Short T.A. Electric Power Distribution Handbook

Подождите немного. Документ загружается.

588

Electric Power Distribution Handbook

The voltage impulse easily breaks down most distribution-class insulation

unless it is protected with a surge arrester. Almost all direct lightning strokes

cause ﬂashovers. In addition, the lightning current may start a pole ﬁre or

burn through conductors. Also, nearby lightning strokes that do not hit the

line may couple damaging voltages to the line. These induced voltages may

fail equipment or cause ﬂashovers.

Figure 12.1(a) shows an example of a lightning ﬂash composed of several

individual strokes hitting a distribution circuit (the movement of the light-

ning channel and camera shaking makes it easier to distinguish the individ-

ual strokes). Two phases initially ﬂashed over, and the third eventually

ﬂashed over (either due to a subsequent stroke or the gasses from the initial

fault). On three-phase circuits, lightning normally causes two- or three-phase

faults. The lightning-caused fault in Figure 12.1(a) resulted in a downed-

wire high-impedance fault that was not cleared by the station protection (the

circuit was tripped manually). Taken moments after the lightning ﬂash,

Figure 12.1(b) shows the glow of an arcing downed conductor. In this case,

the 60-Hz fault arc, not the original lightning impulse, burned the wires

down. Two characteristics made this circuit prone to burndowns: covered

wire and relaying that allowed long-duration faults (the instantaneous relay

in the substation did not cover this portion of the line). The second bright

blob in the lower right of the lightning strike picture is another ﬂashover

further down the line on the same circuit. These supplementary ﬂashovers

— away from the fault point and on the same circuit — are common, espe-

cially in areas with high ground impedances (the ground potential rise

pushes the surge to other locations on the circuit). These pictures were taken

by a lightning-activated camera developed by the Niagara Mohawk Power

Corporation (now part of National Grid) (Barker and Burns, 1993).

Locations with higher lightning activity have higher fault rates on over-

head circuits. Figure 12.2 shows fault rates reported by different utilities

against estimated ground ﬂash density. Utilities in higher lightning areas

have more faults. Not all of these are due to lightning; many are due to wind

and other storm-related faults. Lightning is a good indicator of storm activity

at a location.

The linear curve ﬁt to the data in Figure 12.2 shows line fault rates per 100

mi per year varying with the ground ﬂash density,

, in ﬂashes/km

/year as

12.1 Characteristics

Driven by the need to better protect transmission and distribution lines and

equipment, the electrical industry performed much of the early research on

=+20 6 43.

1791_book.fm Page 588 Monday, August 4, 2003 3:20 PM

Lightning Protection

589

lightning and electrical breakdowns. Charles P. Steinmetz, Charles F. Wagner,

Walter W. Lewis, Karl B. McEachron, Charles L. Fortescue, Basil F. J. Schon-

land — the list of early contributors is long — increased our understanding

of the physics of lightning and its electrical characteristics.

Lightning is the electric breakdown of the air from high electric ﬁelds

generated when electric charge separates within a cloud. Lightning may ﬂash

within a cloud, from one cloud to another, or from the cloud to the ground.

Distribution lines are only affected by cloud-to-ground lightning. In the nor-

mal scenario, charge separates within a thundercloud. The upper portion

becomes positively charged and the lower portion becomes negatively

charged. The ground just underneath the cloud becomes positively charged

FIGURE 12.1

Example of a lightning ﬂash to a 13.2-kV distribution line and the downed wire that resulted.

(a) Lightning strike and flashovers

Supplementary flashovers

(b) Arcing downed conductor

1791_book.fm Page 589 Monday, August 4, 2003 3:20 PM

590

Electric Power Distribution Handbook

(being attracted to the negatively charged lower portion of the cloud). The

lightning breakdown begins in the lower portion of the cloud. The air breaks

down in steps called

stepped

leaders. Each step is about 150 ft (50 m) with

pauses of about 50

sec between steps. The stepped leader may fork and form

branches that each progress towards the ground. As the stepped leader

progresses closer to the ground (see Figure 12.3), more charge is lowered closer

to the ground. More positive charge collects on the earth in response — short

upward leaders extend to meet the downward negative stepped leader.

When the downward leader meets the upward leader, a

return stroke

occurs. The negative charge held in the stepped leader rushes into the

ground, brilliantly lighting the channel and creating a large pressure wave

(thunder). The return stroke propagates up the channel at roughly 20% of

the speed of light, releasing charge as it goes. The charge rushing into the

ground creates a current of tens of thousands of amps peaking in a few

microseconds. The current may extinguish in about 100

sec, or lower-level

continuing current in the range of hundreds of amperes may ﬂow for several

milliseconds (about 25% of the time, continuing currents ﬂow following the

return stroke).

