Page 1 of 13
Archives of Business Research – Vol. 9, No. 6
Publication Date: June 25, 2021
DOI:10.14738/abr.96.10290. Angelopoulos, M. K., & Pollalis, Y. A. (2021). Data Analytics to Improve Customer Energy Efficiency. Archives of Business Research,
9(6). 13-25.
Services for Science and Education – United Kingdom
Data Analytics to Improve Customer Energy Efficiency
Michail K. Angelopoulos
Department of Economics, University of Piraeus, Greece
Yannis A. Pollalis
Department of Economics, University of Piraeus, Greece
ABSTRACT
This research focuses on providing insights for a solution for collecting, storing,
analyzing and visualizing data from customer energy consumption patterns. The
data analysis part of our research provides the models for knowledge discovery that
can be used to improve energy efficiency at both producer and consumer ends. Τhe
study sets a new analytical framework for assessing the role of behavioral
knowledge in energy efficiency using a wide range of Case Studies, Experiments,
Research, Information and Communication Technologies (ICT) in combination with
the most modern econometric methods and large analytical data taking into
account the characteristics of the study participants (household energy customers).
Keywords: Big Data, Data Analytics, Energy Consumption, Energy Customer Experience.
INTRODUCTION
Nowadays, there is an increasing dependency on information and communication technology
(ICT). ICT services and products have changed the way of life dramatically. Any individual must
keep up with the current trend in order to be an effective part of the society and fulfil every
need. There are many research areas as a consequence of this continuous transformation such
as monitoring digital traces of human activities. Every activity produces a digital trace that can
be recorded and analyzed. This ranges from mobile phone activities and internet in general,
information about hobbies, interests, expenses etc. Big Data refers to these digital traces of
human activity. The superabundance of computing resources, fast and high mobile connectivity
has provoked a great surge in the data volumes. Exploiting these circumstances, businesses can
use all those tools as a new revenue stream.
In economic theory, energy use and economic activity in that energy are strongly associated.
The term energy efficiency is a critical factor for economic growth and has different meanings
depending on the state or organization that executes it. In [1] is shown that energy efficiency is
an environmental goal and exhibits the reckless use of energy by the final consumers and how
they respond to the costs. European energy policy looking for a balance in sustainable
development and competiveness, mainly by introducing the use of renewable energies and
other techniques connected with the energy sector. According to [2], the potential for energy
savings in the European Union (EU), calculated that the reduction in energy consumption on
2020-2021 could reach 20% which is translated to about 390-400 Mtoe [3,4].
Page 2 of 13
14
Archives of Business Research (ABR) Vol. 9, Issue 6, June-2021
Services for Science and Education – United Kingdom
Energy efficiency promotion is the most effective energy policy that contributes to all of the
basic energy goals, including the decrease of Greenhouse Gas (GHG) emissions and the
diminution of climate change negative consequences, lowering the cost of consumer energy
services [5,6]. So they become in a cost-effective manner, compliance with the above EU
strategic goals would contribute to [7]:
• Secure supply and independency from energy imports
• Job creativity and economic growth
• Reduction of the harmful impact of energy generation on the environment (mainly
due to GHG emissions)
• Improved level of living conditions
The implementation of all the existing measures and alternative new ones would lead to
financial savings of close to €1000-1500 per household every year.
One of the biggest energy saving potential lies in buildings. The population is constantly
increasing year after year especially in urban areas. For this purposes the construction or
transformation in energy-efficient buildings is essential [8]. The current energy performance of
the building sector in most countries is poor [8]. In the European Union (EU), the building sector
plays a vital role on climate policy [9,10].
The needs for those interested in the building value chain should be constantly integrated
within a concise energy transition framework, in order to guarantee its flexibility and
sustainability along the buildings’ life cycle, meaning all the stages from its conceptualization
to refurbishment/demolition.
The breakthrough of big data platforms and their related technologies formulates a novel
market opportunity for improving the energy efficiency along the lifecycle of building sector
and for energy management at building level. Data are continuously generated into buildings.
This exists because of leading-edge information and communication technologies (ICTs).
