Beruflich Dokumente
Kultur Dokumente
F. D’Alessandro
Abstract
The aim of this study was to assess the performance of one type of air terminal installed on a wide range of
practical structures. With the procurement of reliable raw data from Hong Kong, it has been possible to collate
lightning strike statistics for almost 200 suitable installations over the period 1988-1996. Analysis of these data
shows that there is a highly significant positive correlation between the actual and expected strikes to the
terminal per annum. Furthermore, the data show that the “yield” of the air terminal is higher than the claimed
protection level. The result provides strong evidence that the air terminal being assessed, along with the
lightning protection design principle, is performing according to its specifications. The application of this
analysis and the design concept to the protection of power generation, transmission and distribution facilities is
also described.
1. INTRODUCTION latter is called the keraunic level of the region and the
corresponding contour line plot of TSD is called an
Lightning is an extremely variable phenomenon isokeraunic map. Many empirical relationships have
which necessarily involves the use of statistics. For been derived between TSD and GFD over the last
example, the probability P(n) that a given area will be twenty years or so, e.g., see [1], [2]. Standards
struck by lightning exactly 0, 1, 2, 3 ..... n times over organisations have adopted the more common ones.
a given period of time may be expressed by the The IEC 1024-1-1 states that the flash density may be
Poisson relation, estimated using the relationship:
P(n) = e−µ µn / n! (1) GFD = 0.04 (TSD)1.25 (3)
where µ is the average number of strikes to the area.
The GFD estimates from this relationship are very
The same can be said about lightning protection and
much upper limits. Popolansky [3] recently obtained a
risk assessment where, for example, one is interested
relationship using lightning flash counter data from
in the probability of a strike to a particular building or
14 countries spanning 4 continents. After an
other structure.
exhaustive analysis, the following relationship was
derived:
In general, two factors determine the probability of a
GFD = 0.0086 (TSD)1.45 (4)
strike to a structure, namely the:
2
• Local ground flash density, GFD (strikes/km /yr) The estimates obtained from this relationship are
• Attractive (or exposure) area of the structure, Aeq close to the plotted lower limit found in
(km2). AS 1768-1991.
If these two parameters are known, then the average On perfectly flat ground with uniform properties,
strike frequency (ASF, strikes per annum) can be lightning will strike each square kilometre with equal
computed using probability. However, an object that is taller than the
surrounding area has an “equivalent attractive area”
ASF = GFD x Aeq (2)
greater than the ground area it occupies, as shown in
Figure 1. Hence, the attractive area of a structure is
The average GFD for a particular region can be
defined as the equivalent area on flat ground which
obtained over a period of time (usually years) using
would be struck with the same frequency. It depends
lightning flash counters or lightning detection and
on the structure height, the charge on the lightning
location systems such as LPATS. In many countries,
downward leader (which, in turn, is related to the
this information is not available and only the number
return stroke peak current) and to a lesser extent the
of thunderstorm days per year (TSD) is known. The
ground surface area occupied by the structure.
Aeq
Although the details are beyond the scope of this 3.2 Yield of the Dynasphere
paper, it must be noted that the linear correlation
coefficient, and the test of its significance, is based on This was calculated using all of the attractive radius
the assumption that the data set has a bivariate models (Eqns 5−10). However, the real assessment of
normal distribution. In many cases, this assumption performance can only come from the model on which
cannot be made without casting serious doubt over the the lightning protection design is based, namely that
results. “Nonparametric” methods provide a much of Eriksson [9], summarised by Eqn (7). The yields,
more robust way of assessing the degree of correlation defined as the actual strikes per annum divided by the
between two variables, much like the median is more number expected in that time, are:
robust than the mean in some cases. These methods
Yield obtained using Equation (7): 1.21
correlate the rank of the data rather than the absolute
Yield averaged across all six models: 1.07
values, hence the term “rank correlation coefficient”.
The two most common are the Spearman (rs) and
This is a remarkable result, especially if one considers
Kendall (τ) coefficients. The correlation based on
that at least two of the models are based on striking
rank results in a slight loss of information (it
distance which overestimates the attractive radius
desensitizes the correlation) but this is a small price
(hence giving a greater number of expected strikes
to pay for a major advantage - when a correlation is
and a lower value of the yield ratio). It implies that
demonstrated nonparametrically, it is really there !
the Dynasphere has captured 20 % more strikes than
expected (or 7 % more if the averaged value from all
Hence, in the analysis of the Hong Kong data, all
six models is used). In other words, the Dynasphere
three correlation coefficients were computed
has an enhanced lightning capture ability. This result
(Pearson’s linear, Spearman’s rank and Tendall’s
also indicates that the BENJI lightning protection
rank) along with their respective probabilities. Also, a
designs are relatively conservative, i.e., the attractive
series of experiments were performed in order to
radii computed by BENJI may be smaller than the
assess the effect on the correlation coefficients of: (i)
actual ones.
using building heights derived from the length of
TRIAX according to the 10, 15 and 20% rule of
Hence, it can be confidently claimed that the
thumb, (ii) using different models for the equivalent
Dynasphere successfully captures more strikes than
attractive radius, (iii) combining counts from
would be expected for the unprotected structure. This
installations that are close together, (iv) changing the
claim is supported by the fact that there have been no
(constant) GFD figure used, and (v) changing the
documented bypasses of the Dynasphere on the 185
median peak current used.
installations that were suitable for this analysis.
