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).
本文发布于:2024-09-24 11:25:47,感谢您对本站的认可!
本文链接:https://www.17tex.com/fanyi/11762.html
版权声明:本站内容均来自互联网,仅供演示用,请勿用于商业和其他非法用途。如果侵犯了您的权益请与我们联系,我们将在24小时内删除。
留言与评论(共有 0 条评论) |