【15】Typical load shapes for Swedish schools and hotels_图文

ELSEVIER

Energy and Buildings 28 ( 1998) 145-157

Typical load shapes for Swedish schools and hotels

C. Nor&, J. Pyrko *

Lund Institute of Technology, Dept. of Hear and Power Engineering, SE-221 00, Lund, Sweden

Received 3 February 1997

Abstract

In

this study, typical load shapes for two categories of Swedish commercial and public buildings-schools and hotels-are presented and

discussed. The measurements

from 13 schools and nine hotels in the southern pti of Sweden were andysed Load shqzes were devebpi

for different mean daily outdoor temperatures and different day-types-standard weekdays

and standard

weekends. The load shapes are

presented as non-&mensional normalilse& l-h bact. The typical loa& shapes give a reasonable approximation of the measured SoaCt shapes,

although the relative errors exceed 20% of the mean values during some hours. Daytime (-) results are generally good with

errors of about 10%. Absolute errors remain relatively constant during the year, but as mean values decrease, the relative errors increase,

causing relative errors up to 30% during some time periods.

0 1998 Elsevier Science S.A. All rights reserved.

Keywords:

Hotels; Load shape; Sweden

1. Introduction

Electricity consumption in commercial and public build-

ings is the fastest growing end-use sector in many of the IEA

countries with an annual growth rate of 4-5% [ 11. Several

factors influence the high growth rate, such as:

-increased use of office equipment,

-a growing need of air conditioning and other comfort

equipment.

The fast technical development has made it possible to

reduce electricity consumption by using more efficient equip-

ment, but in many buildings, there is a potential of reducing

the electricity consumption and the electricity costs.

The last Swedish extensive study on commercial buildings

is the study made by the Swedish Association of the Electric

Utilities in 1987-1990 [ 21. Normalised load shapes are pre-

sented for approximately 40 different caiegories of buildings

zu-ib for tifiiererem ~~~~r~~rn-~~~re’~e~~~~.“~~~~ iire one

CD?

tie. car~o~~~~~ha~be.~nv~~~~~e~in~~~

-SmQ~.,bi

the results are not generalisable and applicable in practice

because different types of school buildings cannot represent

a” ki!imcJ~~~ cz&?@a+y L”& plieriT §h4es .4 Fhsd+&%

public and commercial buildings are very limited. For exam-

ple, the Swedish Vattenfall’s study on commercial and public

* Corresponding author. Tel.: + 46 46 2229269/80; fax: + 46 46 22247 17;

re-miix jurek.@k& er

buildings in Sweden [ 31 only presents annual electricity con-

sumption for different end-uses. On the other hand, the con-

tacts with users, distributors and producers of electricity has

indicated a great interest and need for load studies of com-

mercial buildings.

Norwegian EFI has carried out several studies on load

patterns in commercial buildings, but these studies focus

almost entirely on buildings with electrical space heating and

the results are not comparable to the results from this study

In the USA, many studies have been performed by various

companies and organisations, Lawrence Berkeley Laboratory

(LBL) has published several load related documents and

reports. The load shapes presented in the LBL reports are

often separated into end-uses like cooling, ventilation, light-

ing, etc., and usually presented as W/m2 [7-131.

The

object of

this

study

is to

develop

typical

load shapes

5% *2w,%*c&&snf, %&&‘, %ci!xwl,s -ii%Ll3Rkd,s.

Bdi~fisfk~~~‘~~v&~< ~&S?xw3issxws~~ks~

variations are discussed. The outdoor dry-bulb temperature

is assumed to be representative for the entire area where the

5?i%s~~~J~~i~~.N~~~s ‘G~k~%~~!~~

the outdoor temperature, are considered to affect the electric-

ity consumption.

In the beginning of the study, the school category consisted

of23 objects, all-iocated-in the south ofSwed&. But only 13

~5 Vms~ c&d bt c!msm TV a homqgmtms gmip ~5 di+sic~

[4-W.

0378-7788/98/$ - see front matter 0 1998 Elsevier Science S.A. All rights reserved.

PUSO378-7788(98)00011-5

146

C. Nor&, J. Pyrko /Energy and

Buildings 28 (1998) 145-157

heated buildings. This group was finally divided into two

sub-groups since substantial differences were observed

between the sub-groups:

-six schools with district heating and kitchens,

-seven schools with district heating and without kitchens.

The hotel category consists of nine hotels located in the

south of Sweden. The heating systems of the different hotels

are listed below:

-six hotels with district heating,

-one hotel with electrical space heating,

-one hotel with district heating and electricity,

-one hotel with oil furnace and electricity.

A question was raised whether the three hotels with elec-

trical heating should be included in the study. As the large

part of the electrical space heating is not connected to the

same metering unit as the rest of the hotel, this caused no

problems.

2. Methodology

The methodology is briefly described by Noren [ 141.

Some of the most important steps are discussed below.

To compare the load data for different objects, it is nec-

essary to normalise the data by dividing every measuredvalue

with the object’s mean load.

When using the basic statistical equations, an important

question is whether load data are normally distributed. This

is not always the case, but the data material is approximately

normally distributed when the number of observations is large

[ 151. This is discussed in a Norwegian study, where the load

data was considered to be approximately normally distributed

[ 161. However, some skewness of the load distribution was

observed during most hours in this study, but this is very

common due to a few outlying values, to which the skewness

is highly sensitive [ 151. Some few outlying values can make

the standard deviations high and the two-sided confidence

interval gives an incorrect picture of the deviations. Because

of that, it is important to be careful in the cases where it is

one or two deviating objects that caused high standard devi-

ations. Although the standard deviations are high, the spread-

ing might not be as large as indicated by the standard

deviations.

