1989

MODEL FOR LOAD SIMULATIONS BY MEANS OF LOAD PATTERN CURVES

A Andersson A Jansson J Klevgs

Gothenburg Energy Association of South Swedish

Supply Utility, S Swedish Electric Power Company, S
Utilities, S

SUMMARY power demands vary over the day and year.

Also embraced by the project is a survey of
A growing need has been felt in Sweden for how different factors, such as tariffs and
better knowledge of the loads in the power temperature, affect the nature of the demand
distribution networks and for better load curves. The statistical demand curves will
information processing methods. primarily be used in connection with the
The Association of Swedish Electric Utili- planning of distribution networks but they
ties has therefore initiated a nation-wide can also be used for planning of shutdowns
load data measurement programme and research and tariff studies.
project aimed at developing a new load
information processing method.
2. DATA COLLECTION
On the basis of the measurements,
statistical demand curves have now been The measuring equipment was installed adja-
produced for several different customer cent to the electricity meters of the custo-
categories. To facilitate the use of these mers. The average power demand of each
calculated demand curves, a forecast program customer was recorded every 15 minutes for
has been developed. The demand curves and at least one year. In total , about 160
associated forecast program are at present items of measuring equipment were installed.
available for VAX computers and IBM PC:s. A comparatively simple and inexpensive
The objective has been to make the forecast equipment was chosen consisting of an elec-
program as easy as possible to use. The only tronic unit and an optical sensor built into
data that has to be entered is information an electricity meter. The results of the
which is normally available within the measurments are stored in a replaceable
electricity supply utility. EPROM memory which is emptied once a month.
The measuring equipment is shown in figure
1.
1. INTRODUCTION
The measurements are carried out by some 20
For the planning and operation of distribu- Swedish electricity supply utilities located
tion networks, knowledge of the power throughout the country from north to south.
demands of customers and how this demands The measuring equipment was installed on the
are diversified in the network is required. premises of different customer categories
A common procedure in Sweden is to use such as private houses and row houses with
annual energy consumption as a basis and different types of heating, weekend
convert it into power demand with the aid of cottages, farms with different types of
mathematical methods. production, blocks of flats, schools, shops,
offices and industrial manufacturing plants.
Mainly used is Velander's equation:
The power supply utilities collect the
P = c1w + c2fi (1) results of the measurements and information
about the customers. After local processing,
where the measurement and customer data are loaded
into a central computer where the analyses
P = Maximum power demand in kW are performed. The outdoor temperature at
W = annual energy consumption in kwh the localities where the measurements took
Cl, C2 = calculated constants which are place is also entered into the central
dependent on the consumer computer. The organization of the data
category processing is shown in figure 2.
This equation assumes that the maximum
demands of the partial loads show a normal 3. ANALYSIS MODEL
distribution. The equation provides also no
information as to when the summed maximum The purpose of the work of analysis is to
demand occurs. This makes it unsuitable in produce statistical demand curves for a
connection with the summing of power demands forecast model. The statistical demand curve
of a different nature over the day or year. consists of a normalized average curve and a
With the changes in the load characteristics curve which describes the dispersion round
of customers that are taking place in the average.
Sweden, f o r example due to the use of heat
pumps and the affection of power demands by Basic data for the analysis are the recorded
introduction of tariffs with different power demand curves of each individual con-
energy prices at different times of the day, sumer in combination with registrated out-
Velander's equation is growing increasingly door temperature and geographical location.
unsuitable.
In the work of analysis it has been assumed
According to this, the Association of that the demand curves are of the same shape
Swedish Electric Utilities initiated a in different parts of Sweden when the out-
national measurement programme and research door temperature is the same. External
project for the purpose of obtaining stati- conditions are described partly by a sub-
stical demand curves which describe how division into different seasons of the year
545
and partly through dependence on outdoor Annual energy at the actual locality
temperature. The outdoor temperature affects is converted into the level that
the shape and level of the demand curves applies to the standard locality.
while the annual energy only affects the
level. Ea,std Ea,act* (l-P+
P*Gsta/Gact) ( 3 )
The analyses are carried out in five stages where
with the aim of obtaining statistical demand EaIstd Annual energy at
=
curves for different customer categories. standard locality
Each category is dealt with separately when Ea,act = Annual energy at
the data is processed.
actual locality
The recorded daily demand curves are cp = Proportion of the
grouped according to season of the actual annual energy
year (winter, spring/autumn and which is
summer), type of weekday (workday and temperature-dependent
weekend) and outdoor temperature Gstd = Day-degree value, standard
(daily temperature at ten intervals). locality
Gact = Day-degree value, actual
Calculation of the sensitivity of de- locality
mand to outdoor temperature.
The annual energy, corrected for stan-
Calculation of the monthly energies, dard locality, is converted into
corrected for-standard year and seasonal average demand.
standard location, in respect of each
customer. The monthly energy figures
are subsequently converted into
seasonal average demands. where