Subsequent strokes

may follow the ﬁrst stroke. After the current extinguishes

and the channel becomes dark, another pocket of charge may work its way

down the same path. Fast-moving leaders called dart leaders break down

the recently deionized path of the ﬁrst stroke. Subsequent strokes typically

have lower magnitudes of current and charge transferred, but subsequent

stroke currents have higher rates of rise. Subsequent strokes have higher

FIGURE 12.2

Distribution line fault rate vs. ground ﬂash density. (Based on the data in Figure 7.2.)

0 2 4 6 8 10

100

200

300

Ground flash density, flashes/km

/year

Fault rate per 100 miles per year

1791_book.fm Page 590 Monday, August 4, 2003 3:20 PM

Lightning Protection

591

return-stroke velocities, often greater than 50% of the speed of light. The ﬁrst

stroke and subsequent strokes make up a lightning

ﬂash

While the downward negative ﬂash is the most common, other types of

cloud-to-ground lightning occur. About 5 to 10% of cloud-to-ground ﬂashes

are positive. Downward positive lightning lowers positive charge from the

cloud to the ground. Breakdown starts at a positive portion of the cloud

usually near the top of the cloud; a positive downward stepped leader moves

downward until it meets an upward negative leader close to the ground.

Some positive ﬂashes may have very large peak currents and charge. Positive

ﬂashes occur more often during winter storms, especially in some areas.

Positive ﬂashes usually only have one stroke. Cloud-to-ground lightning

may also start at the ground and rise upward, with an upward stepped

leader starting at the ground. These are common on tall objects like the

Empire State Building, but rare from distribution lines.

Normally, the lightning current injection is considered an ideal current

surge (it does not really matter what is struck; the electrical characteristics

of the current stay the same). Table 12.1 shows characteristics of a downward

negative current ﬂash. Many of the characteristics ﬁt a log-normal distribu-

tion, which is common for data bounded at zero. The log standard deviation,

, is shown for the characteristics that have a log-normal char-

acteristic. The 5th and 95th percentiles are shown based on the log-normal

ﬁt. The ﬁrst stroke peak current data does not ﬁt a log-normal distribution,

but Anderson and Eriksson found a good ﬁt using two log-normal param-

FIGURE 12.3

Cloud-to-ground lightning.

Charge rushes

down the channel

Downward

stepped

leader

Return stroke

moves up

the channel

b=sd(ln( ))x

1791_book.fm Page 591 Monday, August 4, 2003 3:20 PM

592

Electric Power Distribution Handbook

eters, one for low currents and one for high currents. Another common

approximation to Berger’s data for the probability of the peak magnitude of

a ﬁrst stroke is (EPRI, 1982):

Table 12.2 shows characteristics for positive strokes. This data is not as

well deﬁned since the data set is more limited.

Although most stroke and ﬂash characteristics are independent of each

other, there are some interdependencies. Cigre (1991) examined correlations

TABLE 12.1

Lightning Current Parameters for Downward Negative Flashes

Parameter

Percent of Cases More than Value

5% 50% (

) 95%

First Strokes

Peak current, kA

Model for I

20kA

Model for I>20kA

8 33.3

61.1

33.3

1.33

0.61

Time to peak,

sec

(virtual front time based on the time

from 30 to 90% of the peak = T

30–90%

/0.6)

1.5 3.83 10 0.553

Steepness, 30–90%, kA/

sec 2.6 7.2 20 0.921

Tail, time to half the peak,

sec 30 77.5 200 0.577

Charge, C 1.1 4.65 20 0.882

dt, (A)

sec

657546 1.373

Subsequent Strokes

Peak current, kA 5.2 12.3 29.2 0.530

Time to peak,

sec

(30–90% virtual front)

0.2 0.67 3.0

Steepness, 30–90%, kA/

sec 4.1 20.1 99 0.967

Tail,

sec 6.5 30.2 140 0.933

Charge, C 0.2 0.938 4 0.882

dt, (A)

sec

0.6 5.5 52 1.366

Flash

Charge, C 1.3 7.5 40 1.02

Flash duration, sec 0.03 0.2 1

Number of strokes 1 2–3 9

Interval between strokes, msec 6 35 202 1.066

Sources:

Anderson, R. B. and Eriksson, A. J., “Lightning Parameters for Engineering Appli-

cations,”

Electra

, no. 69, pp. 65–102, March 1980a; Anderson, R. B. and Eriksson, A. J., “A

Summary of Lightning Parameters for Engineering Applications,” CIGRE Paper No. 33–06,

1980b; Berger, K., Anderson, R. B., and Kröninger, H., “Parameters of Lightning Flashes,”

Electra

, no. 41, pp. 23–37, July 1975.

