Lightning Parameters
Lightning Parameters
Lightning Parameters
to the peak value ( , front time) and by the time to its subsequent decay to its half value ( , tail time). The tail time being several orders of magnitude longer than the front time, its statistical variation is of lesser importance in the computation of the generated voltage. The generated voltage is a function of the peak current for both the direct and indirect strokes. For backashes in direct strokes and for indirect strokes the generated voltage is higher the shorter the front time of the return-stroke current [1]. The front time (and the tail time, to a lesser extent), inuence the withstand capability (volt-time characteristics) of the power apparatus. The charge in a stroke signies the energy transferred to the struck object. The ancillary equipment (e.g., surge protectors) connected near the struck point will be damaged if the charge content of the stroke exceeds the withstand capability of the equipment. The return-stroke velocity will affect the component of the voltage which is generated by the induction eld of the lightning stroke [1]. Field tests have shown that the parameters of the rst stroke are different from that of the subsequent strokes. Lightning being random in nature, its parameters must be expressed in probabilistic terms from data measured in the eld. The objective of this report is to present the statistical data of the signicant parameters collected by many researchers over many years around the world. II. DATA ACQUISITION TECHNIQUES Compilation of lightning parameters is best accomplished by direct measurements on actual lightning. Data gathering can be accelerated by triggered lightning, whereby a rocket trailing a thin conducting wire is shot toward a charged cloud. The rocket is struck by lightning as it approaches the charged cloud and the trailing thin wire is evaporated by the heavy current ow, thus simulating the lightning channel. The rst stroke cannot be simulated by triggered lightning. It does simulate the subsequent stroke. As tall structures are struck more frequently by lightning, the return-stroke current has traditionally been measured by installing current transducers either at the top or the bottom of tall towers. The output of the current transducer is then fed into a recording device. The magnitude of the return-stroke current has also been measured by magnetic links, which are small bundles of high retentivity steel laminations about three centimeters long, placed at various locations on the shield wires and transmission-line tower legs. The currents owing through these parts magnetize the magnetic links, and the peak current can be estimated from the magnetization of the magnetic links. However, such measurements have long been discarded because of unreliability. The peak of the return-stroke current has also been estimated by measuring the radiated magnetic eld of the light, ning stroke. The relationship between the peak current,
I. INTRODUCTION
N accurate knowledge of the parameters of lightning strokes is essential for the prediction of the severity of the transient voltages generated across power apparatus either by a direct stroke to the power line/apparatus, or by a nearby lightning stroke (indirect stroke). However, no two lightning strokes are the same. Therefore, the statistical variations of the lightning-stroke parameters must be taken into account in assessing the severity of lightning strokes on the specic design of a power line or apparatus. The lightning return-stroke current and the charge delivered by the stroke are the most important parameters to assess the severity of lightning strokes to power lines and apparatus. The return-stroke current is characterized by a rapid rise to the peak, , within a few microseconds and then a relatively slow decay, reaching half of the peak value in tens of microseconds. The return-stroke current is specied by its peak value and its waveshape. The waveshape, in turn, is specied by the time from zero
Manuscript received March 28, 2003. Paper no. TPWRD-00144-2003. P. Chowdhuri, J. G. Anderson, W. A. Chisholm, T. E. Field, M. Ishii, J. A. Martinez, M. B. Marz, J. McDaniel, T. R. McDermott, A. M. Mousa, T. Narita, D. K. Nichols, and T. A. Short are members of the Task Force 15.09 on Parameters of Lightning Strokes. Digital Object Identier 10.1109/TPWRD.2004.835039
347
and the radiated electric eld, , was derived from the transmission-line model of the lightning stroke for a lossless earth [2]: and (1)
and =coefcient of correlation. If x and y are independently distributed, then . The cumulative probability that and :
, and
where c=velocity of light in free space, D=distance of the stroke from the antenna, =velocity of the return-stroke, and =peak magnetic induction. III. STATISTICAL DISTRIBUTION OF LIGHTNING STROKE PARAMETERS From eld data on lightning strokes to masts, chimneys, etc., the statistical variation of the lightning stroke parameters can be approximated by a log-normal distribution, where the statistical variation of the logarithm of a random variable, x, follows the normal (Gaussian) distribution. In that case, the probability density function, p(x), of x is given by [1], [3], [4]: (2)
(5) where . Similarly, if probability is given by: , and , the joint cumulative
(6) The conditional probability density function of y for a given can be found by change of variables [5], [6]: (7a)
(7b) where where =standard deviation of , and =median value , the cumulative of x. Putting, probability, , that the parameter will exceed x, is given by integrating (2) between u and , giving:
and (8a)
(3)
As an example, if the critical current of ashover of an overhead power line is 20 kA, then from Table I, and .