Internet of things (IoT), machine learning, artificial intelligence (AI) are falling into this
category too.
The topic of this paper focuses on the adoption of tools and techniques in order to study several
household energy needs and evaluate, quantify, analyze and extract some prediction models for
energy consumption in order to provide the basis for designing, planning and implementing
schemes for improving energy related services for achieving higher efficiency in both
production and usage. The insights generated from these use cases can also help in educating
the consumer about the benefits of energy efficiency and spread awareness about behavioral
changes from which the society and the individuals can benefit.
In the scope of this study is to exploit and combine past surveys and researches regarding
energy consumption patterns for a typical household in order to obtain useful knowledge for
energy efficient techniques. Also, this study exploit and combine data from several countries
from different places around the world such as Finland, Greece, Latvia and Spain. During all
surveys, important quantitative and qualitative data from households were identified by the
questionnaires such as socioeconomic and sociographic characteristics, number and type of
Page 3 of 13
15
Angelopoulos, M. K., & Pollalis, Y. A. (2021). Data Analytics to Improve Customer Energy Efficiency. Archives of Business Research, 9(6). 13-25.
URL: http://dx.doi.org/10.14738/abr.96.10290
electrical appliances and issues related to awareness, attitude and level of knowledge on energy
efficient techniques.
The rest of the paper is organized as follows: The first part describes the climate dependency
on energy consumption patterns. Afterwards, how personal characteristics of each consumer
and level of urbanization affects the consumption is analyzed. In the next part, building
classification on basis of energy efficiency is investigated. Finally, forecasting consumption
techniques and behavioral practices for energy efficiency are proposed followed by conclusions
of the whole report.
SEASONAL VARIATION IN ENERGY CONSUMPTION
Seasonal variation in the use of energy is a very common phenomenon. The energy needs of a
household is not stable every period of the year. The first important aspect of the analysis is to
investigate the sensitivity of energy usage with respect to outside temperature (impact of
seasonality). Another useful point is to separate the energy consumed for electricity and the
energy consumed for heating. All surveys, as expected, indicate that electricity used for heating
is more sensitive to temperatures than electricity used for other purposes. The latter category
shows a more stable trend throughout the year. The improvement of heating distribution and
usage systems can contribute to energy efficiency. Figure 1 confirms the above theory for
Finland and Spain.
Figure 1: Seasonal Variations in Consumption
It is obvious that in winter period the consumption reaches its highest levels nearly 2.5 to 3
times up in contrast to the consumption in summer months. The energy for electricity does not
appear to be influenced by the impact of seasonality.
0
50
100
150
200
250
300
350
0 2 4 6 8 10 12 14
Average Consumption (KWh)
Months
Seasonal Variations-Average Monthly
Consumption
Electricity (Finland) Electricity (Spain)
Heating (Finland) Heating (Spain)
Page 4 of 13
16
Archives of Business Research (ABR) Vol. 9, Issue 6, June-2021
Services for Science and Education – United Kingdom
The majority of households (2/3) cool their dwelling using air conditioning or other cooling
units for 2-4 months annually while 1/5 use such units for less than 2 months. In countries
with lower temperatures the use of cooling equipment is more intense.
In the following Table 1 the unit consumption (toe/dwelling) per dwelling and the percentage
used for heating purposes by the end consumers is presented for the four countries.
Table 1: Unit consumption (toe/dwelling) per dwelling
Heating Total Heating (%)
Greece 0.93 1.36 68.3
Spain 0.41 0.92 44.5
Latvia 1.30 1.74 74.7
Finland 1.46 2.05 71.2
It is easily extracted that in countries with cold climate conditions the total consumption is
higher and accordingly the percentage used for heating prevails the consumption for other
purposes.
CONSUMPTION BASED ON PERSONAL CHARACTERISTICS AND URBANIZATION
According to the survey results, energy consumption is directly connected with the level of
urbanization of the area in which the residency is located. Table 2 presents the annual average
thermal energy and electricity consumption for a typical Greek household in both cases of
urbanization (urban and rural). The need of a household in thermal energy is much higher in
rural areas in contrast to the needs for electricity which are higher in urban areas.