2.1. Load shapes-calculation

of

mean normalised load

cc,*(t)

The normalised load Ci(t) at time t for object

i

can be

calculated as:

ci(t)= y ;

1

where Ci( t) = normalised load at time t for object

i ( - );

Pi(t) = measured load at time t for object

i

(kW) ; pi = mean

annual load for object

i

(kW h/h).

The data are split into different groups, depending on day-

type. The data in every group are sorted by hour, and every

hour is sorted into different temperature intervals. Six inter-

vals for mean daily outdoor dry-bulb temperature are used to

sort the data:

> 20°C.

Now, the mean value, C,,,(t) , can be calculated for every

hour and each temperature interval [ 51:

tciCt)

C,,,(t)=

=-y

(2)

where C,,,(t) = mean normalised load at time

t

for a category

at specified temperature interval ( - ) ; Ci(

t)

= normalised

load at time

t

for object

i

in the category at specified temp.

interval ( - ) ; N= number of observations for time

t

for a

category at specified temperature interval ( - ) .

Standard deviations are calculated as [ 51:

i (Ccat(t>-Ci(t))2

(+cat(t)=

i=

I

(3)

J

(N-1)

where a,,,(t) = standard deviation at time

t

for a category at

specified temperature interval ( - ) .

Eqs. (2) and (3) are repeated for:

0 all temperature intervals

0 all 24 h of the day

0 weekdays and weekends

0 all categories.

2.2. Disadvantages

This method is very sensitive to outlying values and objects

that deviates from the other objects in the category. If one or

two objects out of IO-15 do not fit the category, these objects

will cause high standard deviations. The same problems occur

if there is a high number of erroneous data and for this reason,

it is necessary to exclude the erroneous data, otherwise, the

final results will be affected in a very high degree.

2.3. SRemoval

of

erroneous data

The load data material was plotted for each hour and ‘bad-

quality’ data were only removed when it was absolutely cer-

tain that the measurement errors were the cause, for example,

if a building with a typical daytime load of 100-125 kW h/

h, during a few hours only used O-3 kW h/h, there was no

doubt that the measurement errors were the reason. The data

material quality was very good overall, and less than 5%0 of

the data material was removed for certain objects. Such high

rates of ‘bad-quality’ data were fortunately very rare.

C. Nor&, J. Pyko / Energy and Buildings 28 (1998) 145-157

147

This category consists of seven objects with floor area

ranging from 6700 mz up to 10,&O m’, and all of the objects

are located in the city area.

3.1. YearproJiles

Fig. 1 shows the year profile for schools without kitchens.

Summer holidays are easily distinguished by the drop in the

consumption profile, which is also noticed during Christmas

and the February holidays around day 50. During April and

May, there are a number of drops. First at Easter, and then

during all the holidays in May. During autumn, there are only

two noticeable drops in the consumption profile, both corre-

sponding to school holidays. Peak days are found during the

colder part of the year, usually between November and begin-

ning of February, but no specific day-type could be identified.

Standard deviations are quite high compared to some of

the other categories, indicating a high degree of variation in

consumption between different schools. This is a distinctive

mark for the entire school category.

The activity level is very different, not only at different

schools, but the same school can have large variations from

day to day. Even during the same period of the year, and for

the same temperature, which is further discussed later. Abso-

lute standard deviations are relatively constant during the year

but due to the lower demand during holiday and weekend

periods, the relative standard deviations increase during these

time periods. Some schools are open with several holiday and

weekend activities, while others are closed, and the highest

differences are found during the summer. As shown in Fig.

2, the weekend consumption varies quite a lot during the year,

with the lowest consumption during the summer.

s

8 0.2

0

z 5 6 5 5 5 f f 5 ,$ 5 6

Dnl

Fig. 1. Year profile for district heated schools without kitchen, weekdays.

r, 6 z z4

yg 5

f f

5

5 5 G

orv

Fig. 2. Year profile for district heated schools without kitchen, weekends.

2.5

2

1.5

1

0.5

OJ

I I

!

I

t 2 3 4 6 6 7 6 6 10 11 12 13 14 15 16

17 10 19 20 21 22 23 24

now

Fig. 3. Standard weekday load shape for district heated schools without

kitchen, temperatures below OT, mean temperature - 2.3”C.

3.2. Daily load shapes

Fig. 3 shows a daily load shape with associated standard

deviations for schools with district heating and no kitchens.

Between and , the demand is almost constant,

but after , it begins to increase. At , the demand

has stabilised, and relative standard deviations are now below

10%. Typical daytime (-) standard deviations

in this temperature interval are less than 10% of mean values.

After , the demand is decreasing, while absolute

standard deviations increase, causing 20-30% relative devi-

ations of the mean values during the evening. This may be

explained by the fact that some schools in the study are using

the sports centre until very late in the evening, while others

close earlier. It is not until that the demand is at night-

time levels.

As Fig. 4 shows, the daytime demand is 15% lower in the

highest temperature interval. But absolute standard devia-

tions are at the same level as during low temperature periods,

and the relative standard deviations are l&15% in the middle

of the day. One factor may strongly influence the standard

deviations: two schools with partial electrical heating have a

lower normalised load at high temperatures, compared to the

other five schools.

Unfortunately, this affects the load shape in an adverse

manner: causing high standard deviations and large individ-

ual deviations from the typical load shape. A single school

can, however, have standard deviations of the same magni-

tude (20% of the mean value) for the same hour and tem-

perature interval. During the night and early morning hours,

there is no difference between low and high temperatures.

Differences occur after , which is shown in Fig. 5.

_-~~.

1 2 3 4 5 6 7 6 8 10 II 12 13 14 15 16 17 16 16 20 21 22 23 24

HOW

Fig. 4. Standard weekday load shape for district heated schools without

kitchen, temperatures above lS”C, mean temperature 18.6”C.