This new log-normal distribution of y has then a median value, , which is the antilog of b and a standard deviation, . b can be written in an alternate form:
(8b) ; or . That is, the probability of a negative rst-stroke current greater than 20 kA is 82.11%. The joint probability density function of two stroke parameters, x and y, can be expressed as: or where (8d) and (8e) where Such relationships, i.e., (8c), among lightning parameters have been found and are shown later (Table XI). For cumulative to : probability of y from and , By putting (9) (8c)
(4)
348
If, however, the conditional is for x within a range, e.g., , then (7a) needs to be integrated:
to
(10) Two examples of the conditional probability are shown below. In the rst example, the limiting current is 20 kA, i.e., for a given front time of current, . Assuming , , memedian current from Table I, and , and applying dian front time, (9), the cumulative probability is shown below for three values of the correlation coefcient, , between and .
Such analysis is applicable for estimating outage rates for strokes to nearby ground and also for backash outages. Without the conditional of front time, the cumulative probability of curby applying rent exceeding 20 kA, (3). The second example pertains to gapless MOV surge protectors for the case of a lightning stroke hitting a phase conductor of a shielded line [7]. For a perfectly shielded line, the shielding current will be equal to the critical current, and no insulator ashover will occur. However, currents lower than the shielding current may hit the phase conductors; the charge in the lightning ash will then be absorbed by the surge protector. If this charge exceeds the limit of the protector, then the surge protector , will be damaged. Assuming the shielding current, (i.e., and the limiting charge of the surge protector, ), what is the probability for , given ? The statistical parameters of the , positive-polarity lightning ash are as follows: , , and . From and , taking the lower limit of to (10), putting be very small, e.g., : and
Note 1: References [4], [9][11] suggest that the measured distribution of the rst negative stroke is better approximated by two straight lines intersecting at 20 kA when plotted on a , probability paper. Then, for and ; for , , and . However, the log-normal approximation of the , and entire distribution can be represented by: . Note 2: is time interval between the 10% and 90% of the current peak on the current wavefront, . ; ; ; ; =max. current rate of rise on wavefront. . Note 3: Without the conditional of ,
and
The log-normal characteristics of the negative-polarity rst stroke are shown in Table I where is the standard deviation of the log (base e) of the variate. Most of the data were taken by Berger [8], which were later analyzed and updated [4], [9][11]. Fig. 1 shows the waveshape of the typical return-stroke current as recorded by Berger. , of the reThe cumulative probability distribution, turn-stroke current, , can be estimated from (3) by replacing x with (median value). The values of and with and are given in Table I. A much simpler form for , approximating the log-normal distribution for the return-stroke current in (3), was given by Anderson [12]: (11)
From (10), the probabilities for three values of the correlation coefcient, , are shown below:
Reference [13] provides data from eld tests during 19941997 in Japan. Sixty 500-kV double-circuit transmission towers with overhead shield wires were instrumented. The towers included 1000-kV design, but operating at 500 kV. The tower height ranged from 40 m to 140 m, and the altitude of the observed sites varied from 150 m to 1500 m. The lightning stroke currents were measured by Rogowski coils, attached to
349
The cumulative probability that a subsequent-stroke current will exceed a given level, , can be estimated, similar to (3), with the statistical parameters of (12). A simplied equation, similar to (11) has also been proposed [14]: (13)
Fig. 1.
Note:
;
TABLE III STATISTICAL DISTRIBUTION OF MULTISTROKE NEGATIVE LIGHTNING FLASHES [10]
Although the median value of the subsequent stroke current is lower than that of the rst stroke, the individual value of a subsequent-stroke current can be higher than the preceding rststroke current; factors as high as 200% have been recorded [11]. The physical phenomena associated with articially triggered lightning are believed to be similar to that of the subsequent stroke of natural lightning. However, there may be potential disparities between triggered lightning and the subsequent stroke of natural lightning [15]: (i) the triggered discharge occurs under cloud conditions where a discharge may not have occurred under natural conditions, (ii) the lower portion of the triggered lightning channel may be contaminated by metal vapor. However, in spite of the possible differences between triggered lightning and subsequent strokes of natural lightning, it will be interesting to compare the eld-test results. Fisher et al. [15] have reported extensive test results of triggered lightning, and have compared the various parameters obtained from their tests and those of Berger [8] on subsequent strokes from natural lightning, as reported by Anderson and Eriksson [10]. These comparisons are shown in Table IV. It should be mentioned that in their triggered lightning eld tests, Fisher et al. found very little or no correlation between and time to half value on the current tail. peak current and There were, however, strong correlations between the peak cur(correlation coefcient, ) and rent and .
