More or Less about Data
- Analyzing Load Demand in Residential Houses
Juozas Abaravicius, Kerstin Sernhed and Jurek Pyrko, Lund University
ABSTRACT
Load demand in residential houses is a significant contributor to peak load problems
experienced by utilities. The knowledge about demand variation in households is fairly limited as
well as the use of various tools to analyze the demand. Many utilities have recently installed
interval (hourly) metering at their residential customers. The availability of hourly data is a
significant progress, however, the utilities use this data only to a limited extent, mostly for billing
purposes only. This study aims to discuss the possibilities and the benefits of using this valuable
data.
There are several established load analysis tools, such as load curve, typical load curve,
load duration curve, load factor, superposition factor, etc., which utilities could apply and
develop to provide feedback to small electricity users. Among other benefits, the hourly load
data analysis can provide the detailed characteristics of load demand, define the consumption
patterns and can help to identify which households contribute most to the utility peaks. This
information is essential when developing new energy services, appropriate pricing, load
management strategies and demand response programs.
Through the analysis of strengths and weaknesses of different load analysis tools, this
paper defines the knowledge they could give, how applicable they are and what value they could
have both for the utility and the residential customer. The study is exemplified with ten cases of
households with electric space heating in Southern Sweden.
Load demand and load data in households
Traditionally, when approaching load demand problems, the focus is on bigger electricity
users (industrial). But the fact is that the residential, commercial and services sector accounts for
half of the total electricity consumption in Sweden (Swedish Energy Agency 2003). Electric
space heating currently accounts for just over 30% of the total electricity consumption in the
sector. High electric load demand variations occur in winter season together with temperature
variation. Load demand in Sweden increases by 350 to 400 MW for each ºC of outdoor
temperature drop (Pyrko, 2004). Furthermore, load demand in the residential sector varies
significantly during the day and normally has its peaks during morning and evening hours.
The dominating energy source for heating and domestic hot water for detached residential
houses in Sweden is electricity. An increased number and a variety of household equipment may
also lead to load shortages if used simultaneously.
Most energy experts agree that the residential sector should be seriously considered when
approaching peak demand problems and ensuring a well functioning electricity market.
The existing knowledge about demand variation in households is fairly limited. Many
utilities have recently installed or consider installing interval (hourly) metering at their
residential customers. This is partially enforced by the new law of billing on actual electricity use
(Sernhed 2004). The availability of hourly electricity use data is a big step forward, however, the
use of various tools to analyze the demand is limited. The utilities use collected data mostly for
billing purposes.
The Swedish electricity market was de-regulated in 1996. However, if private customers
wanted to change their supplier (retail company) they were forced to invest in a new electricity
meter with hourly metering. The cost of such a device was typically around 900EUR. Very few
customers changed their supplier at this stage (Matsson 2001). The requirement for hourly
metering was abolished in 1999. It was replaced by a profile-settlement, meaning that different
consumption patterns are applied when estimating the electricity consumption within a specific
period of time. Each pattern is valid for all customers within a specific geographical area (Wallin
2005).
The major advantage of new profile-settlement was that the electricity consumers with
smaller consumption (residential) were able to switch electricity supplier and directly benefit
from the new electricity prices without having to invest in the metering system. One major
disadvantage was that the connection between real physical electricity use and customer
electricity cost in high-peak periods vanished (Wallin 2005).
There are a number of suppliers of new metering systems promoting their products. With
the latest technical development of automated meter reading (AMR) systems the residential
customer becomes more “visible”. Hourly data is available now both to the customer and the
utility. New meters allow even more detailed statistics as well as opens new possibilities for
customer-supplier dialogue and the development of new energy services. This is a new
possibility but at the same time a new challenge, requiring reconsideration of long term customer
– utility relations. In the present market conditions keeping the customer becomes the major task
for a utility. The use of modern metering and communication system could be seen as a
competitive advantage influencing customer choice.
The objective of this paper is to discuss the ways and benefits of using this valuable enduse data. The electricity use of ten pilot households is shown as an example of using the analysis
tools described in this paper.
Analyzing load demand
Several characteristics can be derived when analyzing load on the demand side to describe load
demand conditions as:
• Magnitude of energy use and load demand
• Variation of the demand in time
• Minimum and maximum demand values
• Duration of minimum and maximum load
• Contribution to total pattern/total utility load
There are several established load analysis tools, such as load curve, typical load curve,
load duration curve, load factor, superposition factor, etc, which are used to describe these
conditions. The definitions, interpretation and applicability of these tools are discussed in this
section.