)

131

≥

1791_book.fm Page 592 Monday, August 4, 2003 3:20 PM

Lightning Protection

593

between various parameters. Larger ﬁrst strokes tend to have longer rise

times. For ﬁrst strokes, the equivalent front risetime correlates some with

the peak current; the average rate of rise does not. For subsequent strokes,

the peak current is independent of the rise time, although the peak current

partially correlates with the rate of rise. For both ﬁrst and subsequent

strokes, the peak current correlates to some degree with the maximum rate

of rise. The correlations are not particularly strong in any of these cases.

Cigre used these interdependencies to ﬁnd derived distributions that are

useful in some stochastic simulations.

More than half of cloud-to-ground lightning ﬂashes are composed of more

than one stroke (see Figure 12.4). A quarter of them have at least four strokes.

The subsequent strokes usually have less current than ﬁrst strokes, but the

rate of rise of current is higher (important for the inductive voltage rise,

, in arrester leads, arrester spacings on lines, and traveling-wave issues in

cables). Subsequent stroke characteristics are thought to be independent of

the ﬁrst stroke.

A lightning ﬂash may last more than 1 sec; this impacts line reclosing

practices. If a circuit breaker or recloser clears a lightning-caused fault and

TABLE 12.2

Lightning Current Parameters for Downward Positive Flashes

Parameter

Percent of Cases More than Value

5% 50% (

) 95%

Peak current, kA 5 35 260 1.21

Time to peak,

sec

(30–90% virtual front)

322140 1.23

Stroke charge, C 2 16 84 1.36

Flash charge, C 18 80 350 0.90

Source:

Berger, K., Anderson, R. B., and Kröninger, H., “Parameters of

Lightning Flashes,”

Electra

, no. 41, pp. 23–37, July 1975.

FIGURE 12.4

Number of strokes in a ﬂash. (Data from [Anderson and Eriksson, 1980a,b].)

100

12345678910

Number of strokes in a flash

Probability of exceeding

the x-axis value

1791_book.fm Page 593 Monday, August 4, 2003 3:20 PM

594

Electric Power Distribution Handbook

immediately recloses, a subsequent stroke may ﬂash the line over again (or

continuing current in the ﬂash may maintain the fault arc). About 5% of

lightning ﬂashes will last beyond a typical immediate reclosing time of 0.5

sec (see Table 12.3). Normally, the extra fault will not cause any more dam-

age; the circuit breaker must open and reclose again. The practical effect of

this is to extend a half-second interruption to a 10-sec interruption (or

whatever the next delay time is in the reclosing sequence). It does show that

it is important to have more than one reclose attempt if an immediate reclose

is used.

A good percentage of multiple-stroke ﬂashes have subsequent strokes to

different points on the ground (Thottappillil et al., 1992). This implies that

ground ﬂash densities from ﬂash counters and lightning detection networks

may underestimate the number of lightning ﬂash ground terminations.

The lightning activity in an area can be measured. Many areas of the world

have lightning detection networks that measure the magnetic and/or electric

ﬁeld generated by a lightning stroke, determine if the stroke is from cloud

to ground, and triangulate the stroke’s position. Such systems help utilities

prepare for storms; information on storm intensity, direction, and location

helps determine the number of crews to call up and where to send them.

Maps generated from lightning detection networks of

ground ﬂash density

(GFD or

) are the primary measure of lightning activity. Figure 12.5 shows

a 10-year ground ﬂash density contour map of the U.S. from the U.S. National

Lightning Detection Network (NLDN), which has been operating since

before 1990.

Lightning detection networks are also useful for correlating faults with

lightning. This data helps with forensics and is even used in real time to

direct crews to damage locations. From personnal experience with correlat-

ing faults with the U.S. NLDN and with camera monitoring studies, the

system successfully captures about 90% of strokes. The most important char-

acteristic that allows accurate correlation of faults and lightning is accurate

time tagging of power system event recorders including power quality

recorders, SCADA, or fault recorders (GPS works well). Position accuracy

of detection networks is not good enough to determine if strokes hit a line,

but it is good enough to narrow the choices of strokes considerably — almost

TABLE 12.3

Probability of a Successful Reclose Following a

Lightning-Caused Fault

Duration of the Dead Time

Before the Reclose, sec

Probability of a

Successful Reclosure

0.3 83%

0.4 90%

0.5 95%

Source:

Anderson, R. B. and Eriksson, A. J., “Lightning

Parameters for Engineering Applications,”

Electra

, no.

69, pp. 65–102, March 1980a.