VI. POSITIVE STROKES Less than 10% of the ground ashes are of positive polarity. However, the incidence of positive ground ashes varies seasonally, being more frequent in the winter. It also varies globally. Also, very tall structures produce upward positive strokes, in contrast to the usual downward strokes. Reference [8] states that the analysis was made only on the downward ashes. However, Berger suggested later that these strokes were upward negative leaders followed by downward ash from positively-charged cloud [16]. The parameters of the positive stroke/ash are given in Table V. The median value (35 kA) of the positive-stroke current in Table V is somewhat higher than that of the rst negative-stroke current. However, this could be misleading because the maximum value of the positive-stroke currents are signicantly higher than that of the rst negative-stroke current. According to [8], 5% of the positive strokes exceed 250 kA, the corresponding magnitude of the rst negative strokes being only 80 kA. The incidence of positive strokes signicantly increases during the winter months. Winter lightning data were collected
2.5-m long rods on the top of the towers. The amplitude of the peak current was found to be dependent neither on tower height nor on altitude. The statistical data are shown in Table II. V. SUBSEQUENT NEGATIVE RETURN-STROKE CURRENTS A ground ash very frequently consists of multiple strokes. Based on a survey of almost 6000 ash records from different regions of the world, Anderson and Eriksson estimated the following percentages (Table III) of multiple strokes in a ground ash [10]. In general, there is no correlation between the rst and the subsequent stroke peak amplitudes. The median value of the subsequent stroke is signicantly lower than that of the rst stroke. Assuming log-normal distribution, the median value and the standard deviation of the subsequent stroke have been proposed by Eriksson as [9]: and (12)
350
TABLE IV COMPARISON BETWEEN TWO STUDIES ON NEGATIVE SUBSEQUENT-STROKE CURRENT PARAMETERS [10], [15]
Note:
; VII. TYPICAL LIGHTNING CURRENT WAVESHAPES More than 90 percent of the cloud-to-ground strokes are of negative polarity, except for seasonal and regional variations. According to [8], the positive-polarity stroke currents do not have enough common features to produce an acceptable mean waveshape. This could also be partly due to the small number of positive strokes recorded. The waveshape of the mean negative rst stroke current is shown in Fig. 1. Most of the data came from Bergers work on Mount San Salvatore in the southern part of Switzerland, collected by a 60-m mast. This waveshape has distinctly a concave wavefront with the greatest rate of change near the peak. Many of the current waves have two peaks, the second one being higher in magnitude. The front time is based on the rst peak, and the peak amplitude on the second peak. The negative subsequent stroke current has, in general, shorter wavefront than that of the negative rst stroke current. The negative subsequent stroke currents do not show the pronounced concavity of the wavefront of the rst stroke current. This is shown in Fig. 2 [4]. The concavity of the negative rst stroke current, i.e., the initial slow rise followed by fast rise, may be attributed to the upward streamer from the object to be struck reaching out to the downward streamer from the cloud [4]. The slow-rising upward streamer carries comparatively small current. However, when the upward streamer meets the downward leader, the current rises fast. As the subsequent strokes are not preceded by upward streamers, the wavefront of these strokes do not show the concavity. Several empirical equations have been proposed for the waveshape of the negative rst stroke current [9], [11], [18], of which the equation proposed in [18] has been widely used. This is given by: (14) where =peak current, =correction factor of the peak cur, , =time constants determining current riserent, and decay-time, respectively, and n=current steepness factor. It was stated in [18] that the usual double-exponential function to represent a transient waveshape has a discontinuity of its rst ; therefore, it is not convenient for the LEMP derivative at calculations. This difculty does not arise with (14).
Note 1: (front time) is the time interval between 2-kA point on front and the rst peak. (stroke duration) is the time interval between 2-kA point on front and the 50% of peak current on tail. ; . Note 2: Numbers in parenthesis are for negative rst strokes. Note 3: .
TABLE VI STATISTICAL PARAMETERS OF POSITIVE STROKES IN WINTER [17]
Note: ; is the time interval between the start of the wave and the 50% of peak current on tail. at Fukui (at sea level) in Japan [17]. The height of the measurement tower was 200 m. The statistical data on the winter positive lightning strokes are given in Table VI. No statistical difference was found between the cumulative statistical distributions of the peak values of the positive- and negative-polarity currents. All these incidents were upward strokes. Two types of lightning were reported in Fukui [17]: one type with high peak currents and strong luminosity of the lightning channel (type-A), and the other type with small current peaks and weak lightning-channel luminosity (type-B). Comparing Tables IVVI, it should be noticed that the front time and duration of the positive strokes are signicantly longer and the front steepness is lower than that of the negative strokes. The same is true for the winter positive strokes compared to that of the summer positive strokes.