148

C. North, J. Pyrko/Energy and Buildings 28 (1998) 145-157

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 16 19 20 21 22 23 24

HOW

Fig. 5. Standard weekday load shapes for district heated schools without kitchen for five temperature intervals.

There might exist a season dependency also, during spring

and parts of autumn, when less indoor lighting is required,

and the consumption is therefore reduced.

No weekend load shapes are presented for this category.

Standard deviations are very high, and no conclusions can be

drawn, except that the activity levels vary greatly during

weekends.

3.3. Relationship between outdoor temperature and

electricity consumption

Fig. 5 shows mean daily load shapes for five temperature

intervals. Before , the demand does not depend on

temperature, but after until late at night, the demand

is lower at high temperatures, and some possible reasons for

this have already been discussed.

To determine the relationship between electricity con-

sumption and outdoor temperature, simple linear regression

with daily mean load as response variable, and outdoor dry-

bulb temperature as independent variable, was applied. Other

studies suggest that regression analysis with daily data gives

much better results than with hourly data, since natural vari-

ations tend to be more narrow when using daily data [ 2,7 1.

The regression results are shown in Tables 1 and 2.

During weekdays, the correlation is significant in five of

the seven schools. The difference between the intercepts, PO,

is very obvious, ranging from 1.28 to 1.54-a difference of

about 20%. The differences in temperature dependence are

also quite large, approximately four times (school 4 and 7).

Schools number 4 and 6 use some resistive electrical space

heating.

During weekends, five of the schools show significant cor-

relation between outdoor temperature and electricity con-

sumption. Again, there are large differences of the intercept.

Notice that school 6 has very poor regression statistics, and

how school 2 now shows a significant correlation between

electricity consumption and outdoor temperature.

3.4. Concluding remarks

There are several parameters influencing the load shape

for school buildings. The following parameters have been

considered especially important: (a) type of school: second-

ary schools tend to end later in the afternoon and start earlier

in the morning than primary schools. (b) Operational strat-

egy: parts of the ventilation system is in operation during

night-time for different reasons in some schools, while other

schools shut-off most of the ventilation system during night-

time. This is an important reason why the night-time loads

are different in schools. The night-time ventilation operation

is more common in schools with new ventilation systems.

Table 1

Standard weekday

kitchen

School

1

2

3

4

5

6

7

PO(-)

1.28

1.38

1.48

1.54

1.29

1.32

1.35

regression results for district heated schools without

psp ( -/“C)

- 0.002

0.000

-0.018

- 0.033

-0.011

-0.014

- 0.008

Adj. R’

0.00

0.00

0.71

0.19

0.29

0.64

0.32

Significance

0.013

0.832

0.000

0.000

0.000

0.000

0.000

Table 2

Standard weekend

kitchen

School

1

2

3

4

5

6

7

PO(-)

0.56

0.57

0.62

0.69

0.66

0.48

0.48

regression results for district heated schools without

p,, ( - 1°C)

0.010

0.015

-0.015

- 0.008

- 0.004

- 0.002

0.001

Adj. R2

0.31

0.24

0.65

0.37

0.08

0.01

0.02

Significance

0.000

0.000

0.000

0.000

0.001

0.463

0.080

C. Nor&. J. Pyrko / Energy and Buildings 28 (I 998) 145-157 149

(c) There are major differences during the evening hours

depending on whether the school has some evening activities.

If it is equipped with a sports centre, it is not certain that it is

used in the evening all days. (d) A few schools (typically

secondary schools) have some vocational training activities,

which causes high electrical demand when the machinery are

used. Although the total effects are small, it is a reason for

day-to-day variations. (e) Schools with some electrical heat-

ing (resistive space heating, electrical pre-heaters in the

ventilation system, etc.) have a lower normalised

load

during

the warmer periods of the year when compared to schools

without electrical heating. (f) All schools are not efficiently

operated. It has not been investigated whether the presented

study only has included schools without any operational prob-

lems. Probably, it is an important reason for school-to-school

variations.

The school category is unfortunately characterised by

rather high standard deviations, especially at outdoor tem-

peratures

above

10°C. At temperatures below 10°C the rel-

ative standard deviations are less than 10% between

and , but at higher temperatures and during the evening,

standard deviations can be as high as 30% which mainly

depends on a lower demand. Absolute standard deviations

are less affected. The best way to solve this problem is to

have more load data to be able to divide the school category

into sub-categories like:

0 objects without or with sports centres,

0 objects

with only one type of heating system.

It is important to remember that one single school can

deviate from these typical load shapes, especially if it differs

from the factors joining the seven objects included in this

category:

l

no food cooking,

l

sports centre used in the evening as well,

0 district heating as the main heating system.

4. Results for schools with kitchens

This category consists of six objects, and all are located in

the city area. It is complicated to find representative objects

for this category, since it is quite common today that schools

which cook food have a separate metering unit that does not

collect hourly load data for the kitchen. The floor area ranges

from 9000 m2 to 22,200 m2, and all six objects are district

heated.

The kitchen size varies, from the smaller ones that only

cook a couple of hundred portions per day, to the biggest one

that cooks around 5000 portions per day. The load shapes for

this category are only valid for schools with rather large

kitchens.

4.1. Year projiles

Figs. 6 and 7 show the year profiles for this category. The

difference between this weekday year profile, and the year

profile in Fig. 1, is the slightly higher normalised consump-

Fig. 6. Year profile for district heated schools with kitchen, weekdays.

Fig. 7. Year profile for district heated schools with kitchen, weekends.

observation since schools that cook

weekdays com-

without kitchens. The lower consumption

during holidays, is also a bit more noticeable for this category.