Load curve and load duration curve
Load curve illustrates variation of load demand during a specific period. A load curve can
be converted into a load duration curve showing duration of a particular load demand. A
graphical explanation of load curve and load duration curve is given in Figure 1.
Figure 1. Load curve and load duration curve (Pyrko 2004)
Source: Pyrko 2004
The load curve is a good visual representation of electricity use in a household and its
variation over time. It shows minimum and maximum use values against time-of-day and,
depending on the given resolution, their duration. It could be considered as a “user friendly” way
to explain the customer peak load phenomenon and can be an incentive for the customer to look
at what is beyond the needle peaks in the consumption. Therefore this method should be an
attractive one for the customer (expected to have some sort of behavioral influence) and
convenient for the utility. However, to be realistic, one should keep in mind the question when
and why would customer care to look at their load demand. Three conditions could be
emphasized:
•
•
•
Customer is interested to see the existing pattern
Customer is interested to see the changes when some measures (energy/load saving
etc.) have been implemented in the household
Customer has a direct incentive to care about load pattern when, for example, having
a tariff with load demand component. However, an important prerequisite here is that
customer would have to have instantly available information on a device or screen,
otherwise the feedback would come too late to expect changes in use.
It should also be mentioned that load curve could be a useful way to identify the operational
problems when running the building.
The load duration curve for the utility is essential information about customer’s load
demand, especially looking at the longer perspective (full year or season). A significant benefit
given by the duration curve for the utility is that it directly shows the number of hours when the
demand exceeds a certain load demand level, where there’s the greatest need for load control
actions.
Typical load curves
The typical load curve is different from the simple load curve as it is more than just a
momentary visualization of electricity use. It normally shows the mean load demand value for
each hour during the specified period (week, month, year, etc.). The typical load curves are often
developed to compare the demand on weekdays and weekends. Figure 2 shows the daily load
curve of one household for the winter period. The average and maximum hourly load as well as
standard deviation is shown.
The typical load curve can be used for comparison of “typical” patterns and for a selected
time the actual energy performance. This method therefore is useful both for customer and
utility. It is a good graphical representation and clear picture of household-specific energy use
and daily consumption pattern. The method provides valuable information for a utility for
designing demand response strategies and new value-added energy services.
Figure 2. Daily load curve example
12,00
10,00
kWh/h
8,00
Average
Max
6,00
4,00
2,00
0,00
1
3
5
7
9
11 13
15 17 19
21 23
Hour
Load coincidence
Different customers naturally differ in the load curves/patterns. What principally matters
for the utility is the total load pattern of all the customers – the coincidence load. The
coincidence load curve should be as even as possible. This would be a favorable situation for the
utility. The coincidence factor is the ratio of coincident load maximum value to maximum value
of partial load demand (equation (1)).
CF = Pmax / ΣPimax
where:
Pmax = coincidence load maximum value,
Pimax = maximum value of partial load i.
(1)
This ranges between 0 and 1 as coincident demand should always be less than or equal to the
maximum demand.
Load Factor
Load factor is simply the ratio of the average load during a specific period of time to the
maximum load occurring during that period, as given in equation (2):
Load average
Load Factor =
(2)
Load
max
The load factor is used to demonstrate variations of the household’s load demand. This
factor can range between 0 and 1, where a value of 1 would indicate that the household load
curve was completely flat and no peaks were present. From the supplier’s point of view, it is
preferable to remove peaks and flatten out the load curve, corresponding to an increase in load
factor.
The load factor is a good tool both for utility and customer. However, the major
shortcoming of the load factor is that it does not represent the magnitude of the consumption, i.e.
the value of the highest load and average loads. For the utility, looking at just the load factor,
makes it difficult to judge the load reduction potential for the specific customer and the influence
on the total utility load.
The load factor can be used as a parameter when designing or analyzing new electricity
tariffs. For example, it is a way to evaluate the load pattern before and after the introduction of a
new load tariff.
Exploitation Time
The exploitation time provides information about the shape of the customer’s load
demand curve. The time, calculated in hours, represents the required duration of maximum
(peak) load needed to correspond to the total actual electricity usage during the same period,
which can be represented by the equation (3). A high exploitation time relates to an even load
demand – a preferable situation for the utility (North 2001).
Exploitation Time (h) =
Electricity Consumptio n
Load max
(3)
Exploitation time has more value for the utility than the household customer as is more
complex parameter than load factor and would not provide meaningful information for the
customer.