Estd = Erec + Psp*(Grec-Gstd)*24 (2) Pseason = Seasonal average demand at

standard locality
Erec = recorded energy consumption per Qlseason = Seasonal proportion of
month annual energy
nday = number of days in the
Estd = energy consumption corrected season concerned
for standard month
The daily demand curve for each cate-
Psp = sensitivity of demand to gory is calculated by means of the
temperature statistical demand curve for the selec-
ted month, type of day and tempera-
Grec = recorded day-degree value per ture. The average demand curve is
month calculated with the following equa-
tion:
Gstd = standard day-degree value per
month
Normalization of the recorded daily i=l
demand curves of each customer with where
seasonal average demand corrected for
standard locality. n - number of customers in the
Calculation of the normalized average PtYP
- category
normalized average curve
curve and standard deviation per cate- for the category
gory, season of year, type of weekday
and temperature interval. A total of Pseason seasonal average demand
about 18 average curves are calculated i for a customer
per category. Example of normalized t - quarter-hour or hour
average curves for a winter weekday is concerned during 24-hour
shown in figure 3 . period
The standard deviation is calculated
4. FORECAST MODEL with the equation:
The forecast model makes it possible to fore-
cast daily demand curves for an arbitrarily Scat(t) = Styp(t) * / i=1
S ( 6 )
selected group of customers. It is assumed
that information are available in regard to
the composition of the group of customers, where Styp = standard deviation round
annual energy demand and the day-degree the average curve of the
value applicable to the locality. The user category
can freely select external conditions, i.e.
the outdoor temperature and the month or The average demand and standard devia-
season of the year. tion curves for the various categories
are merged to make a total demand
Forecasts are first made for each category curve in accordance with the following
and these forecasts are then merged to make equations:
a total forecast. The procedure is as
follows: Ptot(t) = Pcatl(t) +..+ Pcatn(t) (7)

546
 On the basis of the merged average Computer programs are presently available in
demand curve and its standard devia- standard FORTRAN for VAX computers and IBM
tion curve, it is possible to calcu- PC: s
late for every time segment a demand
which, with a certain degree of pro- The input data required in the present
bability, will not be exceeded. version is:
This is done as follows:
the standard day-degree value for
Pmax ( $ I t ) = Ptot(t) + a * Stot(t) ( 9 ) the locality
month
where $ = the desired degree of weekday/weekend
probability outdoor temperature
a = a factor depending on$ probability limit
category
+ a annual energy or size of the
(8) building
number of customers
04 1.00
90 1.28 Input data are given for every customer or
95 1.64 group o f ' customers who are to be included in
99 2.33 the calculation.
99.9 3.09
Output data are produced in the form of a
average curve and a curve which are not
5 USING THE FORECAST MODEL exceeded by the specified degree of
probability.
The aim is that the user can obtain output
data which can be used for thermal and In the example in figure 4 are shown a
economic dimensioning, planning of shutdowns number of forecasted daily demand curves
and long term forecasts. which, with a probability of 9 9 % , will not
be exceeded. The following external
Dimensioninq conditions apply to the example:
For thermal dimensioning, details are locality = Stockholm
required of the highest demand value that is day = weekday in January
attained with a certain degree of proba- outdoor temperature = -15'C
bility. Normally, this is the highest value time frame = 1 hour
on a winter weekday. However, certain types
of loads may reach a demand peak at some
other time.
REFERENCES
The thermal dimensioning limit is of parti-
cular interest when a choice has to be made 1. The Association of Swedich Electric
between reinforcing the existing network or utilities, 1 9 8 7 , "Load simulations by
not reinforcing it. means of load pattern curves", 35
pages + 9 appendixes (in Swedish).
At valuation of losses in economic dimensio-
ning, the most probable demand value and its 2. Larsson, L., 1 9 8 2 , "Computer model for
duration curve should be taken as starting simulation of power demands in private
point. houses with electric heaters", LUTMDN,
TMVK-3105, 6 4 pages (in Swedish).
Forecasts
3. Kalevi, J . , Karkkainen, S. and Juuti,
In connection with rebuilding and P. , 1 9 8 7 , "Load curve simulation
maintenance work, it is often of interest to experiences in network calculations
be able to decide whether a demand can be and load management in Finland", CIRED
supplied via an alternative feeder path for report 1 9 8 7 .
a given period. It may be a matter of weeks
during a low-load season or hours in the
night during a weekend or public holiday. To
see whether such alternative feeding is
possible or not, forecasts of the maximum
power demand that will not be exceeded du-
ring the period with a certain degree of
probability are required.
Forecasts are also needed for long-term
assessments of the consequences of, for
example, the introduction of tariffs with
varying energy prices at different times of
the day or in the case of transition from
one form of heating to another.
Comwuter Droarams
The method of determining the total load by
summing the partial load curves requires
access to a large quantity of tabular data.
The statistical calculations are also compara-
tively extensive. Processing by computer is
therefore suitable.

547
4.

-ao
- 5 O

- 09

1. Electronic device

2. EPROM-memory

3. Electricity meter with

integral optical sensor
and power unit
4. Connecting lead

Figure 1 Measuring equipment Figure 3 Example of normalized average

curves for a winter weekday

CUSTOMER MEASUREMENT
DATA DATA

-T DATA COLLECTION

LOCAL PROCESSING

CENTRAL PROCESSING

DATA
PROCES- * DATA - ANALYSIS
SING BASE PROGRAM

Figure 2 Organization of the data processing

548
 -
5 customers who
combine oil and
electricity f o r

~tLJ
heating and have
cheap electricity
20 rates at night.
Annual electric
10 energy consump-
*’ ’
tion = 15000 kWh

*’
/’ 3 6 9 12 15 18 21 24 h
I
per customer.

I
I

,j
10
I

I
...
...
..-
3 6 9 12 15 18 21 24 h

40F
30
5 customers using
electric heaters.

’‘ 10 t Annual electric
energy consump-
tion = 25000 kwh
per customer
60 .’ 0‘
v
3 6 9 12 15 18 21 24h

I
40 I

lL++-
10

;.“i
I.
3 6 9 12 15 18 21 24 h

60

30

5 customers using

2o t oil-fired furnaces.
Annual electric
lo LczzzZ- 3 6 9 12 15 18 21 24 h
energy consump-
tion = 5000 kwh
per customer
I
I
I
i I

Figure 4 Example of forecasted and diversified daily demand curves

549