Absolute standard deviations are of the same size as for

schools without kitchens, even slightly lower from time to

time.

If Figs. 7 and 2 are compared, the difference is very small.

This is an expected observation since none of the kitchens

normally are in operation during weekends. Again, the rela-

tive standard deviations are higher than during weekdays,

depending on the more varying activity levels during the

weekends.

4.2. Daily loud shapes

As Fig. 8 shows, the demand in the middle of the day is

approximately 20% higher than in schools with no kitchens.

A distinctive mark for these schools is the very high standard

deviations between and because some of schools

in the study begin cooking very early and the different schools

cook different number of portions/day. It is the one that cooks

35

(-1

3

1

I

2.5

P *

f “1

0.5

0

I I

I

I

1 2 3 4 5 6 7 6 9 10 11 12

13 14 I5 16 17 16 19 20 21 22 23 24

“mu

Fig. 8. Standard weekday load shape for district heated schools with kitchen,

outdoor temperature below 0°C. Mean temperature - 2.1”C.

tion. This is an expected

food use more equipment during standard

pared to schools

150

C. Nor&, J. Pyrko /Energy and Buildings 28 (1998) 145-157

2.5

0,

1 2 3 4 5 6 7 8 9 10 11 12 13

TltM

14 15 16 17 18 19 20 21 22 23 24

Fig. 9. Standard weekday load shape for district heated schools with kitchen, outdoor temperature above 15°C. Mean

temperature -

18.3”C.

I

0

1 2 3 4 5 6 7 8 9 10 11

I

12 13 14 15 16 17 18 19 20 21 22

I

23 24

Fig. 10 Standard weekday load shapes for district heated schools with kitchen for five temperature intervals.

5000 portions/day that affects the load shape mostly during

these 2 h, since cooking begins already at in that school.

Absolute errors are higher compared to schools without

kitchens during daytime and the demand variations from day-

to-day are higher for these schools. The major difference,

compared to schools without kitchens, is the higher normal-

ised load between and , and the lower normalised

load between and On single days, the devia-

tions from the typical load shape can be very large since the

schools in the study showed a very varying demand, even

during the same hour and temperature interval. Measure-

ments in one of the school kitchens during 2 weeks in spring

1996 [ 171, showed that the kitchen demand varied very much

depending on the kind of food that was served. On days when

simple courses like soup was served, the demand was much

lower compared to when more advanced courses were served,

like fried fish with boiled potatoes.

Fig. 9 shows the daily load shape for the highest temper-

ature interval. The maximum demand has decreased by 15%,

but standard deviations are of the same size as at the low

temperature interval.

No weekend load shapes for this category are presented

here. All the shapes can be found in Ref. [ 141. The differ-

ences between schools with and without kitchens are negli-

gible on weekends.

4.3. Relationship between outdoor temperature and

electricity consumption

Fig. 10 shows daily load shapes at different temperature

intervals. Except for the higher demand in the middle of the

day, the load shapes are quite similar to the ones in Fig. 5:

hourly loads after are most affected by temperature.

The regression results are presented in Tables 3 and 4. All

holidays and days with a mean outdoor temperature above

17°C were excluded.

As shown in Table 3, the R2-values are low for four of the

schools, but the correlation is significant in five schools. The

temperature dependence varies very much in the different

schools, almost six times between school 5 and 6.

Table 3

Standard weekday

School

1

2

3

4

5

6

regression results for district heated schools with kitchens

Psp(-IT)

- 0.011

-0.012

- 0.016

- 0.022

- 0.028

- 0.005

Adj. p

0.28

0.36

0.31

0.61

0.58

0.05

Significance

0.000

0.000

0.000

0.000

0.000

0.002

PO(-)

1.41

1.45

1.52

1.54

1.58

1.41

C. Nor&, J. Pyrko /Energy

and

Buildings 28 (1998) 145-157

151

Table 4

Standard weekend regression results for district heated schools with kitchens

School

P”(F) psp ( - /“Cl

Adj. R2 Significance

1

0.68 -0.008 0.23 0.000

2 0.58 -0.007 0.17 0.000

3 0.54 -0.002 0.02 0.089

4 0.7

1

-0.012

0.34 0.000

5 0.61 -0.014 0.45 0.000

6 0.49 0.008 0.15 0.000

Table 4 shows that temperature dependency generally is

lower at weekends, and R2-values are now low for all the

objects, although the correlation is significant for five of

these. Notice how school 6 now has a significant positive

correlation between electricity consumption and temperature.

Because of the great variations in temperature dependence,

no general rule predicting how school buildings respond to

temperature changes can be determined.

4.4. Concluding remarks

Just like the other school categories, this category is char-

acterised by high standard deviations. The maximum demand

is approximately 20% higher in schools with kitchens during

daytime, but after , the demand is lower when compared

to schools without kitchens. Very large variations can be

observed from day to day, even in the same school. There are

several reasons for the deviations and some have already been

discussed, but the most important reason for the day-to-day

demand variations in schools with kitchens is the fact that

different kind of food requires a different amount of the

kitchen capacity and different schools cook different number

of portions per day.

5. Verification

To verify the typical load shapes, measured data from 1995

was used. Data from 14 schools that was not used to develop

the load shapes were used to verify the load shapes for schools

without kitchens. The measured total demand for the 14

schools was compared to the total model demand. Four week-

days were randomly chosen at different outdoor temperature

levels.

Fig. 11 shows the typical load shape compared to the meas-

ured one. Although none of the objects really belongs to that

category, the typical load shape matches the measured load

shape very well. During some hours, errors exceed 10% but

are below 10% most of the hours. The high errors in the

evening is due to the sports centre; only some of the 14

schools are open in the evening. Several of the previously

discussed factors can influence the errors.