Superposition factor
Superposition describes one specific customer's influence on a total utility load curve or
the contribution of partial load to the total load. Superposition factor is the ratio between the
partial load demand during the total peak and the maximum partial load during the same time
period as it is expressed in Figure 3 and equation (4).
SF = ΔPmax/pmax
(4)
where:
ΔPmax = increase of total load peak value due to partial load p,
pmax = partial load maximum value
Figure 3. Superposition and superposition factor (Pyrko 2004)
Source: Pyrko 2004
The range of values that this factor can take is between 0 and 1, where a value of 1 would
indicate, that the peak of the partial load coincided with the peak of total load. What the
superposition factor doesn’t tell is how big the partial load demand is that contributes to the
maximum load of the system, doesn’t tell the magnitude of the use of the specific customer.
Superposition factor is, of course, a more “utility oriented” tool. This factor could be the
major decision driver to select customers for participation in DR programs. The factor identifies
which customers should be addressed first and where the load management activities would
actually give the desired results. Having this knowledge, for example, the utility could approach
the specific customer, or group of similar customers, with the special proposals for DR actions
(tariffs, load control programs, etc.). Obviously, it is necessary to look at many peaks to be able
to determine if the peaks coincide by chance or if there is an evident correlation.
Examples of ten households
This study is exemplified with ten cases of households with electric space heating in
Southern Sweden.
Our associated utility Skånska Energi AB has installed advanced metering system
“CustCom” to all its customers (99% of those are residential houses)(Abaravicius, 2004). The
system is able to provide automatic hourly measurements, as well as electricity control and
information services. At the moment it is used only for measurements and billing purposes.
Hourly use data is automatically collected by the utility and is available for each customer via
Internet. The hourly data is available in form of tables and load curves during a specified period.
Figure 4 shows the typical daily load curves (average value during specific hour) and
coincident load curve for the10 analyzed households during December 2003.
Figure 4. Households’ typical daily load curves and coincident load curve during
December 2003 (Sernhed 2004)
45
6
kW
5
4
3
2
01:00
04:00
07:00
10:00
1
40
K10
35
K9
30
K8
25
20
K7
15
K5
10
K4
5
K3
0
K2
K6
13:00
0
22:00
01
:0
0
04
:0
0
07
:0
0
10
:0
0
13
:0
0
16
:0
0
19
:0
0
22
:0
0
16:00
19:00
K1
Source: Sernhed 2004
Table 1 shows an example of load factor and the exploitation time for 10 households in Southern
Sweden calculated for period April 1, 2003 – Feb 15, 2004.
Table 1. Load factors and Exploitation time for 10 analysed households
Household
H1
H2
H3
H4
H5
H6
H7
H8
H9
H10
Total
consumption
kWh
16906
21910
16525
17506
18695
13405
12757
13966
11874
15422
Average
load
kWh/h
2,19
2,84
2,14
2,27
2,43
1,74
1,66
1,81
1,54
2,00
Load
Factor
0,17
0,24
0,23
0,22
0,22
0,28
0,29
0,27
0,23
0,22
Exploitation
time
h
1339
1877
1798
1683
1698
2128
2254
2060
1770
1719
The example shows that households H6, H7, H8 have the highest load factors and highest
exploitation times among the analyzed objects. From the load demand point of view, these
customers could be seen as the most favorable for the utility. On the other hand, these
households (together with H9) also have lowest total electricity consumption, and average load
demands, that, in turn, decreases their significance for the total demand conditions of the utility.
Table 2 shows results of a superposition factor analysis performed for 10 households in
order to observe which of them are mostly contributing to the utility peaks. The utility’s ten
highest hourly peaks within the analyzed period were selected and the superposition factors for
the households were calculated. The results indicate that households H10, H4, H5, H6, H8 were
mostly contributing to the selected utility peaks (SF ≈ 1,0). What superposition factor doesn’t tell
is how big the partial load demand is that contributes to the maximum load of the system.
Table 2. Superposition factor for 10 analysed households
10 utility
peaks
date
2004-01-22
08:00
2004-01-22
18:00
2004-01-21
19:00
2004-01-21
08:00
2004-01-26
18:00
2004-01-27
18:00
2004-01-05
18:00
2004-01-23
08:00
2004-02-12
08:00
2004-01-02
18:00
kWh/h
Out.