351
E = I dt [8], [15]
VIII. STROKE CHARGE Most of the charge delivered by lightning ashes does not occur during the current pulses with the high current peaks. Instead, it is contained in the slow continuing low-magnitude currents between or after the high current peaks. To some extent, a ash behaves like an arc welder as far as surface ablation and arc ignition is concerned. Reference [8] provides observational results for a large number of ashes. However, for delivered charge, statistics of the highest magnitudes of charge are of most concern, and only a few observations always exist at the end of any probability curve. Hence, for the data of most interest, the probable error is the highest. Following [8] and assuming log-normal probability distribution, the parameters for the statistical distribution of the stroke/ash charge were developed and given in Table VII. The numbers in parenthesis in Table VII are from [15]. The following approximate cumulative probability equations for delivered charge were developed from data in [8], where is the probability that the charge Q (in coulombs) will be exceeded in a single ash. Total charge delivered by a negative ash: (15) Total charge delivered by a positive ash: (16) Charge delivered by a negative rst stroke: (17) The charge delivered by positive and negative strokes is only within the rst two milliseconds. Charge beyond that time is classied as in a continuing current. Another way to assess the thermal severity of a lightning ash of the ash. Table VIII shows is to estimate the integral of is the median value of . The numthe data from [8]. bers in parenthesis are from [15]. is a measure of thermal It should be borne in mind that severity if the current ows into a constant resistance. For most lightning strikes the current ows into either a cathode spot whose voltage drop is quasiconstant or into an impedance that reduces dramatically as current increases making much less heating. IX. RETURN-STROKE VELOCITIES The eld data from four papers [19][22] were investigated. In [19], both the straight-line velocity and the track (two-dimensional) velocity were tabulated for 36 strokes each. Of the 36 points, only 7 were for the rst stroke. In [20], 16 more measurement points were given. However, they were not given in tabular form, and the velocities were plotted without differentiating between the rst and the subsequent strokes. Therefore, the data from [20] could not be used. Of the 14 data points in [21], only 4 were from the rst stroke. In [22], of the 63 data points, 17 were for the rst strokes. Hence, of the 113 measured velocities, 28 were for the rst stroke and 85 were from the subsequent strokes. Table IX compares the mean and the standard deviation of the return-stroke velocity for both the rst and the subsequent strokes. It has been observed that the return-stroke velocity, for both the rst and the subsequent strokes, decreases as the stroke progresses upwards toward the cloud [22]. Therefore, the average
352
TABLE IX COMPOSITE FIELD DATA ON VELOCITY OF RETURN STROKES NEAR GROUND [19], [21], [22]
TABLE X COMPOSITE FIELD DATA ON RETURN-STROKE VELOCITY [19], [21], [22] DATA FROM REF. [22] FOR CHANNEL LENGTH OF AT LEAST 0.7 km
velocity measured over a longer channel length will be lower than that for a shorter channel length. In [22], two sets of data were given; one set for observations at ground levels, and the other set for channel lengths of at least 0.7 km. These data, together with the data from [19] and [21] are shown in Table X. There is signicant disparity in results among the three studies. These differences may be attributed to: i) region; ii) sample size; iii) channel length; iv) experimental error. The tests in [19] were performed in South Africa; the tests in [21] were performed in Albany, NY; and the tests in [22] were at the Kennedy Space Center in Florida and at the Langmuir Laboratory near Socorro, NM. The mean rst return-stroke velocities and 150 , in Florida and New Mexico were 66 respectively; similarly, for the subsequent strokes 110 and 130 , respectively. The measurement error in [21] was estimated to vary between 30 to 60%, and the maximum error in [22] was estimated to be 35% or less. The estimated error in [19] is not known. In [21], some measurements were taken within 300 m of the ground, and some within 1 km of the ground. In [22], some measurements were taken near ground (1.3 km or less), and some were taken over a minimum of 0.7 km of channel measured from the ground. For [19], the channel length and height are not exactly known, but is estimated to be longer [22].
As the return-stroke currents were not measured concurrently, the cumulative distribution of velocity was calculated rst from the eld data, and then this distribution was matched with the CIGRE cumulative distribution of current [2], [9]. The pertinent log-normal parameters of the currents have been shown in Table I. Two empirical equations relating the velocity to the current of the rst stroke are widely used. One equation was proposed by Lundholm [23] and Rusck [24], and the other by Wagner [25]. These equations are plotted in Fig. 3. The disparity is caused mainly because the old AIEE current distribution was assumed in the derivation of these equations. A relationship between the return-stroke current and its velocity is proposed: (20) The velocity is plotted as a function of the return-stroke current, , in Fig. 4. X. CORRELATIONS BETWEEN LIGHTNING PARAMETERS As shown in Section III, correlation between lightning parameters signicantly inuences the estimation of the cumulative probability. Once the correlation coefcient, , between
353
Fig. 3.
equation.