Fig. 12 shows the typical load shape and measured load

shape. Errors are quite high during most of the day and, when

trying to apply the typical load shape in this temperature

; 2 3 4 5 8 7 n 9 10 ,, 12

13

I4 15 16 17 18 19 m 21 22 23 24

HOW

Fig. 11. Total demand from 14 schools (thick line) not included in the

‘schools without kitchen’ category, compared to the typical load shape (thin

line with boxes) of that category. Thursday 951214, mean daily outdoor

temperature - 3.0%

1 1 2 2 3 3 4 4 n n n n 7 7 n n 9 9 10 II II 12 13 1. 15 16 17 18 19 x) 21 P 23 24

mu

Fig. 12. Total demand from 14 schools (thick line) not included in the

‘schools without kitchen’ category, compared to the typical load shape (thin

line with boxes) of that category. Friday 95 1110, mean daily outdoor tem-

perature 7.3”C.

interval for other days, the results were identical: quite high

errors during most of the day.

Fig. 13 shows a comparison between the typical load shape

and a measured load shape from one of the hottest days during

autumn 1995. Errors are well below lo%, except between 4

p.m. and , when the operating hours of the sports

centre have great influence on the load shape.

Fig. 14 shows a measured load shape compared to the

typical load shape during a day when, for some reason, some

of the schools with kitchens had a quite low electricity con-

sumption. This caused the high errors between and 9

a.m. The errors in the evening are smaller compared to the

other typical load shapes presented in Figs. 11-13.

c

%

lam

no3

I

I

9 2 3 4 5 8 7 n 9 10 11 12 13 14 15 16 17 In 19 20 n P 23 24

Hov

Fig. 13. Total demand from 14 schools (thick line) not included in the

‘schools without kitchen’ category, compared to the typical load shape (thin

line with boxes) of that category. Monday 950821, mean daily outdoor

temperature 20.6”C.

152 C. Nor&, J. Pyko /Energy and Buildings 28 (1998) 145-157

al

I

, 2 3 4 5 5 7 8 9 10 I1 12 13 14 IS 15 17 18 19 20 21 22 23 24

-

Fig. 14. Total demand from 14 schools (thick line) not included in the

‘schools without kitchen’ category, compared to the typical load shape (thin

line with boxes) of that category. Wednesday 951011, mean daily outdoor

temperature 12.8”C.

6,

Results for hotels

The hotel category consists of nine objects with floor area

ranging from 5000 to 37,500 m2, eight of them are located

within city area and one in the countryside.

6. I _ Year profiles

Fig. 15 shows the year profile for Mondays to Fridays.

Christmas and New Year are easily recognised, when elec-

tricity consumption is much lower than for the rest of the

days during the same time period. Standard deviations are

much higher during Christmas and New Year when the activ-

ity levels vary greatly between different hotels.

During Easter, there is another obvious drop in the con-

sumption profile, and the third drop of the year takes place in

the beginning of May, around May day. During summer,

standard deviations increase and the electricity consumption

is lower than during the rest of the year. The major reason for

the higher standard deviations during summertime is that one

of the hotels has a positive correlation between electricity

consumption and outdoor temperature. This hotel is the most

recently built of the nine included in this study. Seven of the

nine hotels showed a negative correlation between tempera-

ture electricity consumption. Temperature may not be the

only reason of lower consumption, as hotels with much light-

ing equipment will be affected by more hours of daylight,

and consumption decreases. Akbari et al. [9] report that

indoor lighting and miscellaneous equipment shows a lower

consumption during summer months, compared to the rest of

the year. These results are valid only for the USA, but it is a

possible reason forthe variations in observed Swedishresults.

c-j

Fig. 15. Year profile for hotels, weekdays.

The highest daily consumption occurs during the winter.

No special day-type or time period can be identified. Gener-

ally, the highest daily consumption can be found in December

to February.

During standard Mondays-Fridays, the standard devia-

tions are generally low and during spring and autumn com-

prise to 6-10% of the mean values. The absolute standard

deviations increase slightly on non-standard weekdays,

depending on the activity level at the different hotels. Some

are open as usual, but others only serve food for guests staying

at the hotel while the public restaurant closes early.

There is also a noticeable difference between different

weekdays. The consumption on Mondays and Fridays is

clearly lower compared to Tuesday-Thursday. The differ-

ence is approximately 5%, and this is another possible source

of errors when analysing the daily load shapes, where all

standard weekdays are included.

Fig. 16 shows the year profile for Saturdays and Sundays.

Again, consumption decreases during Christmas and New

Year. The reasons are the same as previously discussed.

The electricity consumption is

almost

constant between

April and October, with a tendency to increase during the

winter months. Standard deviations are almost constant,

except during Christmas and New Year and in the beginning

of August. The reason for this sudden increase of standard

deviations and electricity consumption during the beginning

of August, cannot be explained. The relative standard devi-

ations are higher during most of the year than for Monday to

Friday, mainly due to a slightly lower consumption during

weekends. This depends on the activity levels. On weekdays,

most hotels are open until late in the evening, but during

weekends, some are still open until late in the evening, while

others close earlier. The most probable cause of the lower

electricity consumption during weekends is the kitchen, since

it is generally open for fewer hours during weekends than

during the weekdays.

Differences in electricity consumption on any day might

depend on several factors, and some of the most probable are

listed below:

l

Different numbers of occupied beds

l

Different operating hours for the kitchen, some of the hotel

kitchens are open until , while

some

close earlier in

the evening

C-1

- & 6 & 5 5

5 E f

5 5 2

6

m

Fig. 16. Year profile for hotels, Saturdays and Sundays.

C. Nor&, J. Pyrko/Energy and Buildings 28 (1998) 145-157 153

0

Some hotels have installed electricity saving equipment,

four of the hotels report having installed electricity saving

equipment.