Temp.,°C
H1
H2
H3
H4
H5
H6
H7
H8
H9
H10
80387
-15
0,81
0,95
0,81
1,00
1,00
0,65
0,66
0,91
0,81
1,00
76400
-9
0,78
0,79
0,89
1,00
0,59
0,70
0,82
0,78
0,89
0,80
74019
-8,1
0,69
0,66
0,92
1,00
0,75
0,56
0,74
0,88
0,70
0,73
73773
-9,8
0,78
0,86
0,76
0,83
0,99
0,55
0,59
0,86
0,57
0,95
72202
-3
0,75
1,00
0,90
0,95
0,81
0,68
0,78
1,00
0,60
0,60
71429
-4,3
0,66
0,92
0,82
0,81
0,76
0,64
0,94
1,00
0,56
0,93
71128
-7,4
0,81
0,91
0,83
0,59
0,97
1,00
0,85
0,79
0,68
0,93
70934
-3
0,88
0,86
0,81
0,86
0,82
0,66
0,55
0,68
0,55
1,00
70614
-4,3
0,31
0,80
0,73
0,87
1,00
0,72
0,79
0,53
0,65
1,00
69494
-3,9
0,74
0,89
1,00
0,92
0,59
1,00
0,86
0,97
0,71
1,00
What’s beyond the peaks?
For experimental purposes two extra meters were installed to measure load demand for
heating and hot water. Figure 5 provides the overview of the average of 10 highest peaks of
every household and the peaks’ composition. The available metering of partial loads allows
insight into the origin of the peaks, i.e. whether it is a climate dependent or behavior dependent
peak. The composition is presented in two ways – in kW, in order to see the peak value and as a
share of the load demand.
Figure 5. Ten highest peaks at the analyzed households
10 peaks average composition
10 peaks average composition
100%
12,00
10,00
80%
kWh/h
8,00
60%
household
Hot water
Heating
6,00
household
Hot water
Heating
40%
4,00
20%
2,00
0%
0,00
H1
H2
H3
H4
H5
H6
H7
H8
H9
H1
H10
H2
H3
H4
H5
H6
H7
H8
H9
H10
This information has primarily value for the customer as they could directly see what
kinds of end-uses consume most energy. This kind of information could be treated as a specific
utility service. It should be mentioned, nevertheless, that the method might be costly as it
requires installation of extra meters.
The utility also benefits from this data as it provides good grounds and information to
develop value-added energy services, as promotion of specific appliances, installation of storage
and load control equipment, etc.
A very interesting and, in a way, unexpected result in Figure 6 is the electricity demand
for household needs which can be as high as 60% of the total demand during some periods.
Furthermore, these periods were measured during the winter season when one would normally
expect the heating demand to take the highest share. The important conclusion therefore is that
not only the climate (outdoor temperature), but the behavior related electricity use could be a
serious cause of load peaks. Especially risky is the coincidence of both.
In order to get a more detailed energy use description, diaries were filled in for 4 days by
household members. Based on this information and measurements it was possible to specify all
end uses’ contribution to the highest peak during the four-day period. The results are given in
Figure 6 (Sernhed 2004)
Figure 6. Composition of 10 household’s highest peak load
Drying cupboard
16
W ashing machine
14
Dishwasher
Co ffee maker
12
M icrowave oven
kW
10
Kitchen range
Oven
8
Remaining load
6
Sundry, hot water
4
Reheating, ho t water
Shower, hot water
2
Electric heater
0
Sauna
K1
K2
K3
K4
K5
K6
K7
K8
K9
Source: Sernhed 2004
K10
Space heating
Heating load
The measured load data for heating could be extended to various outdoor temperatures
using the regression analysis. This is a good tool for a utility to predict the demand related to
temperature variation, as well as to estimate the expected load savings by controlling (turning off
or partial decreasing) load for heating in load management programs. Heating systems are
normally the primary subject for load management/control programs as they have highest load
demands and could be manipulated with the least negative consequences for a customer.
One interesting idea to look into is if the separate measurement of heating load could
create a possibility to charge it separately. Different contracts for heat supply in this case could
be developed. Consequently, it could stimulate for instance the investment in heat storage
technologies, system automation, load control technologies, etc.
Hot water load
Hot water use has a tendency to increase during morning and evening hours in most
households and thus contribute to the total load peaks. Similar conclusion could also be drawn
from an analysis of probability that hot water boiler is on full power, presented in Figure 7.
An important question, when analyzing load demand for hot water, is when the hot water
boilers are on full power. Having this knowledge it is easier to create load management
strategies, as it gives a suggestion when to control the units in order to get a maximum load
savings. Using the daily load curve and the hourly data, the following methodology for finding
the probability if the hot water boiler is on full power was developed for our case studies: it is
assumed that boiler is on full power if the hourly load exceeds 2,5 kWh/h. Every hour of the day
through the investigated period is analyzed. Number of hours when the load reached this value is
divided by the total number of recorded hours.
