ChapterPDF Available

Abstract and Figures

Home electricity demand has increased uninterrupted and is expected in 2050 to doubles the demanded in 2010. Making reasonable use of electricity is increasingly important and, in that way, different policies are carried out based on knowledge of how it is used. This article presents a procedure for measuring the potential electricity consumption in Uruguay. The study takes as main axis the appliance ownership information revelled by a national survey about severe socioeconomic aspects, and combines it with data on the characteristics of appliances, collected from local shops with an internet presence. Based on this data, an index of potential electricity consumption is performed for different census areas. To validate the analysis, it uses electricity consumption data from the ECD-UY (Electricity Consumption Data set of UruguaY) dataset and performs OLS linear regressions to evaluate real consumption and index correlation. The implementation uses Jupyter notebooks, language Python version 3, and utils libraries such as Pandas and Numpy. Results indicate that the departments with the highest index score are located on the West/Southwest coastlines. About census sections and segments in Montevideo, results show that the highest score areas are located in the South/Southeast coastlines, while lowest score ones are located in the outskirts. The validation process was limited by the lack of real consumption data.
Content may be subject to copyright.
Analysis of residential electricity consumption
by areas in Uruguay
Juan Chavat[0000000199252651] and Sergio Nesmachnow[0000000281464012]
Universidad de la Rep´ublica, Montevideo, Uruguay,
{juan.pablo.chavat,sergion}@fing.edu.uy
Abstract. Home electricity demand has increased uninterrupted and is
expected in 2050 to doubles the demanded in 2010. Making reasonable
use of electricity is increasingly important and, in that way, dierent poli-
cies are carried out based on knowledge of how it is used. This article
presents a procedure for measuring the potential electricity consumption
in Uruguay. The study takes as main axis the appliance ownership infor-
mation revelled by a national survey about severe socioeconomic aspects,
and combines it with data on the characteristics of appliances, collected
from local shops with an internet presence. Based on this data, an index
of potential electricity consumption is performed for dierent census ar-
eas. To validate the analysis, it uses electricity consumption data from
the ECD-UY (Electricity Consumption Data set of UruguaY) dataset
and performs OLS linear regressions to evaluate real consumption and
index correlation. The implementation uses Jupyter notebooks, language
Python version 3, and utils libraries such as Pandas and Numpy. Results
indicate that the departments with the highest index score are located
on the West/Southwest coastlines. About census sections and segments
in Montevideo, results show that the highest score areas are located in
the South/Southeast coastlines, while lowest score ones are located in
the outskirts. The validation process was limited by the lack of real con-
sumption data.
1 Introduction
Word Energy Outlook report, by the International Energy Agency [6], states
that residential electricity demand has increased uninterrupted worldwide. It is
expected to be double in 2050 than what it was in 2010 [7]. For that reason, it
is important to make responsible use of electricity. In that way, multiple inves-
tigations have been carried on with the purpose of apply policies that motivate
saving and reducing climate impact in factories, buildings, and homes [3,5,9, 12].
The population of Uruguay is 3.4 million inhabitants. Electricity in country
is provided by the state-owned company, UTE. In 2020, the company provides
electricity to a total of 1,498,164 customers throughout Uruguay, where 1,355,995
(90.5%) are residential customers. About 1.5 million people live in the capital
city, Montevideo. The city presents an electrification rate of 99.8%, including
urban and rural areas, according to data of 2018. In average, and according
Proceedings of the III Ibero-American Conference on Smart Cities (ICSC-2020)
Page 137
ISBN 978-9930-541-79-1
2J.ChavatandS.Nesmachnow
to 2017 stats, UTE serves per month 246 kWh to each residential customer in
Montevideo.
Energy consumption data analysis and characterization are needed to apply
demand management techniques oriented to a better use of energy resources. A
possible approach for demand management is to motivate behavioral changes in
customers that lead to electricity savings. Data analysis provides precise infor-
mation on how customers consume electricity, which can be used to elaborate
eective policies to consider for promotion of behavioral changes, elaboration of
new plans and taris, etc.
In this line of work, this article presents an index of residential electricity
consumption based on statistics about appliance ownership. The data of the
2017 National Continuous Household Survey (ECH, by its Spanish acronym) is
used in combination with appliance characteristics information collected from
local shops with a presence on the internet. The index is calculated to three
census area levels: by departments, by sections and by segments. Real electricity
consumption data, gathered from a subset of the ECD-UY dataset [2], is finally
used to validate the correlation of the index results with real data.