The only parameter that could not be included in the
3.3 Correlations
correlation tests is the specific region-to-region GFD
within Hong Kong, where geographical features can
Both the linear and rank correlation coefficients were
result in considerable departures from the average
computed along with their respective probabilities. As
strike density. This detailed information was not
mentioned earlier, the usual way of expressing the
available at the time of writing. Using a constant
latter is by stating the percentage probability that the
computed correlation coefficient is exceeded if the
two variables are uncorrelated. Hence, a small Figure 2(a) shows a BENJI lightning protection
probability such as 1 % indicates that the correlation design for a power plant in the Philippines [Fig.
is highly significant. This logic can be reversed by 2(b)]. In this design, a single Dynasphere is mounted
stating that the “probability the correlation is on an appropriate structure in the complex and its
significant is 99 %”. The latter is the manner in attractive area (16,300 m2) is such that it protects the
which the probabilities are expressed in Table 2, whole facility. The smaller circles in Fig. 2(a) show
along with the correlation coefficients. The results the attractive area of the “competing features” in the
quoted in this table were based on equivalent radii complex.
computed with Equation (7) which, as stated earlier,
is used in the BENJI computer designs.
(b)
4. APPLICATION TO POWER FACILITY
PROTECTION Figure 2: BENJI lightning protection design for a
power plant in the Visayas region of the Philippines.
Whilst the data described and analysed above do not (a) Plan view of the protection or attractive area
relate directly to power facility lightning protection, provided by the Dynasphere. (b) 3D view of the site.
exactly the same principles will apply. ERICO has a
large portfolio of BENJI designs compiled for
customers which cover a wide range of power Dynasphere installations such as the one shown in
stations, substations and control buildings housing Figure 2 are providing effective protection against
switchgear, communications and control cabling. In lightning strikes for a wide variety of power utility
the case of control buildings, the technique would be customers both in Australia and throughout Asia,
identical to that applied to the structures in Hong from hydro, geothermal and steam, to photovoltaic
Kong, i.e., an evaluation of the attractive area of the systems. Just a few of the customers that are protected
building and the substation area to be protected. In a with ERICO’s System 3000 include:
typical installation for substation protection, a
Dynasphere would be mounted on a mast of a given In Australia: Pine Creek, Standwell and Collie power
height so that its attractive area encompassed the stations; Hasper and Primrose substations in
matrix of transformers and HV transmission lines. Tasmania.
The mast is simply a very slender structure and so In Asia: Coloane (Macau), PLN steam (Indonesia),
similar principles would once again apply. Hydrolic (Taiwan), Visayas region (Philippines),
Pontianak gas turbine/Darajat geothermal/Wayang
Windu geothermal (Indonesia) power plants. [4] Mousa, A.M. & Srivastava, K.D., “The
implications of the electrogeometrical model
regarding the effect of height of structure on the
5. CONCLUSIONS median amplitude of collected lightning strokes”,
IEEE Trans. Pow. Del., Vol. 4, 1989, pp. 1450-
A statistical analysis of lightning strikes to 1460.
Dynasphere air terminal installations on structures in
Hong Kong has been carried out. The mean yield of [5] Bazelian, E.M., Gorin, B.N. & Levitov, V.I.,
the Dynasphere is effectively 100% and there is a “Physical and engineering foundations of
highly significant positive correlation between the protection against lightning”, Leningrad,
actual and expected strikes over a wide range of Gidrometeoizdat, 1978.
structure heights and locations. In fact, the
correlation is significant at a level of at least 99.7%, [6] Rakov, V.A. & Lutz, A.O., “A new technique for
and it persists even when the assumptions are put to estimating equivalent attractive radius for
the test. downward lightning flashes”, 20th ICLP,
Switzerland, 1990, Paper 2.2.
These results strongly suggest, if not conclusively
prove, that the design principle behind the [7] Petrov, N.I. & Waters, R.T., “Determination of
Dynasphere is sound and the collection volume the striking distance of lightning to earthed
concept employed for the placement of air terminals structures”, Proc. R. Soc. Lond. A, Vol. 450,
on structures is an effective tool in BENJI lightning 1995, pp. 589-601.
protection designs. This conclusion is further
strengthened if one considers that there are no [8] Golde, R.H. (editor), “Lightning”, Academic
documented cases of bypasses to the Dynasphere. Press, London, 1977.
Hence, it can be claimed with a high degree of
confidence that ERICO’s System 3000 designs offer a [9] Eriksson, A.J., “The incidence of lightning
protection level of at least I (98%) according to the strikes to power lines”, IEEE Trans. Pow. Del.,
IEC 1024-1-1 scheme. PWRD-2, 1987, pp. 859-870.
It has also been demonstrated that these basic [10] Rizk, F.A.M., “Modelling of lightning incidence
principles are easily extended (and are applicable) to to tall structures. Part II: Application”, IEEE
the protection of a wide variety of power stations, Trans. on Power Delivery, Vol. 9, 1994, pp. 172-
substations and control buildings operated by power 193.
utilities throughout Australasia.
[11] Torres, H., Rondon, D., Briceno, W. & Barreto,
Finally, to the author’s knowledge, such a study has L., “Lightning peak current estimation analysis
never been carried out by any other lightning from field measurements in tropical zones”, 23rd
protection company or manufacturer of air terminals, ICLP, Florence, 1996, pp. 181-185.
be they passive systems or other active systems. This
is another example of ERICO’s leadership in air [12] Rizk, F.A.M., “Modelling of lightning incidence
terminal research and, in particular, the assessment of to tall structures. Part I: Theory”, IEEE Trans. on
air terminal performance. Power Delivery, Vol. 9, 1994, pp. 162-171.