The result for one of the households (H5) is given in Figure 7. Values on Y axis show the
probability (%) that the water heater is on full power versus the hours of the day (0-24). The
pattern of load demand for hot water varies from household to household, principally depending
on the behavioral factors. There is a tendency for the highest demand to occur during morning
and evening hours, therefore these periods could have the highest potentials for load control.
Figure 7. Probability that hot water load is on full power during a day
40,0%
35,0%
30,0%
25,0%
20,0%
15,0%
10,0%
5,0%
0,0%
1
3
5
7
9
11
13
Hour
15
17
19
21
23
Concluding discussion
New automatic meter reading technologies allow to measure more than the total
electricity demand. High resolution load data is now available both for the customer and the
utility. The open question remains how the data should be provided in order to make customer
interested in better understanding their energy use? Is it sufficient only to make it available
online? The research on this particular issue is on the way, however, it requires broad efforts and
trans-disciplinary cooperation of engineers, economists and behavioral scientists.
The load analysis tools, touched in this paper, are not new. The novelty here is their use
when analyzing residential load demand and their applicability in the residential market
conditions.
Load curves, daily load curves, load factor, and superposition factor should be used as
good representations of the household load demand. Load curve is more useful tool for the
customer, while the duration curve is more useful for the utility. Load factor is a useful indicator
both for the utility and the customer. However, the major shortcoming of the load factor is that it
does not represent the magnitude of the consumption, i.e. the value of the highest load and
average loads. For the utility, looking at just the load factor value makes it difficult to judge the
load reduction potential for the specific customer and how would it influence total utility load.
One interesting possibility is that the load factor can be used as a parameter when evaluating new
electricity tariffs.
Even though ideal market development demands broad participation in DR programs,
with the help of such a method as superposition, the utilities can start it with the households and
their energy uses that contribute the most to their peaks.
Typical load curves provide valuable information for a utility for designing demand
response strategies and new value-added energy services.
Partial loads measurement has primarily the value for the customer as they could directly
see what kinds of end-uses consume most energy. This kind of information could be treated as a
specific utility service. It should be mentioned, nevertheless, that it requires installation of extra
meters. The utility also benefits from this data as it provides good information to develop valueadded energy services, as promotion of specific appliances, installation of storage and load
control equipment, etc.
Special attention should be given for the electricity used for heating (as the highest
demand). One possibility is to measure and charge it separately. Different contracts for heat
supply could be developed.
There’s evidently higher interest for the utility in analyzing load demand or, in another
words, it is a utility oriented process. It provides a utility an essential grounds, sources to create
appropriate DR strategies (pricing, direct control, etc.).
References
Abaravicius, J. 2004. Load Management in Residential Buildings Considering TechnoEconomic and Environmental Aspects. ISRN LUTMDN/TMHP—04/7024—SE, Department
of Heat and Power Engineering, Lund, University, Lund, Sweden.
Matsson, P. 2001, Elstatistik som energitjänst (Electricity statistics as energy service) ISRN
LUTMDN/TMVK—7049—SE, Department of Heat and Power Engineering, Lund,
University, Lund, Sweden (in Swedish)
North, G. 2001, Residential Electricity Use and Control, Technical Aspects, ISRN
LUTMDN/TMVK--7051—SE, Division of Energy Economics and Planning, Department of
Heat and Power Engineering, , Lund University, Lund, Sweden.
Pyrko, J. 2004, Eleffekthushållning i byggnader (Load management in buildings), ISBN 91-6315905-8, Lund University, Lund, Sweden (in Swedish)
Sernhed, K. 2004. Effekten av effekten. Elanvändning och laststyrning i elvärmda småhus ur
kund och företagsperspektiv. Fallstudier. (Effects of load. Electricity use and load
management in electrically-heated detached houses from customer and utility viewpoints.
Case studies). ISRN LUTMDN/TMHP-04/7025-SE Department of Heat and Power
Engineering, Lund University, Lund, Sweden.
Swedish Energy Agency 2003. The Electricity Market 2003, ET 13:2003
Wallin, F. 2005, Added Values of Remote Collected Electricity Consumption Data. Software
Solutions for Demand –Side Management, ISBN 91-88834-91-3, Dept of Public Technology,
Mälardalen University, Västerås, Sweden