The study applies a data analysis approach [10] over appliance ownership
statistics, together with appliance characteristics information, to evaluate the
potential electricity consumption by census area. Also, a validation method is
proposed for the index, using real consumption data.
Results show that the departments with the highest index score are located on
the West/Southwest coastlines, while the ones with the lowest scores are located
in the East of Uruguay. Regarding census sections and segments in Montevideo,
results show that the highest score areas are located in the South/Southeast
coastlines of the city, decreasing progressively as it approaches the outskirts
of the city. Score at the segment level shows great dierences, up to six times,
between the highest and lowest extreme. The index validation process was limited
by the lack of more real consumption data.
This work is developed in the context of the project “Computational intelli-
gence to characterize the use of electric energy in residential customers”, funded
by the National Administration of Power Plants and Electrical Transmissions
(Spanish: Administraci´on Nacional de Usinas y Trasmisiones El´ectricas, UTE),
and Universidad de la Rep´ublica, Uruguay. The project study how computational
intelligence techniques can be used to process household electricity consumption
data and characterize energy consumption. It also focuses on determining which
appliances have the greatest impact on household electricity consumption and
in the identification of patterns in residential consumption.
The article is structured as follows. Next section describes the problem of an-
alyzing residential electricity consumption and reviews the main related work.
The proposed approach for analyzing the electricity consumption in Uruguay
is described in Section 3. The datasets and the processing are described in Sec-
tion 4. The main results are reported and analyzed in Section 5. Finally, Section 6
presents the conclusions and the main lines of future work.
Proceedings of the III Ibero-American Conference on Smart Cities (ICSC-2020)
Page 138
ISBN 978-9930-541-79-1
Analysis of residential electricity consumption by areas in Uruguay 3
2 Analysis of residential electricity consumption
This section describes the problem addressed in this article and reviews relevant
related works.
2.1 Main research question and hypothesis
This work analyzes the electricity consumption based on an index built from ap-
pliance ownership statistics. The statistics were obtained from a national survey
implemented year by year by the National Statistics Institute (INE), Uruguay.
The formulated question is: Can an index build from appliances ownership statis-
tics model the electricity consumption per census area in Uruguay?
Some energy-intensive appliances, such as air conditioner or electric water
heater, determine the electric consumption of a household. Some of these appli-
ances are not a basic need, and therefore not every household count with them. If
a degree of the appliances ownership is calculated for census areas, is expected to
determine an average level of electricity consumption. From the previous ques-
tion and considerations, and based on intuitive ideas, the following hypothesis
was formulated to work on it.
Hypothesis: The more energy-intensive appliances owned, the higher the po-
tential electricity consumption.
2.2 Related works
The analysis of the related literature allowed identifying several approaches for
electricity consumption characterization in several countries. Most of the ap-
proaches have applied statistical tools (e.g., multilevel and logistic regression),
such as in this article. Some relevant related works are reviewed next.
Ch´evez et al. studied the electricity consumption in Great La Plata, Ar-
gentina [4]. Two relevant problems of the Argentinean electricity sector were
identified: i) consumption peaks, that increased 5% per year, could not be sat-
isfied, and ii) a poor diversification of the electricity generation matrix. 1010
census areas with similar electricity consumption were identified and clustered
in eight groups applying the k-means algorithm. Results were related to socio-
demographic variables and its relevance in electricity consumption was studied.
The article concluded that electricity demands grow quickly as the ratio of peo-
ple per home and people per room increases. The greater the presence of flats in
the area, the lower the electricity consumption. In turn, the more precarious the
buildings, the greater the electricity consumption. Concerning unsatisfied basic
needs, at higher the index level, proportionally higher is the electricity demand.
McLoughlin et al. [11] analyzed energy consumption data from 3941 smart
meters in Ireland, and socio-economic, demographic, and dwelling characteris-
tics. Four parameters were considered in the study: i) total electricity consumed,
ii) maximum demand, iii) load factor (the lower, the more ”peaky” of the con-
sumption), and iv) time of use of maximum electricity consumption. Linear
Proceedings of the III Ibero-American Conference on Smart Cities (ICSC-2020)
Page 139
ISBN 978-9930-541-79-1
4J.ChavatandS.Nesmachnow
regression algorithms were developed to study how the dwelling/occupant char-
acteristics and how the owned appliances aect on the electricity consumption.