1791_book.fm Page 594 Monday, August 4, 2003 3:20 PM

Lightning Protection

595

all strokes found by the U.S. NLDN are accurate to within 1 mi (1.6 km),

with most accurate to 2000 ft (0.5 km).

If direct measurements of ground ﬂash density are unavailable, meteoro-

logical records of thunderstorms are available, most commonly the number

of days with thunderstorms (or keraunic level). Thunderstorm days approx-

imately relate to ground ﬂash density as (Eriksson, 1987):

Records of the number of thunderstorm hours per year relate to

(MacGorman et al., 1984):

where

is the number of thunderstorm hours per year.

Another crude estimate of the lightning level is from NASA’s Optical

Transient Detector, which measured worldwide lightning activity for 5 years

(Chisholm et al., 1999).

Lightning is highly variable. It takes several hundred lightning ﬂash counts

to obtain modest accuracy for an estimate of the average ﬂash density. A

FIGURE 12.5

Ground ﬂash density from the United States National Lightning Detection System. (Lightning

map provided by Vaisala.)

= 004

125

= 0 054

1791_book.fm Page 595 Monday, August 4, 2003 3:20 PM

596

Electric Power Distribution Handbook

smaller geographic area requires more measurement time to arrive at a

decent estimate. Similarly, a low-lightning area requires more measurement

time to accurately estimate the lightning. Standard deviations for yearly

measurements of lightning activity range from 20 to 50% of the mean (IEEE

Std. 1410-1997). Figure 12.6 shows the variability of ground ﬂash density in

a high-lightning area. Lightning and storms have high variability. Lightning

and weather patterns may have cycles that last many years.

Distribution lines cover small geographic areas, so lightning damage and

lightning ﬂashover rates show high variability on a circuit. One year a circuit

may get nailed with damage coming from many storms; the next year it may

have next to nothing.

The variability of lightning and the variability of storms is also important

for utility planning regarding regulatory incentives for reliability and for

performance guarantees for customers. Just a few years of data usually does

not accurately depict the performance of weather-related events for a circuit

or even for a whole system.

12.2 Incidence of Lightning

Lightning ﬂashes hit typical-height distribution lines that are in the open at

the rate of about 17 ﬂashes/100 circuit mi/year for an area with

= 1

FIGURE 12.6

Estimated annual ground ﬂash density for Tampa, Florida based on thunderstorm-hour mea-

surements. (Data from [MacGorman et al., 1984].)

1950 1960 1970

Year

Ground flash density, flashes/km

/year

1791_book.fm Page 596 Monday, August 4, 2003 3:20 PM

Lightning Protection

597

ﬂash/km

/year (11 ﬂashes/100 km/year). Flashes to lines vary directly as

the ground ﬂash density. Almost all of these ﬂashes to lines cause faults,

and some damage equipment. Taller lines attract more ﬂashes. The most

accepted model of lightning strikes to power lines in open ground is Eriks-

son’s model, which ﬁnds the number of strikes as a function of the line

height (Eriksson, 1987):

where

= ﬂashes/100 km/year to the line

= ground ﬂash density/km

/year

= height of the conductor (or overhead groundwire) at the tower, m

= overhead ground wire separation, m

For distribution lines, the

term can be ignored, which gives:

for

in meters and

in ﬂashes/100 km/year, or

for

in feet and

in ﬂashes/100 mi/year.

For a 30-ft (10-m) line, about 17 ﬂashes/100 mi/year (10.6 ﬂashes/100 km/

year) hit a distribution line for

= 1 ﬂash/km

/year. The main data point

for distribution lines is a South African test line heavily monitored in the

1980s (Eriksson, 1987). The test line had 19 ﬂashes/100 mi/year (12 ﬂashes/

100 km/year) normalized for

= 1 ﬂash/km

/year for a line height of 28

ft (8.6 m).

The equivalent shadow width or total width that a distribution line attracts

lightning is on the order of 360 ft (110 m). This is about 12 times the line

height. For many years, a shadow width of four times the line height was

used; this model underestimates the number of ﬂashes hitting lines.

More exposed lines — those with no trees nearby or lines on the tops of

ridges — attract more ﬂashes. Eriksson’s equation is for a line in open

ground. Many lines have fewer hits because of shielding from nearby objects

(mainly trees, but also buildings and other power lines). For lines in forested

areas, the number of strikes is signiﬁcantly reduced, so this equation should

be an upper limit for most lines. Some very exposed lines on hills may have

more hits than Eriksson’s equation predicts. The number of hits to a line in

a shielded area is:

06.

NNh

=◊28

NNh

=◊22

1791_book.fm Page 597 Monday, August 4, 2003 3:20 PM