Table 2: Average thermal and electricity consumption in both cases of urbanization
Urban areas (%) Rural (%)
Thermal energy (kWh) 8453 (68%) 16923 (84%)
Electricity (kWh) 4000 (32%) 3070 (16%)
Total 12453 19993
In Figure 2 and 3 there is a comparison between rural and urban areas on how total energy
consumption is distributed for two countries: Greece and Latvia.
Page 5 of 13
17
Angelopoulos, M. K., & Pollalis, Y. A. (2021). Data Analytics to Improve Customer Energy Efficiency. Archives of Business Research, 9(6). 13-25.
URL: http://dx.doi.org/10.14738/abr.96.10290
Figure 2: Total Energy Consumption by end use and degree of urbanization in Greece
Figure 3: Total Energy Consumption by end use and degree of urbanization in Latvia
Profitable research findings are extracted by studying the correlation of socioeconomic
characteristics of household members with their energy consumption. In order to have a
complete view we compare counties with different climate behaviors (Greece and Latvia for
instance). The correlation shows that there are many common points:
• Household with older members consume more thermal energy and less energy for
electricity. Households with at least one member aged more than 65 years the thermal
consumption is 8-10% higher, compared to those with no 65 aged, and the electricity
consumption is reduced 15-17% respectively.
• When the average income is increased the consumption is increased too.
• Households with unemployed persons present higher electricity consumption and
lower thermal by percentages 14-16% and 10-12% respectively.
0
10
20
30
40
50
60
70
80
Space Heating Dometic Hot Water Cooking Appliances Other Use
Total Energy Consumption(%) by end use and degree of
urbanization
Urban Areas Rural Areas
0
10
20
30
40
50
60
70
80
Space Heating Dometic Hot Water Cooking Appliances Other Use
Total Energy Consumption(%) by end use and degree of
urbanization
Urban Areas Rural Areas
Page 6 of 13
18
Archives of Business Research (ABR) Vol. 9, Issue 6, June-2021
Services for Science and Education – United Kingdom
• The number of electrical appliances is analog to the consumption of the household. The
highest reduction in electricity consumption was achieved in private houses which use
heat pumps for heating purposes.
• The average consumption (both thermal and electricity) per person in one member
households is 1.5 times higher than in multi-person households.
• Consumption is increased by 8-12% when a member is added.
• In households with working members the average electricity consumption and the
average thermal energy consumption are higher by 30-32% and 12-15%, respectively,
compared to households without working members.
• Owned dwellings show double thermal energy consumption in contrast to rented ones
and slightly higher electricity consumption.
From the current surveys is implied that all countries exhibit similar consumption behavior
regarding the different socioeconomic and personal characteristics. The thing that differs is the
clean amount of energy consumed which depends on the climate conditions of each country.
BUILDINGS CLASSIFICATION ON BASIS OF ENERGY EFFICIENCY
Classification of the buildings on the basis of energy efficiency can be used to evaluate the
inefficient consumption units from the efficiently performing buildings. To test this
performance we visualize the available datasets for the hourly energy consumptions for each
building under study (Figure 4).
Figure 4: Average consumption for electricity and heating for each building (Finland)
The data are acquired with the following processing steps:
• Electricity and electricity used for heating are separated in the analysis.
• The average hourly consumption of all days for each building and each energy type is
calculated. Then, the average hourly consumption of all months are done in similar way.
• The data are normalized by taking the average of observations available for the hours
within a day and the days within the month.
0
10
20
30
40
50
60
70
80
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Average Consumption (Wh per m.sq)
Building under study
Heating Electricity
Page 7 of 13
19
Angelopoulos, M. K., & Pollalis, Y. A. (2021). Data Analytics to Improve Customer Energy Efficiency. Archives of Business Research, 9(6). 13-25.
URL: http://dx.doi.org/10.14738/abr.96.10290
It is easily occurred that the building itself affects the energy efficiency pattern. The buildings
are classified into four categories from poor to high efficiency (K value of 4). Then K-means
clustering is applied to on the input.