= 31:1 kA,
Velocity vs. rst-stroke current from composite eld data. = 0:48. (a) Lundholm-Rusck equation; (b) Wagner
tional Lightning detection network (NLDN) by Global Atmospherics, Inc. The recent improvements of NLDN has been described in [33], [34]. The data shown in Tables XII and XIII are from the central, northwest and southeast regions of U.S.A. for four lightning seasons, represented in two 2-year periods (19971998 and 19992000). These three regions were selected to represent the most extreme differences in the characteristics. The areas of the three regions are rectangular, designated with the southwest and northeast corners by the latitudes and longitudes of these corner points. The log-normal plots of the cumulative probabilities are shown in Figs. 57. The absolute uncertainty in peak current is 2030% which is due mainly to modeling errors. The random error between regions is small due to the large number (typically 67) of sensors that are used to estimate the peak current for each individual ash. were The median current and the standard deviation computed from the raw data provided by Global Atmospherics, Inc. As there is no signicant regional variation in the instrumentation, the differences in the lightning parameters are predominantly due to the difference in the climates in the three regions. It should be noted that the cumulative probability proles do not entirely t the log-normal distribution. They seem to have different slopes in the entire range of current, similar to the two-slope characteristic of the Berger data [4]. It should also be noticed that the median value of the positive strokes does not always exceed that of the negative strokes, e.g., southeast region of the USA. The small percentage of positive ashes is probably biased by the misclassication of some small positive cloud to-cloud discharges as cloud-to-ground ashes [33]. XII. DISCUSSION Most of the measurements reported here were taken on tall towers with current transducers either located at the top or the bottom of the structure. There are several sources of error associated with such measurements. First, the measured median current will be different from that to at ground [26]. Second, reections at both ends of the tower of the traveling current waves along the tower will distort the recorded current wave. In recent years, from the National Lightning Detection Network (NLDN), the return-stroke current is estimated from the radiated magnetic eld of the lightning stroke by (1), assuming the transmission-line model of stroke channel. Several errors are encountered in this method of measurement: i) the return-stroke velocity is a function of the peak current; therefore, the assumption of a constant velocity is incorrect; ii) several models of the return stroke have been proposed; none has been accepted as superior to the others; iii) for nearby strokes, the assumption of the radiation eld is not acceptable; iv) even when the stroke is distant, the radiated eld is attenuated when it reaches the antenna, the degree of attenuation being a function of the ground resistivity. The NLDN system was calibrated with peak currents from triggered lightning return strokes lowering negative charge measured at the NASA Kennedy Space Center, Florida. The radiated eld of the triggered lightning was measured by six sensors, one in Georgia and ve in Florida, ranging from 117.9 km to
Fig. 4.
= 0:48.
= 31:1 kA,
the current and another parameter, y, is known, then the effective median value of the variate can be found from (8a), and the probability density function can be estimated from (7b). It should be borne in mind that certain uncertainties exist in the estimation of . Table XI shows , a and d of (8c), and of (8a). was taken from [4] and [8]; a and d were computed was computed from from (8d) and (8e), respectively; was taken from Tables I and V for the negative (8a) where rst strokes and the positive strokes. The values of for the negative subsequent strokes were computed from the 95% and 5% cumulative probabilities given in Table I of [8]. XI. REGIONAL VARIATION OF RETURN-STROKE CURRENT The regional variation of the return-stroke current is illustrated in Tables XII and XIII. The data was taken from the Na-
354
y x = ax
426.8 km from the trigger site [36]. The tests were later repeated with about three fold larger data set [37]. A relationship between the peak current and the magnetic signal strength was proposed [2]: (21) where SS is the signal strength of the magnetic eld in arbitrary . This assumed a return-stroke velocity to units and
. However, a triggered lightning does not repbe resent a natural lightning. Moreover, the return-stroke velocity in a natural lightning is related to the peak current. Using this relationship from data on negative triggered lightning to positive strokes is highly unjustied. The attenuation of the radiated eld will depend upon the soil resistivity as well as the frequency (waveshape) of the radiated signal. Therefore, application of (21) to other natural lightning and to other regions would result in signicant error.
355
Fig. 5.
Cumulative probability distribution of lightning strokes in the central region of U.S.A. (a) Negative strokes; (b) positive strokes.
Fig. 6.
Cumulative probability distribution of lightning distribution in the northwest region of the USA. (a) Negative strokes; (b) positive strokes.
Additionally, this method estimates only the current peak; it cannot estimate the waveshape of the current. Reference [27] provides a comprehensive discussion on the limitations in the measurement of lightning parameters. The amplitude of the return-stroke current being the most important parameter of lightning in estimating the severity of the overvoltage across insulators, an urgent need exists to develop new techniques to measure lightning return-stroke current. One possibility is to measure the intensity of luminosity of the lightning channel and relate it to the current amplitude [17]. Several attempts have been made to measure the return-stroke luminosity [17], [28][30]. The proles of the channel luminosity against time showed striking resemblance to the double-exponential impulse current wave. The cumulative probability distribution of the channel luminosity distribution also showed resemblance to the cumulative probability distribution of the cur-
rent [30]. However, the analysis of [30] showed the relationship between the luminosity and current is neither linear nor quadratic. Although a denite correlation was found in [17], no mathematical formulation was given. However, as was pointed out in [17], atmospheric conditions, such as rain and fog, will distort the luminosity and will pose a problem in the calibration. Another possibility is the spectroscopic study of the lightning channel to determine its electrical characteristics. The front time of the return-stroke current is another important parameter which is often overlooked. Shorter front time will produce higher voltages across insulators for both direct and indirect strokes [1]. Therefore, this parameter needs to be measured accurately, and an analytical expression which closely follows the eld data should be specied. The present standards specify a double-exponential mathematical expression to represent the lightning return-stroke
356
Fig. 7. Cumulative probability distribution of lightning distribution in the southeast region of the USA. (a) Negative strokes; (b) positive strokes.