6.2. Daily load shapes

As shown in Fig. 17, the demand does not fluctuate very

much for hotels. The difference between minimum and max-

imum demand is less than 100% which should be compared

to me school category where me difference was 4-6 times.

Between to , the demand is almost constant.

Night-time demand is quite high, depending heavily on light-

ing equipment and ventilation. The hotel kitchen equipment

and indoor lighting are assumed to be the reason for the major

part of the daytime demand variations.

Absolute standard deviations are constant during early

morning hours. It is not until that they begin to increase

from 0.1 during night-time, to 0.14 during the later part of

the morning (-), and then stabilises after 10

a.m., at around 0.12. The higher standard deviations during

morning hours depend to a great extent on breakfast time.

Survey answers indicated that breakfast time starts between

and 7: at the different hotels.

After , me demand is constant until lunch is over at

, which is followed by a slight demand decrease. At 5

p.m., another increase occurs, and standard deviations also

become larger at this time, depending on whether the hotel

serves dinner or not. One hotel in me study did not serve any

food in the evening, which affected the errors.

After , the demand decreases rapidly, some of the

kitchens are beginning to close, while others are open until 1

a.m. This is assumed to be the major reason of the higher

(0.16-0.20) standard deviations during the evening.

Fig. 18 shows another hotel load shape for standard week-

days, but this time, for higher outdoor temperatures. The load

shape is displaced about 0.1 compared to the load shape in

Fig. 17, which may depend on temperature, but also on other

factors like different operating hours for lighting equipment

because of more hours with daylight. Absolute standarddevi-

ations are slightly higher during daytime than for lower tem-

perature, but again, it is one single hotel that affects the

standard deviations. If this hotel is excluded, the standard

deviations are of the same size as for lower temperatures.

Fig. 19 shows a daily load shape for standard weekend

days. The weekend shapes are characterised by lower daytime

1.6

t-1

P 1.2

1.4

i 0.:

1 1:

0.2

0

1 2 3 4 5 6 7 6 9 10 11 12 13 14 15 16 17 16 19 20 21 22 23 24

HOW

Fig. 17. Standard weekday load shape for hotels, temperature below 0°C.

Mean temperature - 2.I”C.

0

1 2 3 4 5 6 7 6 9 10 11 12 13 14 15 16 17 16 19 20 2, 22 23 24

HCIW

Fig. 18. Standard weekday load shape for hotels, temperature interval 15-

20°C. Mean temperature 17.3”C.

(-)

1 2 3 4 5 6 7 6 9 IO II 12 13 14 15 16 17 16 19 20 21 22 23 24

HOW

Fig. 19. Standard weekend load shape for hotels, temperature below 0°C.

Mean temperature - 2.3”C.

demand than on weekdays and slightly higher standard devi-

ations. Night-time demand is almost equal to the weekday

shape, but 15-20% lower during the rest of the day. This

indicates that night-time loads are constant over the year and

little affected by the occupancy rate.

A possible reason for the higher errors is the opening hours

of the kitchen. During the morning, breakfast is served as

usual, but after , the standard deviations increase,

indicating that some hotels do not use as much electricity as

others during weekends. This is the case for hotels located

outside the centre of the city. Generally, it was found that

hotels with public restaurants located in the centre of the city

have higher consumption during weekend days, compared to

the ones located outside the city centre.

6.3. Relationship between outdoor temperature and

electricity consumption

Fig. 20 shows weekday load shapes for six temperature

intervals. No standard deviations are shown in the figure. As

the figure shows, the demand is lower at higher temperatures,

except between and Between and ,

it is the demand at lower temperatures that is most affected,

but in daytime and in the evening, only the demand at tem-

peratures above 5°C is affected by temperature.

During daytime, the lowest demands corresponds to high

temperatures, but hotels with a high amount of cooling equip-

ment will not be affected to such a high degree. The results

are presented in Tables 5 and 6. Days with mean outdoor

temperatures above 17°C and holidays are excluded.

Five of the hotels show a significant negative correlation

between temperature and electricity consumption, three

hotels shows a non-significant correlation, and the last one, a

154

C. Nor&, J. Pyrko/Energy and Buildings 28 (I!W8) 145-157

% ’

0.8

I

$f 0.6

s

0.4

1 I

0

1 2 3 4

I

5 6

7 8

I

9 IO 11 12 13 14 15 16 17 18 19 20 21 22 23 24

HOW

Fig. 20. Standard weekday

Table 5

Standard weekday

Hotel

1

2

3

4

5

6

7

8

9

load shapes for hotels, for six different temperature intervals.

regression results for hotels

Adj. R2

0.73

0.04

0.07

0.23

PO(-)

1.18

1.06

1.05

0.98

Psp ( -/“C)

-0.018

-0.002

0.003

0.009

Significance

0.000

0.006

0.005

0.000

0.678

0.000

0.000

0.000

0.000

general rule, predicting how electricity consumption depends

on outdoor temperature, can be determined for hotels.

The regression results for weekends are presented in Table

6. The intercept, PO, is slightly lower compared to weekdays,

and only four objects show significant correlation. Still, it is

only the two hotels with some electrical space heating, that

have high R2-values.

7. Verification

To verify the typical load shapes, measured data from 1993

was used. One hotel with gas furnace was used to verify the

hotel load shapes. Three different weekdays were randomly

chosen.

Fig. 21 shows the typical hotel load shape compared to a

measured load shape. The typical load shape matches the

measured load shape quite well during most of the day. Dur-

ing late night and morning hours, errors are approximately

10%.