Kmeans clustering is an algorithm for cluster analysis. Is a method of vector quantization,
originally from signal processing, that aims to partition n observations into k clusters in which
each observation belongs to the cluster with the nearest mean (cluster centroid) [11]. In our
case the objects are energy efficiency values which are divided into predefined number of
classes. K depicts the number of clusters or groups that we can set in start of the process. There
are two prevailing methodologies: the Forgy [12] method and the Hartigan-Wong method [13].
Hereon the latter method is used.
Given a set of observations (x1, x2, ..., xn), where each observation is a d-dimensional real vector,
k-means clustering aims to partition the n observations into k (≤n) sets S = {S1, S2, ..., Sk} so as
to minimize the within-cluster sum of squares (WCSS/variance). Formally, the objective is to
find:
μi is the mean of points in Sι. This is equivalent to minimizing the pairwise squared deviations
of points in the same cluster:
The equation above used the Euclidean distance formula for calculating distance between
centroid and data point. The input to the K-means is a set of feature vectors along with the
number of necessary clusters. As written, we were required to classify pilot site buildings into
four groups (high-moderate-low-poor efficiency).
The Figure 5 [14] summarizes the results of the whole procedure. Each bubble on the graph
represents the hourly average energy efficiency for a particular building. The data are existing
for the 12 month period. The color of the bubble represents the energy class/cluster, while the
size of it goes along with the size of the building. Figure 6 [14] shows the one month subset of
the clustered values for January.
(1)
(2)
Page 8 of 13
20
Archives of Business Research (ABR) Vol. 9, Issue 6, June-2021
Services for Science and Education – United Kingdom
Figure 5: Clustering and average hourly energy efficiency for each month per each building.
Figure 6: Clustering for one month view
From the previous figures some useful insights are extracted. The bigger sized buildings are
more efficient in comparison with the smaller ones. The inefficient buildings can be a good
target for further study in order to improve their efficiency. Possible factors for poor
performance may be energy leakages, bad usage practices and inefficient equipment units. The
tactics followed by high performance buildings may be a guide for proper energy usage in the
lower performance ones.
Page 9 of 13
21
Angelopoulos, M. K., & Pollalis, Y. A. (2021). Data Analytics to Improve Customer Energy Efficiency. Archives of Business Research, 9(6). 13-25.
URL: http://dx.doi.org/10.14738/abr.96.10290
A useful approach could be to examine how the results of consumption per building are
differentiated from country to country. This is a difficult procedure for two main reasons.
Firstly, the buildings must appear similar characteristics regarding size, construction etc. in
order to be compared, and also information about consumption needs to be adequate on daily
basis and during all hours. To overcome those problems some extra mathematical techniques
are applied.
The lack of data or untrustworthy data from energy equipment is dealt with forecasting
methods. Estimating device-specific energy consumption has been a key focus area for the
energy providers. It helps in load forecasting and comprehend end user manners. The most
widespread methods are using auto regression in combination with moving averages in the
shape of a model known as ARIMA (Auto-regression Integrated Moving Averages) to estimate
consumption of a device given previous usage.
The concept behind regression is that we try to forecast a variable ‘y’ based on another variable
‘x’. A linear regression model is expressed in the equation below:
� = �! + �"� +ε (3)
where �! and �" represents the intercept and slope respectively for the line representing the
linear relationship.
Auto-regressive model is based on the scenario of a variable regressing on itself. Many observed
time series exhibit serial autocorrelation, that is a linear association between lagged
observations. This suggests past observations might predict current observations. The
autoregressive (AR) process models the conditional mean of yt as a function of past
observations, yt-1, yt-2,..., yt-p. The form of the model is:
yt= c+φ1 yt-1+...+φp yt-p+εt (4)
where ετ is an uncorrelated innovation process with mean zero.
The aim for good estimation is to select values of 0 and 1 that can minimize the sum of the
square of errors. The above equation can be used to estimate the value based on previous
values.