The analysis concluded that electricity consumption was negatively influenced
by a higher number of bedrooms, head of households between 36–55 years, and
a higher presence of professionals. On the other hand, it is positively influenced
by dwelling type apartments and lower/middle social classes. About appliances,
households using electricity for water heating or cooking consumed more elec-
tricity than the rest. Load factor, a measure of daily mean to daily maximum
electricity demand, was sensible to the dwelling type and the number of bed-
rooms. Time length of maximum demand is more by the number of occupants
than the dwelling type. It occurs during the morning for older heads of house-
holds, and late in the day for middle age heads of households.
Anderson et al. [1] explored inferring household characteristics of census ar-
eas from electricity consumption and number of residents for Ireland too. Data
was limited to three days (Tuesday, Wednesday, and Thursday) of October 2009.
Indicators were generated to describe household electricity consumption, consid-
ering load magnitude, summary statistics, and temporal properties. First, house-
hold characteristics were identified to infer profile indicators, applying multilevel
regression considering several explanatory variables: income, employment status
of the household response person (HRP), presence of children, and the number
of residents. Then, the most likely profile indicators to reverse the direction of
the prediction model were selected by logistic regression. Results showed an ac-
curacy close to 60% to classify the employment status of the HRP. The work
concluded that, despite the accuracy achieved, it is a feasible approach to infer
household characteristics from the electricity consumption profiles.
Villareal and Moreira [13] studied residential electricity consumption in Brazil
in 1985–2013. Residential consumption represented 26% of the electricity used
in country, and the most demanding appliances were electric shower (19%), re-
frigerator (18%), lamps (15%), TV (11%), air conditioning and freezer (5%).
Elasticity values were obtained from processing explanatory variables into linear
regressions, and used to relate variables to consumption behaviours. The follow
variables were used for the analysis: number of households on the country, avail-
able family income, electricity taris, appliances ownership, and social/economic
policies that aect electricity consumption directly. About extra factors, the fol-
lowing three social policies were chosen: restraining of electricity consumption,
facilitate access to electricity for low incoming families, and energy eciency pro-
grams. Three models were developed to describe the consumption: i) considering
variables represented by time series only, ii) considering electricity restraints, and
iii) considering all the extra factors. Authors concluded that a rise of 1% in the
number of residences increases electricity consumption by 1.53%, a rises of 1% in
family income increase consumption in 0.19%, and a rise of 1% in the taricause
a decrease of 0.23% in the consumption. The models presented high coecients
of determination (0.968 the first, 0.989 the rest), showing a strong relationship
between explanatory variables and electricity consumption.
Proceedings of the III Ibero-American Conference on Smart Cities (ICSC-2020)
Page 140
ISBN 978-9930-541-79-1
Analysis of residential electricity consumption by areas in Uruguay 5
In Uruguay, Laureiro [8] analyzed residential electricity consumption based
on socioeconomic characteristics, dwelling characteristics, energy uses, and tem-
perature. Ordinary Least Square (OLS) and Quantile Regression (QR) were
applied on data from 2994 houses. A cursory analysis yielded that income per
capita is a relevant factor but not the unique, owning certain appliances (electric
water heater/air conditioner) directly impacts over electricity consumption, and
thermal comfort appliances are more common in dwellings with high electricity
consumption. The OLS analysis concluded that: i) per capita income has high
elasticity, ii) an increment of 1% in the square meters of a dwelling, increments
0.06% the electricity consumption, iii) houses consume 10.8% more than apart-
ments, iv) electricity consumption increases 8.2% for each extra air conditioner
and 17.2% for each extra electric water heater, and v) regional variables do not
impact significantly in the consumption. The QR analysis concluded that: i) the
impact of income per capita over consumption is lower in high quartiles than in
low/medium ones, ii) dwelling size impact more in higher than in lower deciles
iii) the dwelling type impacts only in medium/high deciles while building mate-
rials do not impact at alliv) air conditioners impact more in lower deciles and
electric water heaters impact equally in all deciles, vi) the impact of cooking,
washing/dryer machines, and sanitary heating have an inverted ‘U’ behaviour
(low in extreme deciles, high in medium deciles). The work concluded that al-
though the income per capita is a determining variable, it is not the only one
that impacts on electricity consumption, and other characteristics must be take
into account (e.g., family composition, dwelling characteristics, and energy uses).