currents. However, questions have been raised about the adequacy of this double-exponential waveshape since the publication of Bergers data showing concave wavefront of the negative-polarity rst stroke. Of the several analytical expressions suggested for the concave current wavefront, the one proposed by Heidler [18] and shown in (14) has been widely used. Three examples of waveshape plotted by using (14) are shown in Fig. 8. None of the three examples in Fig. 8 resembles Fig. 1. The following questions need to be addressed for considering a concave wavefront to be a standard: a) Is the concavity caused by the upward streamer from the struck tower? If the upward streamer is responsible for the concavity, then the concave wavefront should not be standardized. Many, perhaps most, wavefronts of the return stroke do not show the concave characteristic. b) How will the concave wavefront be specied? The front . In addition, the time may be specied as in should be specied maximum steepness ( along with its location on the wavefront. The severity of insulator voltage stress caused by direct strokes is not a function of the return-stroke velocity. However, the induced voltage is a function of return-stroke velocity for indirect lightning strokes [1]. Moreover, it has been postulated that the return-stroke velocity is a function of the return-stroke current, increasing with increase of the current peak [23][25]. Therefore, the relationship between the current and the velocity of the return stroke needs to be known to estimate the voltage induced by the indirect stroke. Simultaneous measurement of the return-stroke velocity and the current has not been done in the previous studies; velocity and current were matched on the basis of equal probability of occurrence, e.g., the median value of the velocity was matched with the median value of the current [23][25]. Simultaneous measurement of velocity and current is highly desirable. All eld data show that the rst stroke peak current is significantly higher than the subsequent stroke currents for the nega-
tive strokes; however, the steepness of the rst negative stroke is less than that of the subsequent negative strokes. Therefore, it is possible for an insulator to survive the rst stroke but to ash over during the subsequent stroke. The volt-time characteristics of the insulator under voltages of different front times will also play a decisive role in its survival. The median value of the peak positive stroke current is somewhat higher than that of the negative stroke (Table V). The steepness of the positive stroke current is signicantly lower and its duration is longer than that of the negative stroke. Therefore, the voltage across an insulator will be lower under a positive stroke. However, it may spark over because of the longer front time and time to half value of the applied voltage. Therefore, research on the volt-time characteristics of insulators under nonstandard lightning voltages for both polarities of voltage should have priority. Because of the signicantly longer duration of the positive are higher than that of the negative stroke, its charge and stroke. This may increase ablation damage at its terminal point. Worse still, a positive stroke may exceed the thermal capability of a surge protector because of larger charge (Table VII). The NLDN data shown in Tables XII and XIII, and in Figs. 57 are widely different from the data for the other parts of the world, shown in the previous Tables. The NLDN median currents of both polarities are signicantly lower than those of the other parts of the world. It appears that lightning statistics vary signicantly from one region to another and also from one season to another in the same region, such as: (i) return-stroke velocities (Tables IX and X) in South Africa [19], Albany, NY. [21], Florida and New Mexico [22], (ii) median currents (Tables XII and XIII). Latitudinal variation of lightning characteristics has been suggested [31]. By analyzing data from New York to Florida and to the west up to the Mississippi River, Orville suggested that the peak return-stroke current is higher in the southern latitudes and decreases with increase in the latitude [32]. He proposed that the longer lightning channels in the south, caused by the higher alti-
357
It should also be borne in mind that the instrumentation used by the various researchers at different times were different. The measurement accuracy in most cases is not known. One obvious difference is in the trigger level. Bergers experiments had a trigger level of 2 kA [8], [10], whereas those in [13] were 9 kA. Uniform standards for instrumentation should be formulated. Lastly, correlation among the various lightning parameters is an important parameter which should not be ignored. Two examples were given in Section III of the signicance of correlation on conditional probabilitycurrent front time and charge current. These were simple computations. Computations can get more involved in the estimation of outage rates. As an example, the outage rates caused by lightning strikes to nearby are given below ground of a 10-m high line of [35]. for a ground ash density,