Fig. 22 shows another hotel load shape, this time at colder

weather conditions. Errors are of the same size as in Fig. 21,

except during late night and early morning hours. Altogether,

the typical load shape gives a fair approximation of a load

shape for a hotel. Errors are below + 10% during most of the

day, except during late night and early morning hours.

The last hotel load shape is shown in Fig. 23. Errors during

the chosen Saturday are less than IO%, except during late

1.05

1.08

1.05

1.19

1.05

0.000

-0.003

-0.003

-0.016

-0.003

0.00

0.06

0.13

0.70

0.08

Table 6

Standard weekend regression results for hotels

Hotel

1

2

3

4

5

6

7

8

9

PO ( - 1

1.1

0.92

0.89

0.88

0.91

0.9

0.99

0.99

0.97

psp ( -/“Cl

-0.016

0.000

0.004

0.007

-0.002

0.000

-0.003

-0.013

-0.002

Adj. RZ

0.77

0.00

0.05

Significance

0.000

0.802

0.039

0.10

0.02

0.00

0.18

0.60

0.04

0.000

0.043

0.767

0.000

0.000

0.018

significant positive correlation (hotel 4). Hotel 4 is the one

most recently built and the most luxurious of the hotels in the

study. The intercept load at 0°C varies by more than 20%,

and it is the two hotels with partial electrical heating, that

have the highest consumption at 0°C and the highest temper-

ature dependence.

R*-values are generally poor, except for the two hotels with

partial electrical heating. Although the regression line cannot

explain more than 23% of the variations in electricity con-

sumption for the other seven hotels, the correlation between

electricity consumption and outdoor temperature is signifi-

cant in four of these seven. The conclusion of this is that no

data

from October

Fig. 21. Typical load shape for a hotel (thin line with boxes) and measured

weekday in 1993 (thick line). Temperature interval 5-

10°C. A 95% probability for the associated standard deviations.

C. Nor&, J. Pyrka /Energy and Buildings 28 (1998) 145-157 155

tion varies only slightly during the year, except during Christ-

mas and New Year, when consumption is much lower than

during the rest of the year.

8. Comparison to other studies

0

I I !

I

I 2 3 4 5 6 7 6 9 to 11 12 13 14 15 16 17 1s 19 20 21 22 23 24

In order to compare the load shapes from this study to load

Hour

Fig. 22. Typical load shape for hotels (thin line with boxes) and measured

shapes developed within other studies, usually presented in

data from December weekday 1993 (thick line). Temperature below 0°C.

W/m* terms, it is necessary to use an annual electricity con-

A 95% probability for the associated standard deviations.

sumption to obtain a W/m’-load shape. Lawrence Berkeley

Laboratory (LBL) has developed load shapes for schools

and lodging buildings, where the lodging category includes

hotels and motels, and measured load data for these two

categories are compared to the results from this study. The

temperature interval 5-10°C was chosen for the representa-

tion of the Swedish load shapes to provide the best represen-

tation of the annual mean outdoor temperature. LBL load

shapes for coastal buildings were assumed to be most similar

1 2 3 4 5 6 7 6 9 10 1, 12 13 14 15 16 17 16 19 20 21 P P 24

to Swedish conditions and all objects in this study are located

Hour

in the coastal area too. The comparisons are shown in Figs.

Fig. 23. Typical load shape for hotels (thin line with boxes) and measured

24 and 25.

data from December Saturday 1993 (thick line). Temperature below 0°C.

The school building load shape from this study is correlated

A 95% probability for the associated standard deviations.

to the LBL school load shape, although there is a vertical

evening hours, and twice in the middle of the day. The unex-

displacement during hours to The highest dif-

plainable high demand that occurred between and 11

ferences occur in the morning and an important reason is the

p.m., caused very high errors during these hours.

fact that this study also includes secondary schools which

were observed to begin earlier in the morning compared to

the primary schools. The reasons are the same for the differ-

7. I. Concluding remarks

ences that occur in the late afternoon. An interesting obser-

vation is the effect of the cooling equipment. If cooling

A typical daily load shape can be developed for hotels with

equipment is included, there is a 2 W/m* displacement during

reasonable accuracy, and typical daytime standard deviations

daytime, but if cooling is excluded, the load shape is very

are approximately 8-10% of the mean values. Errors begin

similar to the LBL load shape. Swedish schools are very

to increase when temperatures rise above 15°C which

seldom equipped with cooling systems.

depends very much on the air conditioning system in the

If the hotel load shape is compared to the LBL results for

hotels. During weekends, errors are higher than during week-

the lodging category, there are some noticeable differences.

days. This is quite natural since weekend activity levels are

Night-time demand match the LBL load shape well but major

very different for different hotels. The electricity consump-

differences occur during daytime when there is a 3-4 W/m*

20

18

-___-

16

14

I2

“e

5

10

8

6

4

2~ --

- ~- -__

0, : : : : :

12 3 4 5 6 7 8 9 IO 11 12 13 14 IS 16 17

18 19 20 21 22 23 24

Hour

Fig. 24. Load shape for schools without kitchen from this study compared to results from Akbari et al. [9].

156

C. Nor&, J. Pyrko /Energy and Buildings 28 (1998) 145-157

I 2 3 4 5 6 7 8 9 10 II I;..;3 14 IS 16 17 It7 19 20 21 22 23 24

Fig. 25. Load shape for hotels from this study compared to results from Akbari et al. [9]

difference between the two load shapes. The LBL lodging

load shape is more flat and the daytime demand is lower

compared to the load shape developed within this study. The

most probable single reason for the differences is less usage

of electrical cooking equipment in the lodging buildings.

Cooking equipment is reported to only account for approxi-

mately 1% of the annual electricity consumption in lodging

buildings [ 91. The Swedish Vattenfall study reports 30% for

kitchen equipment in Swedish hotels [3]. All of the hotels

in this study are equipped with kitchens and most of them

have quite large kitchens, providing the hotels’ public restau-

rants with food.