The autoregressive integrated moving average (ARIMA) process generates nonstationary
series that are integrated of order D, denoted I(D). A nonstationary I(D) process is one that can
be made stationary by taking D differences. Such processes are often called difference- stationary or unit root processes.
A series that you can model as a stationary ARMA(p,q) process after being differenced D times
is denoted by ARIMA(p,D,q). The form of the ARIMA(p,D,q) model is:
ΔDyt=c+φ1ΔDyt−1+...+φpΔDyt−p+εt+θ1εt−1+...+θqεt−q, (5)
where ΔDyt denotes a Dth differenced time series, and εt is an uncorrelated innovation process
with mean zero. More details about ARIMA are shown in [14] and [15].
Page 10 of 13
22
Archives of Business Research (ABR) Vol. 9, Issue 6, June-2021
Services for Science and Education – United Kingdom
Following the strategy described previously, simulated data are acquired for buildings with
similar characteristics (e.g. size) with those in Figure 4 for other two countries, and normalized
using the forecasting techniques. It is noticed that present similar energy consumption trend
but different values due to the climate that exists in each country (Figure 7 and Figure 8). As
explained in the previous parts, colder countries have increased energy requirements.
Figure 7: Average consumption for electricity and heating per building (Greece)
Figure 8: Average consumption for electricity and heating per building (Spain)
Another classification is according to the type of building (flat or private house). Surveys show
that there is no dependency from the type of building as the observations from several
households does not indicate continuity to one tendency. Nevertheless, all households appear
to have reduced consumption year after year (Table 3).
0
5
10
15
20
25
30
35
40
45
1 2 3 4 5 6 7 8 9 10
Average Consumption (Wh per m.sq)
Building under study
Heating Electricity
Page 11 of 13
23
Angelopoulos, M. K., & Pollalis, Y. A. (2021). Data Analytics to Improve Customer Energy Efficiency. Archives of Business Research, 9(6). 13-25.
URL: http://dx.doi.org/10.14738/abr.96.10290
Table 3 Consumption for different type of buildings in 2012 and 2013
Type of building Consumption(kWh) in
2012
Consumption(kWh) in
2013
Flat 281 205
Flat 770 680
Flat 100 88
Private house 607 570
Private house 568 555
Private house 791 551
FORECASTING CONSUMPTION IN HOUSEHOLDS AND BEHAVIORAL PRACTICES FOR
ENERGY EFFICIENCY
Recent techniques are implemented in order to monitor consumption patterns for the
households. Nonintrusive load monitoring (NILM), is a process for analyzing changes in the
voltage and current going into a house and deducing what appliances are used in the house as
well as their consumption [16]. Electric smart NIALM devices provide a continuous
communication between the user and the energy supplier, providing metering information at
regular intervals (order of minutes).
Several smart metering projects indicate that smart meters have great potential to decrease
energy use by 2-8% according to each country policies and manners. In most European
countries installation of smart meters is widely established. Italy and Finland use smart meters
under a strict legal framework. Latvia and Greece is still at the first steps of smart metering
development compared to other countries.
Mathematical prediction models are implemented using the data collected by the NIALM
devices. The datasets are from different appliances, some are in continuous use (freezer,
refrigerator) and some others are used when needed (tv, laundry machine, coffee maker). The
models used for the forecast devices which are in continuous usage.
The establishment of behavioral energy practices would lead to rational use of energy
consumed in the residential sector. The surveys study different disciplines in order to reduce
the energy usage of a household. Some are listed below:
1 Setting of thermostats on heating and cooling systems
2 Tents and shutters during the summer months
3 Use of washing machines with full load
4 Consult energy label before buying an electrical appliance
5 Open windows and roof openings for natural cooling
6 Proper maintenance of all appliances to be on optimal operation
CONCLUSIONS
Global energy needs are continuously rising. Energy efficiency has become a major concern for
minimizing the need for producing more energy. The consumers must understand the energy
patterns, identifying consumptions problems and establishing good energy practices. The
conducted surveys used different datasets that helps improving energy efficiency. There can
also be used alternative sources of data.