This article contributes by studying the electricity consumption based on
appliance ownership data processed from national surveys in Uruguay.
3 Proposed approach for electricity consumption analysis
The proposed approach for the analysis consists of building an index that scores
the electricity consumption degree, per census areas. For the construction, data
provided by the ECH national survey from the year 2017 is used. ECH counts
with several variables, described in the following section, that quantify the appli-
ances ownership of the households. The surveyed households have geo-referenced
information in at least three census levels: departments, census sections and cen-
sus segments. Further details about the census areas are provided next.
Data is grouped by census area and the likelihood of owning the surveyed
appliance is calculated. Besides, each appliances power consumption is collected
from many local shops to weight the impact of each appliance in the final value.
For example, owning an air conditioner aect more on electricity consumption
than owning a flat TV. In the same way, each appliance is categorized by its fre-
quency of use between low, medium or high. This represents a second weighting
on the appliance consumption impact over the final result. Therefore, a fridge
that is always on aects more than a notebook computer (sporadically used)
on the final results. Frequencies are assigned as a rule of thumb guided by the
authors own experience.
Proceedings of the III Ibero-American Conference on Smart Cities (ICSC-2020)
Page 141
ISBN 978-9930-541-79-1
6J.ChavatandS.Nesmachnow
The index scores are calculated as shown in Equation 1. Given a type of cen-
sus area rwith mdierent areas (e.g., m= 19 if r= departments), A(r)2Rmxn
a matrix with one row per census area and one column per appliance likelihood
information; c(r)2Rna vector with the consumption of the nappliances; and
f(r)2Rna vector with quantified frequency of use for the nappliances. The
result is a vector index (r)2Rmwhere each value in position imeans the index
score for the area i.
index (r)=A(r)
m,n ·c(r)·f(r)·2
6
4
1
.
.
.
1
3
7
5(1)
4 Data collection and processing
This section describes the data used for the analysis and how it was prepared to
be processed.
4.1 Census data
Used census data is provided by the National Institute of Statistics (INE, by its
Spanish acronym). INE collect data of dierent index with monthly, quarterly,
half-yearly and annual periodicity. The information is presented as a continuous
household survey (ECH, by its Spanish acronym) every year. The ECH collects
data about the labour market and income of households and individuals, from
a representative set of households distributed around the country.
Information in ECH is georeferenced by, at least, the department, the census
section, and the census segment. The definition of these georeferenced levels are
provided next:
Department: Coincides with the nineteen dierent political-administrative
borders of the country.
Census section: Corresponds to the first division level of the departments.
Each section area can be cut into blocks or not. Its borders coincide with
the ones used in the national census of 1963.
Census segment: The segments are the subdivision of the sections. In census
locations or areas cut into blocks, corresponds to a set of blocks, otherwise,
the segments are a portion of territory that groups minor units with recog-
nisable physical limits in the terrain and can include population centres.
Only a subset of the indexes in ECH was selected for the analysis. The se-
lected indexes focus on georeferencing the data and quantifying appliance own-
ership. Table 1 list detailed information about the selected indexes.