Because of this signicant inuence of the correlation coefcient, , on the lightning performance of power lines, this parameter needs to be estimated accurately. XIII. CONCLUSIONS AND RECOMMENDATIONS Negative rst strokes have traditionally been considered to produce the worst stress on transmission-line insulation. Subsequent negative strokes have signicantly lower peak current but shorter wavefronts. These subsequent strokes may stress the system insulation more in some cases, particularly for low footing resistances and tall structures. Positive strokes have about the same median current value as the negative rst strokes and longer fronts. However, the extreme current values of positive strokes tend to be higher than the negative strokes; hence both positive and negative strokes should be considered in the lightning simulations of overhead power lines. Positive strokes may also cause more thermal damage because of their signicantly higher delivered charge . and Although it has been postulated that the return-stroke velocity is related to the return-stroke current, the current and the velocity have not been measured simultaneously. Since the return-stroke velocity is a signicant parameter in estimating the lightning-induced voltages and also in estimating the return-stroke currents from measurements of the radiated electromagnetic eld of the lightning channel, more research is needed to relate the currents and their associated velocities. Better methods for making remote measurements of stroke current magnitudes and waveshapes need to be developed, as well as formulation of lightning parameters according to geographic region and season instead of assuming that they are a globally unied data set. In making simulations of lightning performance of overhead power lines, conservative values of stroke parameters are advised in presence of the many uncertainties that presently exist. Until these uncertainties are resolved, it is prudent to use those stroke values obtained by direct oscillographic measurements
Fig. 8. Examples of return-stroke current plotted from (14). (a) I = 30 kA; = 0:96; n = 1:0 = 0:03; = 68:0; (b) I = 30 kA; = 0:97; n = 10:0 = 1:0; = 68:0; (c) I = 30 kA; = 0:79; n = 10:0 = 10:0; = 68:0.
tude of the center of the negatively-charged region in the cumu) may contribute to the higher peak lonimbus cloud (at current in the southern latitudes. Apart from the meteorological conditions, the soil resistivity may also be a factor in inuencing the lightning stroke characteristics (e.g., front time). Therefore, it may be appropriate not to have global statistical parameters for lightning, but regional and seasonal.
358
and to recognize that approximations are inevitable. It is recommended that until more data are available: 1) The CIGRE waveshape (Fig. 1) be used whenever possible. 2) Table I be used for negative rst strokes, the AndersonEriksson part of Table IV be used for negative subsequent strokes, and Tables V and VI be used for positive strokes. 3) The eld-test return-stroke velocity as a function of return-stroke current in Fig. 4 be tentatively adopted. 4) The NLDN data on stroke magnitudes be viewed with caution until the validities of the various assumptions made in the analysis can be resolved. 5) The approximation equations [(11) and (13)] and [(15)(19)] be used for cases where local data are not available. However, it should be recognized that the extreme values at very low and high magnitudes are inadequate. ACKNOWLEDGMENT The raw data of the NLDN system was provided by the Vaisala-GAI, Inc. The Task Force acknowledges the fruitful critique provided by Dr. K. L. Cummins. REFERENCES
[1] P. Chowdhuri, Electromagnetic Transients in Power Systems. Taunton, U.K.: Research Studies, 1996. [2] Performance Evaluation of the National Lightning Detection Network in the Vicinity of Albany, New York, Electric Power Research Institute, Palo Alto, CA, EPRI Rep. TR-109 544, 1997. [3] A. R. Hileman, Insulation Coordination for Power Systems. New York: Marcel Dekker, 1999. [4] R. B. Anderson and A. J. Eriksson, Lightning parameters for engineering applications, Electra, no. 69, pp. 65102, Mar. 1980. [5] A. Hald, Statistical Theory With Engineering Applications. New York: Wiley, 1952. [6] R. V. Hogg and A. T. Craig, Introduction to Mathematical Statistics, 5th ed. Englewood Cliffs, NJ: Prentice-Hall, 1995. [7] M. Bernardi, L. Dellera, and E. Garbagnati, Lightning parameters for protection: an updated approach, in Proc. Int. Conf. Lightning Protection, Birmingham, U.K., 1998. [8] K. Berger, R. B. Anderson, and H. Kroninger, Parameters of lightning ashes, Electra, no. 41, pp. 2337, Jul. 1975. [9] A. J. Eriksson, Notes on Lightning Parameters for System Performance Estimations, CIGRE Rep. 33-86 (WG 33-01)IWD, 1986. [10] R. B. Anderson and A. J. Eriksson, A summary of lightning parameters for engineering applications, in Proc. CIGRE, 1980, Paper no. 33-06. [11] Guide to Procedure for Estimating the Lightning Performance of Transmission Lines, CIGRE Brochure 63, Oct. 1991. [12] J. G. Anderson, Lightning performance of transmission lines, in Transmission Line Reference Book 345 kV and Above, 2nd ed. Palo Alto, CA: Elect. Power Res. Inst., 1987, ch. 12. [13] T. Narita, T. Yamada, A. Mochizuki, E. Zaima, and M. Ishii, Observation of current waveshapes of lightning strokes on transmission towers, IEEE Trans. Power Delivery, vol. 15, pp. 429435, Jan. 2000. [14] IEEE Guide for Improving the Lightning Performance of Transmission Lines, IEEE Std. 1243-1997.