Page 12 of 13
24
Archives of Business Research (ABR) Vol. 9, Issue 6, June-2021
Services for Science and Education – United Kingdom
Ecological factors like climate and external temperatures affects the consumption of specific
energy types. The visualization of the average daily consumption of each building suggested the
respective use of the buildings. Optimizing the load of the buildings improves their energy
efficiency. The classification of buildings provides us view to identify possible faults, energy
leakages and inefficiencies and isolate those buildings.
A repetition of the survey of the selected households was conducted in order to clarify
behavioral factors that might influence changes in electricity consumption. The result was
encouraging as the consumption was reduced by 15%, but consumers said they don’t apply
saving measures or changed their consumption philosophy. That means that people do not
realize that they have taken some measures. The main factors for incentive to undertake
energy-saving actions is for saving money, spirit of competition and real examples on how to
save. Consumer’s attitude regarding smart meters is positive, but is not the main motivation to
decrease their consumption.
The operation of various systems and infrastructure is based on the use of electricity. Life
without electricity is infeasible. Unfortunately in practice even very high efficiency of electrical
equipment does not guarantee expected savings. It is due to the consumers’ behavioral aspects.
Inefficient use of electricity probably will lead to even more inevitable consequences in the
future, therefore increased attention should be paid how to find new innovative solutions for
increasing costumers motivation. Smart meters can play a significant role on promoting energy
saving in households. They are contributing not only to optimize the operation of electricity
systems, but also leading to socio-economic platform for the promotion interaction between
supplier and consumer.
References
[1]. Oikonomou V. Becchis F, Steg L, Russolillo . Energy saving and energy efficiency concepts for policy making. Energ
Pol 2009;37:4787-4796.
[2]. European Union. Green paper on an energy strategy for sustainable, competitive and secure energy.
COM(2006)0105; Brussels; 2006.
[3]. Streimikiene D, Šivickas G. The EU sustainable energy policy indicators framework. Environ Int 2008;34:1227–1240.
[4]. European Communities. Action Plan for Energy Efficiency: Realising the Potential. COM(2006)545 final; Brussels;
2006.
[5]. Hierzinger R, Albu M, van Elburg H, Scott A, Lazicki A, Penttinen L, Puente F, Sæle H. European Smart Metering
Landscape Report 2012.
[6]. Siano P. Demand response and smart grids – A survey. Renewable sustainable energy Reviews 2014;30:461–478.
[7]. Wissner M. The Smart Grid – A saucerful of secrets? Applied Energy 2011;88:2509–2518.
[8]. United Nations. World’s Population Increasingly Urban with More Than Half Living in Urban Areas. 2014. Available
online: https://www.un.org/en/development/desa/news/population/world-urbanization prospects-2014.html
(accessed on 18 January 2020).
[9]. European Commission. Energy Use in Buildings. Available online: https://ec.europa.eu/energy/en/eubuildings- factsheets-topics-tree/energy-use-buildings (accessed on 12 January 2020).
Page 13 of 13
25
Angelopoulos, M. K., & Pollalis, Y. A. (2021). Data Analytics to Improve Customer Energy Efficiency. Archives of Business Research, 9(6). 13-25.
URL: http://dx.doi.org/10.14738/abr.96.10290
[10]. United Nations Climate Change. What Is the Paris Agreement? Available online: https://unfccc.int/processand- meetings/the-paris-agreement/what-is-the-paris-agreement (accessed on 18 December 2019).
[11]. https://en.wikipedia.org/wiki/K-means_clustering
[12]. Forgy, E. W. Cluster analysis of multivariate data: efficiency versus interpretability of classifications. Biometrics 21
(1965), 768-769.
[13]. Hartigan, J. A., and Wong, M. A. Algorithm as 136: A k-means clustering algorithm. Applied statistics (1979), 100-
108.
[14]. Hussnain Ahmed,Applying Big Data Analytics for Energy Efficiency, Master's Thesis Espoo, August, 2014
[15]. https://www.mathworks.com/help/econ/arima-model.html
[16]. https://en.wikipedia.org/wiki/Nonintrusive_load_monitoring