bowes, 1990;

Chong, 2000; Courtois, 2001; Courtois and Deslauriers, 1997; EPRI PEAC Brief No. 7, 1992].)

Regulated

dc bus

regulator

logic

circuits

Unregulated

dc bus

Switch-mode power supply

ITI threshold

0.0 0.1 0.2

100

Ride-through duration for an interruption, seconds

Percent of tested samples

exceeding the x-axis value

1791_book.fm Page 512 Monday, August 4, 2003 3:20 PM

Voltage Sags and Momentary Interruptions 513

An important factor regarding the ride-through capability of computers is

that it varies signiﬁcantly depending on the voltage just before the interrup-

tion. The energy storage in a switch-mode power supply is from the front-

end rectiﬁer capacitors. The energy stored in a capacitor is 1/2CV

. Power

supplies typically have two 470-mF capacitors in series, and the voltage across

the two capacitors in series is . We can estimate the

ride-through capability of a computer as:

where

t = ride-through duration for an interruption, sec

P = load on the computer, W

C = capacitance on one half of the bridge rectiﬁer, mF (470 is common)

= peak of the ac voltage, V (339.4 V for 120 V nominal)

= voltage on the unregulated dc bus where the computer will drop out

(use half of V

if unknown or 0 for the maximum ride through)

Since the energy is a function of V

, a voltage of 90% of nominal means

the capacitor stores only 81% of the energy that it would at nominal voltage.

FIGURE 10.23

Volt-time characteristics of several computers tested in different studies. (Data from [Bowes,

1990; Chong, 2000; Courtois, 2001; Courtois and Deslauriers, 1997; EPRI PEAC Brief No. 7, 1992;

of Extremely Reliable Power Delivery Systems. Reprinted with permission.)

ITI curve

0.0 0.1 0.2 0.3

100

Duration, seconds

Magnitude, percent

1990: Northeast Utilities

1992: EPRI PEAC

1992: Japanese

1997: Hydro Quebec

2000: U. of Queensland

=◊=22120 339 4. V

CV V

()

410

1791_book.fm Page 513 Monday, August 4, 2003 3:20 PM

514 Electric Power Distribution Handbook

Even worse, the computer drops out before all of the energy in the capacitor

is used. Figure 10.24 shows the relative ride through as a function of the

voltage prior to the interruption, assuming the computer drops out when

the unregulated dc bus voltage reaches half of nominal.

The pre-disturbance voltage affects ride through for any device that has

capacitance for energy storage, including most computer power supplies,

programmable-logic controllers, digital clocks, and adjustable-speed drives.

So, either on the utility side or the customer’s side, raise voltages to inex-

pensively increase ride through of devices.

Even more ride-through capability is possible with computers. EPRI PEAC

has done tests of a computer power supply modiﬁed with extra ride-through

capability (EPRI PEAC Brief No. 12, 1993). The enhanced supply, developed

by the New England Electric Company, had an extra 4500 mF of capacitors

installed in parallel with the existing capacitors, which increased the ride

through from 0.175 sec to 1.8 sec. In the near future, ultracapacitors may

supply an even more economical ride-through enhancement.

Intelligent power management might also increase ride through. Laptops

and most desktops have sophisticated ways of managing power to conserve

energy. We could apply similar techniques to short-duration power inter-

ruptions. The processor, disk, and other power-hungry equipment could be

“quick suspended” during a power interruption to increase the normal 0.05

to 0.2-sec ride-through capability. Just suspending the processor (30–50 W

typically in fast, hot chips) would extend the ride through considerably. This

enhancement requires very little extra hardware; a sensor to measure the

incoming ac power or the unregulated dc bus voltage would be needed —

no microprocessor-level changes are required.

FIGURE 10.24

Change in the ride-through capability of computers vs. the voltage prior to the interruption.

0.90 0.95 1.00 1.05 1.10

0.8

1.0

1.2

Per-unit voltage before the interruption

Interruption ride-through relative

to the ride through at nominal voltage

1791_book.fm Page 514 Monday, August 4, 2003 3:20 PM

Voltage Sags and Momentary Interruptions 515

Industrial dc power supplies share the same characteristics as the com-

puter power supply. Heavily loaded power supplies are more susceptible to

voltage sags and interruptions. Use a supply rated at twice the load on the

supply to increase ride through.