[15] R. J. Fisher, G. H. Schnetzer, R. Thottappillil, V. A. Rakov, M. A. Uman, and J. D. Goldberg, Parameters of triggered-lightning ashes in Florida and Alabama, J. Geophys. Res., vol. 98, no. D12, pp. 22 88722 902, Dec. 20, 1993. [16] K. Berger, The earth ash, in Lightning, R. Golde, Ed. New York: Academic, 1977, vol. 1, ch. 5. [17] A. Asakawa, K. Miyake, S. Yokoyama, T. Shindo, T. Yokota, and T. Sakai, Two types of lightning discharges to a high stack on the coast of the sea of Japan in winter, IEEE Trans. Power Delivery, vol. 12, pp. 12221231, Jul. 1997. [18] F. Heidler, J. M. Cvetic, and B. V. Stanic, Calculation of lightning current parameters, IEEE Trans. Power Delivery, vol. 14, pp. 399404, Apr. 1999. [19] B. F. J. Schonland and H. Collens, Progressive lightning, in Proc. Royal Society, vol. 143, Ser. A, 1934, pp. 654674. [20] B. F. J. Schonland, D. J. Malan, and H. Collens, Progressive lightning II, in Proc. Royal Society, vol. 152, Ser. A, 1935, pp. 595625. [21] J. S. Boyle and R. E. Orville, Return stroke velocity measurements in multistroke lightning ashes, J. Geophys. Res., vol. 81, no. 24, pp. 44614466, Aug. 20, 1976. [22] V. P. Idone and R. E. Orville, Lightning return stroke velocities in the Thunderstorm Research International Program (TRIP), J. Geophys. Res., vol. 87, no. C7, pp. 49034916, Jun. 20, 1982. [23] R. Lundholm, Induced overvoltage-surges on transmission lines and their bearing on the lightning performance at medium voltage networks, in Trans. Chalmers Univ. Technol., Gothenburg, Sweden, no. 188, 1957. [24] S. Rusck, Induced lightning over-voltages on power transmission lines with special reference to the over-voltage protection of low-voltage networks, in Trans. Royal Inst. Technol., Stockholm, Sweden, no. 120, 1958. [25] C. F. Wagner, Relation between stroke current and velocity of the return stroke, AIEE Trans., pt. III, vol. 82, pp. 606617, 1963. [26] A. M. Mousa and K. D. Srivastava, The implications of the electrogeometric model regarding effect of height of structure on the median amplitude of collected lightning strokes, IEEE Trans Power Delivery, vol. 4, pp. 14501460, Apr. 1989. [27] Characterization of Lightning for Applications in Electric Power Systems, CIGRE Brochure 172, Dec. 2000. [28] E. P. Krider, Time-resolved spectral emissions from individual return strokes in lightning discharges, J. Geophys. Res., vol. 70, no. 10, pp. 24592460, May 15, 1965. , Some photoelectric observations of lightning, J. Geophys. Res., [29] vol. 71, no. 12, pp. 30953098, Jun. 15, 1966. [30] C. Guo and E. P. Krider, The optical and radiation eld signatures produced by lightning return strokes, J. Geophys. Res., vol. 87, no. C11, pp. 89138922, Oct. 20, 1982. [31] E. T. Pierce, Latitudinal variation of lightning parameters, J. Appl. Meteor., vol. 9, pp. 194195, 1970. [32] R. E. Orville, Peak-current variations of lightning return strokes as a function of latitude, Nature, vol. 343, pp. 149151, Jan. 11, 1980. [33] K. L. Cummins et al., A combined TOA/MDF technology upgrade of the U.S. national lightning detection network, J. Geophys. Res., vol. 103, no. D8, pp. 90359044, Apr. 27, 1998. [34] K. L. Cummins, E. P. Krider, and M. D. Malone, The U.S. national lightning detection network and applications of cloud-to-ground lightning data by electric power utilities, IEEE Trans. Electromagn. Compat., vol. 40, pp. 465480, Nov. 1998. [35] P. Chowdhuri, Estimation of ashover rates of overhead power distribution lines by lightning strokes to nearby ground, IEEE Trans. Power Delivery, vol. 4, pp. 19821989, Jul. 1989. [36] R. E. Orville, Calibration of a magnetic direction nding network using measured triggered lightning return stroke peak currents, J. Geophys. Res., vol. 96, no. D9, pp. 17 13517 142, Sep. 20, 1991. [37] V. P. Idone, A. B. Saljoughy, R. W. Henderson, P. K. Moore, and R. B. Pyle, A reexamination of the peak current calibration of the national lightning detection network, J. Geophys. Res., vol. 98, no. D10, pp. 18 32318 332, Oct. 20, 1993.