Contents Telektronikk - Telenor
Contents Telektronikk - Telenor
Contents Telektronikk - Telenor
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
74<br />
Si:<br />
extreme<br />
high<br />
n s<br />
n s<br />
normal<br />
s n<br />
0<br />
0 20 40 60 80 100<br />
% of circuit-groups<br />
Figure 11 Distribution (s) of the observed skewness indices (Si)<br />
compared with the skewness of the normal distribution function<br />
(n) [7]<br />
then compared with the basic variations<br />
D. D = (2tmY/T) 1/2 , where tm = average<br />
holding time, Y = measured intensity in<br />
the peak-hour, T = integration period, or<br />
the read-out period.<br />
The peak-hour’s internal skewness was<br />
studied by comparing its half-hour intensities.<br />
It was seen that only in 20 % of the<br />
circuit-groups the skewness was normal,<br />
i.e. in measures of the basic variation<br />
(Figure 11). The skewness was remarkably<br />
higher in 58 % and very remarkably<br />
higher in 22 % of the circuit-groups,<br />
when according to the Poisson-process,<br />
the corresponding percentages were 83.8,<br />
15.7 and 0.5. The variations are thus<br />
clearly higher than ideal. This can be<br />
understood so that the offered intensity is<br />
not constant but changing during the<br />
peak-hour. Thus, the congestions are not<br />
divided evenly over the full hour, but<br />
concentrated to one of the half-hours.<br />
Why dimension during the full hour<br />
when the process, relevant to the quality,<br />
happens during one half-hour only? –<br />
This alone is a reason for reconsidering<br />
the modelling.<br />
In lack of more detailed data, the peakhour’s<br />
internal distribution was studied<br />
as the quarterhours’<br />
relative<br />
range W, i.e. the<br />
difference of the<br />
highest and lowest<br />
quarter-hour<br />
intensity in relation<br />
to the basic<br />
variation. The<br />
observable range<br />
W depends both<br />
on the offered<br />
traffic’s variations,<br />
and the<br />
number of circuits<br />
if it is<br />
dimensioned to limit the variations. In<br />
order to distinguish these two reasons,<br />
the circuit-groups were classified according<br />
to their loading. The Christensen formula’s<br />
goodness factor h is here used for<br />
load classification:<br />
h = (n - Y) / Y 1/2 ,<br />
where h = the load index, n = number of<br />
circuits, Y = average of the round’s days’<br />
peak-hour intensities.<br />
Here the dimensioning is called tight,<br />
when h ≤ 1.5, but normal or loose when<br />
h ≥ 2. Tight dimensioning is used typically<br />
in overflowing high-congestion<br />
first-choice circuit-groups, whereas loose<br />
dimensioning is used in low-congestion<br />
last-choice, and in single circuit-groups.<br />
According to the mathematical tables of<br />
Poissonian distribution, probabilities<br />
10 %, 40 %, 40 %, 10 % are reached by<br />
the range of four samples in corresponding<br />
classes having W = < 1.0, 1.0 to 1.9,<br />
2.0 to 3.2, and > 3.2. The numbers of circuit-groups<br />
in the same classes are given<br />
in Table 1.<br />
It indicates that the circuit-groups in all<br />
are rather well in line with the ideal<br />
Table 1 The range in measured circuit-groups and in the Poissonian distribution,<br />
depending on the dimensioning<br />
Classes of range 3.2 total<br />
Circuit-groups 19 50 26 25 120<br />
% 6 42 22 21 100<br />
Of them tight 19 11 0 0 30<br />
% 63 37 0 0 100<br />
loose 0 39 26 25 90<br />
% 0 43 29 28 100<br />
Poissonian % 10 40 40 10 100<br />
ranges. However, this is only partly a<br />
consequence of ideal offered traffic, but<br />
caused by the combined effect of tight<br />
dimensioning and peaky traffic, namely<br />
- tight dimensioning limits the variations<br />
to narrower than the Poissonian<br />
- loose dimensioning allows the unlimited<br />
variations of the offered traffic,<br />
being generally larger than the Poissonian<br />
- role of the circuit-group – first-choice<br />
or individual or last-choice – has<br />
minor influence on the range, as was<br />
seen separately.<br />
The explanation of the large range in<br />
offered traffic is that the intensity is not<br />
constant during a period of one hour, but<br />
changes by more than the normal variation.<br />
One hour is thus too long a period<br />
for the assumption of equilibrium. – This<br />
result is thus reverse to the English<br />
school which has favoured the smooth<br />
models.<br />
When the distributions in reality differ so<br />
much from the model in common use,<br />
there are two choices:<br />
- to develop the non-Poissonian models<br />
which better describe the large variations<br />
of the hour. The overflow dimensioning<br />
methods give some experiences<br />
in this way, or<br />
- to diminish the dependability of the<br />
model by shortening the read-out<br />
period so far that the assumption of<br />
constant intensity is valid, from one<br />
hour to e.g. a quarter-hour.<br />
A similar phenomenon was observed in<br />
Helsinki at an early stage [1] when the<br />
observed and the calculated congestions<br />
were compared. The calculated congestions<br />
were too few when the read-out<br />
period was half-hour, less differing in<br />
1/8, and well fitting by 1/24 hour. The<br />
only one analysed circuit-group does not<br />
give basis for wide conclusions, but<br />
offers an interesting view for the modellers.<br />
7 The day’s peaks’<br />
concentration to the<br />
peak-hour<br />
The ElCo about the peak-hour concentration<br />
is that the day’s peaks are concentrated<br />
into the peak-hour or to its<br />
vicinity. If the ElCo about peak-hour concentration<br />
is valid, special attention must<br />
be paid to the perfect peak-hour timing.