A power supply with a universal input operates over a wide range of

voltages (85 to 264 V typically), but the ride-through capability changes

dramatically with operating voltage. Operation as close to the upper end as

possible improves ride through. For this reason, prefer a line-to-line connec-

tion (208 V) over a line-to-ground connection (120 V). The low-voltage limit

of 85 V is 71% of nominal at 120 V, but we obtain much better ride through

when applied at 208 V (now the lower limit is 41% of nominal). The difference

in 1/2CV

is dramatic in the two cases. By the same token, if the power supply

has alternate settings, use the setting that positions the actual voltage near

the high end of the range. Consider a power supply with a 95 to 250-V range

designed for Japanese and European loads and a 110 to 270-V range designed

for America. The range with an upper limit of 250 V for a 208-V line-to-line

connection results in the maximum ride through (McEachern, 2001).

Another option for some industrial supplies and large computer systems

is a three-phase power supply instead of a single-phase supply. A three-

phase supply is less sensitive to voltage sags. Single-phase sags only slightly

depress the dc bus voltage of a three-phase rectiﬁer because the remaining

two phases can fully supply the load. Even a two-phase sag is signiﬁcantly

less severe than a three-phase sag.

Linear power supplies have much less ride-through capability than switch-

mode power supplies (switch-mode supplies may have 100 times the capac-

itance). Fortunately, most power supplies are switch-mode supplies (prima-

rily because they are lighter, more efﬁcient, and cost less).

10.9.2 Industrial Processes and Equipment

A variety of industrial equipment is sensitive to voltage sags. Some of the

main sensitive equipment used in industrial facilities are

•Programmable logic controllers (PLCs)

• Adjustable-speed drives (ASDs), also called variable-speed drives

(VFDs)

• Contactors

• Relays

• Control equipment

Depending on the process and load, any number of devices can be the

weak link. Table 10.10 shows the breakdown of weak links for semiconductor

tools serving the semiconductor manufacturing industry.

1791_book.fm Page 515 Monday, August 4, 2003 3:20 PM

516 Electric Power Distribution Handbook

10.9.2.1 Relays and Contactors

Contactors are electromechanical switches used for a variety of power and

control applications. A contactor uses a solenoid to engage when an appro-

priate voltage is applied. More voltage is required to close the contactor than

is required to keep it closed.

Relays and contactors can drop out very quickly. Figure 10.25 shows the

ride-through duration for an interruption for several relays and contactors,

and Figure 10.26 shows the dropout levels for voltage sags. The devices are

somewhat dependent on the point on the wave where the voltage sag starts.

Ride through is longest for sags starting at the voltage zero crossing; but

unfortunately, faults tend to occur when the voltage is near its peak. The

fast dropout of contactors limits some of the utility-side solution options —

faster relaying, smaller fuses, or 1.5-cycle transfer switches may provide

good improvement to computers but offer little help for many relays and

contactors. Because they trip very quickly, voltage mainly dominates, not

the duration.

The volt-time capability of relays and contactors approximates a rectan-

gular shape. Contactors can have the unusual property that the ride-through

capability improves at lower voltages. An example volt-time ride-through

characteristic is shown in Figure 10.27. The reason for this property relates

to the fact that current, and not voltage, holds a contactor in. A contactor

contains shading rings, which are analogous to damper windings in a rotat-

ing machine. A shading ring is a shorted winding around the magnetic core.

In response to a voltage transient, the shading ring produces a back emf that

opposes the transient. A larger transient (deeper sag) creates more current

that holds the contactor in (Collins Jr. and Bridgwood, 1997).

A larger relay generally has more ride through; a contactor usually has

more ride through than a relay. Some of the most sensitive relays are small

industrial relays with clear plastic cases referred to as ice-cube relays.

Several options are available to help hold in contactors and relays (St.

Pierre, 1999):

TABLE 10.10

Breakdown of Semiconductor-Tool Voltage Sag Sensitivities (n = 33)

Weak Link

Overall

Percentage

Emergency off (EMO) circuit: pilot relay (33%), main contactor (14%) 47%

dc power supplies: PC (7%), controller (7%), I/O (5%) 19%

3-phase power supplies: magnetron (5%), rf (5%), ion (2%) 12%

Vacuum pumps 12%

Turbo pumps 7%

ac adjustable-speed drives 2%

Source: Stephens, M., Johnson, D., Soward, J., and Ammenheuser, J., Guide for the Design

of Semiconductor Equipment to Meet Voltage Sag Immunity Standards, International SE-

MATECH, 1999. Technology Transfer #99063760B-TR, available at http://www.semat-

ech.org/public/docubase/document/3760btr.pdf.

1791_book.fm Page 516 Monday, August 4, 2003 3:20 PM