This preview shows page 1. Sign up to view the full content.
Unformatted text preview: either in or out from
South Africa. If the power is flowing into South Africa, it is included in the
generation pattern, else it is part of the load for International Customers.
However, there are now predefined rules to take any decision whatsoever. In some cases a neighbouring country can take its total demand from the
distribution networks, or a portion thereoff, and sometimes from the
transmission network. At one of the transmission substations (275/132 kV), a
neighbouring country has the option to take supply from the 275 kV or 110 kV 42 (through a 132/110 kV transformer). To complicate matters further, it is also
possible that the 275 kV can supply into South Africa.
Therefore, for the purpose of the balancing algorithm, these loads are treated
separately. The agreements between the utility and the neighbouring countries are studied and the best judgement is made to allocate the loads to
the transmission network. The expected loads are treated similarly.
The total area loads are approximately 93 percent of the transmission system
load. Each area load is the sum of a number of transmission substation loads
that are connected directly to the distribution networks. The sum of the customer loads in the area can be either less or more than the area load
because of import or export of loads through the distribution networks. The
area load is thus the total load that flows from the transmission network into
the distribution networks in that area.
The area loads are a further breakdown of the transmission system load and
therefore the same factors should be considered, but more on a geographical
basis. Furthermore, an area load is also a benchmark for the transmission
and distribution loads in that area. The accuracy of the area loads is only
important within the first five years. Thereafter it only provides information on
how much the area load will be (see the section on the distribution
Any forecast has a degree of uncertainty and obviously the longer the forecast
horiz on and the larger the variances, the larger the degree of uncertainty. It is
quite a challenge to produce a twenty-year area forecast and also give a
reasonable and acceptable maximum transmission system load. This is the
dilemma for producing area forecasts that will stand the test of time. In Chapter 1 it is mentioned that the objective of this thesis is to give more
insight into the electrical industry and therefore reduce the uncertainty in the
future expected loads. The forecasts have to be more accurate and should 43 include more information. The reason is to meet three key goals: retain customers, enhance revenue and reduce costs. 
A number of forecasting articles (mainly short-term) have been studied. The
articles however, confirm that tempera ture, economical cycles and long-term
growth trends are important to consider. [18 - 27] Referring to Annexure A, it is not possible to use historical trends in the area
data and make predictions of future load growths. The different forecasting
techniques evaluated for the expected area loads are discussed next. Decomposition Method
The decomposition method, although more often classified as a short-term
forecasting technique, has been evaluated to predict the area loads (long-term
forecasting horizon). The basic concept of the decomposition method is firstly
to remove the trend (T), then the cycle (C), and finally the seasonal
component (S). Any residual (I) is assumed to be random and cannot be
predicted. ACTUAL = T × C × S × I (3.2.5) The trend represents the long-running behaviour of the data and can increase,
decrease or remain stationary. The cyclical factor represents the ups and
downs of the economy. The seasonal factor relates to the periodic fluctuations of constant length that are caused by temperature, rainfall, climate
variations, etc. The factor TC is determined by calculating the centred moving averages of
the actual data. The trend of TC (that is T) is determined next by linear regression. C is determined by dividing TC by T. If the data is monthly totals,
then the period length is twelve and each of the twelve Ss is calculated by the
average of SI (ACTUAL/TC) for that given month.
determined by dividing SI with S. The random factor is 44 No demographic, economic or sector data is available on national level to
develop area load models. 
Decision analysis helps the forecaster to systematically structure any possible
demand growths in an area. This provides a basis for a thorough understanding of the key issues for future demand growths. Incorporation of
subjective judgements is an important aspect of decision analysis.
Decision analysis has the following three steps: 1) decisions to make 2)
uncertain events and 3) the value of specific outcomes. The relevance trees,
cross -impact matrix and surveillance methods assist the decision-maker to
understand the problem, identify the uncertainties and determine the best
outcome. The scenario in Figure 3.2.1 provides a useful structure to work
View Full Document
This document was uploaded on 03/04/2014.
- Spring '14