Two load shapes are very well-correlated: the LBL lodging

load shape compared to the weekend load shape developed

within this study. An interesting observation from the LBL

study is that no differences could be observed between stan-

dard days and non-standard days. In this study, major differ-

ences between standard weekdays and weekends were

observed and previously, it was stated that less cooking equip-

ment is used during weekends in Swedish hotels compared

to weekdays. The major reason for the differences between

the LBL results and the weekday load shapes from this study

seems to be less usage of electrical cooking equipment for

the lodging category.

LBL load shapes and the load shapes developed in this

study are quite similar although there are substantial differ-

ences. The similarities between the load shapes are better

than initially expected by the research group, though different

conditions in Sweden and the USA.

but if O&M problems should be detected, hourly load shapes

probably are among the best tools. There are several appli-

cations for the load shapes presented in this study, such as:

(a) conversion from annual electricity consumption to hourly

load shapes. The set of typical load shapes can be used to

estimate the building load profile without load measurements.

For example, the school load shapes have been used to eval-

uate the possibilities to install a gas engine for small scale co-

generation in a school building. (b) Identification of O&M

problems. If O&M problems are suspected, it is possible to

compare the load profiles obtained by the normalised load

shapes with measured data. In case of differences, the time

of day and magnitude of the deviations can be identified. (c)

Evaluation of energy efficiency projects. If an energy effi-

ciency retrofitting is carried out and no hourly measurements

before the retrofitting were carried out, the electricity con-

sumption from the previous year can be used to estimate the

hourly load shape before the retrofitting which then can be

compared to the hourly load shape after the retrofitting.

10. Conclusions

The typical load shapes give a reasonable approximation

of the measured load shapes, although the relative standard

deviations are quite high during certain hours, but this is

mainly due to decreasing demand. Daytime (-)

results are generally good with standard deviations below

10% of mean values but there are quite high day-to-day var-

iations, where some possible reasons have been discussed.

Only a limited amount of data was available for verification,

although the typical load shapes provided quite good approx-

imations of the measured load shapes. Comparisons with

results from Lawrence Berkeley Laboratory showed both

similarities and differences, especially for the hotel category,

but it is important to consider that Swedish and American

conditions are different. The developed load shapes can be

applied if a fair approximation of a school/hotel load shape

is required. The presented methodology is simple to use for

similar studies.

9. Applications of the results

Knowledge of electricity consumption patterns and elec-

tricity consumption indicators is necessary for developing

new tools for energy auditors and for identifying operational

and maintenance (O&M) problems. Several types of indi-

cators are available, such as: annual electricity consumption

figures, hourly load shapes and other efficiency indicators,

C. Nor&, J. Pyrko / Energy and Buildings 28 (1998) 145-157 157

Acknowledgements

This work was granted by the Swedish Council for Build-

ing Research and Swedish Electrical Utilities’ R&D Com-

pany Elforsk, under Contract 9402.57-g.

References

[II

Electricity End-use Efficiency, International Energy Agency, Organ-

isation for Economic Co-operation and Development, Paris, 1989.

(21

Belastningsberaekning med typkurvor, Svenska Elverksfoereningen,

Stockholm, 1991 (in Swedish),

[31

Lokalerna och Energihushaallningen, Vattenfall U, Vaellingby, 1991,

p, 70 (in Swedish)

[41

K. Livik, Main findings of load research in Norway between 1980-

1985, EFI TR 3411, Trondheim, 1987.

151

N. Feilberg, K. Livik, Energy and load structure at various categories

of end-users, EFI TR 4074, Trondheim, 1993.

[61

K. Livik, 0. Rismark, Belastningskurver for bygg med forskjellige

oppvarmingssystemer, EFI TR 3726, Trondheim, 1990 (in

Norwegian).

[71

H. Akbari, Validation of an algorithm to disaggregate whole-building

hourly electrical load into end-uses, Energy 20 ( 12) ( 1995) 1291-

1301.

[81

H. Akbari, et al., Analysis of commercial whole-building 15-minute-

interval electric load data, ASHRAE Trans. 94 (2) ( 1988)

[91

H. Akbari, et al., Integrated Estimation of Commercial Sector End-

Use Load Shapes and Energy Use Intensities, Phase 2, LBL-30401,

CA, 1991.

H. Akbari, et al., Integrated Estimation of Commercial Sector End-

Use Load Shapes and Energy Use Intensities, LBL 27512, CA, 1989.

H. Akbari, et al., A Review of Existing Commercial Energy Use

Intensity and Load Shapes Studies, LBL 29209, CA, 1990.

H. Akbari, et al., Integrated Estimation of Commercial Sector End-

Use Load Shapes and Energy Use Intensities in the PC&E Service

Area, LBL 34263, CA, 1993.

[I31 H. Akbari, et al., A New Approach to Estimate Commercial Sector

End-Use Load Shapes and Energy Use Intensities, LBL 35372, CA,

1994.

[I41

C. Nor&n, Typical Load Shapes for Six Categories of Swedish Com-

mercial Buildings, Lund Institute of Technology, LUTMDN/TMVK-

5279-SE, Lund, 1997.

[I51

Selected statistical methods for analysis of load research data, EPRI

EA-3467, CA, 1984.

11’31

N. Feilberg, K. Livik, Sammenlagring av effekt, Beskrivelse av Sta-

tistisk Metode, EFI TR 3806, Trondheim, 199 1 (in Norwegian).

[I71 K. Johansson, M. Mohammadian, Jaernaakraskolans koek i Lund,

Lund Institute of Technology, Lund, 1996 (in Swedish).