Proceedings of the III Ibero-American Conference on Smart Cities (ICSC-2020)
Page 142
ISBN 978-9930-541-79-1
Analysis of residential electricity consumption by areas in Uruguay 7
Table 1: Description of data from ECH (2017) used to build the index
name description type of value
dpto Code of the department Number (1-19)
nomdpto Name of the department String
secc Census section Number
segm Census segment Number
nombarrio Name of the neighbourhood String
d9 Number of residential rooms Number
d18 Energy source for lighting Number (1: electric; 2-4: other)
d260 Energy source for heating Number (1: electric; 2-6: other)
d20 Energy source for cooking Number (1: electric; 2-6: other)
d21 1 Electric water heater Number (1: yes; 2: no)
d21 2Showerwaterheater Number (1: yes; 2: no)
d21 3 Fridge Number (1: yes; 2: no)
d21 4TubeTV Number (1: yes; 2: no)
d21 41NumberoftubeTVs Number
d21 5 LCD/Plasma TV Number (1: yes; 2: no)
d21 5 1 Number of LCD/Plasma TV Number
d21 6Radio Number(1:yes;2:no)
d21 8videocassetteplayer Number (1: yes; 2: no)
d21 9DVDplayer Number (1: yes; 2: no)
d21 10 Washing machine Number (1: yes; 2: no)
d21 11 Clothes dryer Number (1: yes; 2: no)
d21 12 Dishwasher Number (1: yes; 2: no)
d21 13 Microwave Number (1: yes; 2: no)
d21 14 Air conditioner Number (1: yes; 2: no)
d21 14 1 Number of air conditioners Number
d21 15 Notebook computer Number (1: yes; 2: no)
d21 15 2 ‘Plan Ceibal’ laptops Number
d21 15 4OtherNotebooks Number
Data preparation. Preliminary analysis showed that records outside Montevideo
do not have census section nor segment values set. Therefore, the index for these
areas is evaluated only for Montevideo. The Yes/No columns were transformed
from {1,2}values to {0,1}to facilitate the multiplication by the columns that
indicate the number of appliances. Additionally, columns with common and ‘Plan
Ceibal’ laptops were merged into one with the sum of both and the lighting
columns was multiplied by the number of residential rooms to represents a light
per room. Also, to discriminate between the air conditioner and other electric
heating sources, the column indicating source was set to ‘No’ if the column
of the air conditioners has a ‘Yes’ value. The final transformation consisted of
multiplying all the columns that indicate the presence of an appliance by the
corresponding column that indicates the number of appliances in the household.
Finally, several validations were processed to assure the integrity of the in-
formation. For example, columns that indicate the number of an appliance in a
household were checked that if the value is greater than zero, then the column
indicating the presence of this appliance have the corresponding ”Yes” value. No
integrity errors were found in this last step.
Proceedings of the III Ibero-American Conference on Smart Cities (ICSC-2020)
Page 143
ISBN 978-9930-541-79-1
8J.ChavatandS.Nesmachnow
4.2 Appliance characteristics data
ECH surveys gather data about the ownership of certain household appliances.
Based on these appliances and using the information of local shops with presence
on the Internet, power consumption data was collected. Up to five dierent
appliance models were gathered to define the median power consumption of
each appliance. Table 2 lists the result of the data collection process, and Fig. 1
presents a bar graph of the mean power consumption together with its standard
deviation measure. It can be observed how some appliances are more energy-
intensive than others.
Table 2: List of appliances information used to build the index
appliance Mean power (W) Frequency
of use
Power weighted
by frequency of use
lighting 11.8 medium 8.85
heating 1200.0 high 1200.00
oven 1380.0 high 1380.00
electric water heater 1600.0 high 1600.00
shower heater 1810.0 medium 1357.50
fridge 199.4 high 199.40
tube TV 124.8 medium 93.60
flat TV 85.6 medium 64.20
radio 20.2 low 10.10
VHS player 10.0 low 5.00
DVD player 10.5 low 5.25
washing machine 740.0 medium 555.00
clothes dryer 3154.0 medium 2365.50
dishwasher 1409.6 medium 1057.20
microwave 1068.0 medium 801.00
air conditioner 1290.0 high 1290.00
notebook 57.0 medium 42.75
Fig. 1: Electricity consumption of the appliances used to build the index
Proceedings of the III Ibero-American Conference on Smart Cities (ICSC-2020)
Page 144
ISBN 978-9930-541-79-1
Analysis of residential electricity consumption by areas in Uruguay 9
4.3 Electricity consumption data
Real electricity consumption data is gathered from the ECD-UY dataset [2] and
corresponds to the electric water heater consumption subset. The records origi-
nate from dierent clamp/meters installed by the National Electricity Company
(UTE) in the households of their customers. The subset consists of mainly two
parts, the records of the appliance consumption disaggregated and the total ag-
gregated household consumption. The location of the households varies among
the main Uruguayan cities. For this work, only the total aggregated consumption
and the customer georeferenced information were used.
The subset contains the total consumption of 541 households on which only
242 are georeferenced. These georeferenced households are distributed into 6
departments. Household located in Montevideo are located along 12 census sec-
tions and 38 census segments. Fig. 2 shows three maps at dierent area level. The
marked areas in each map correspond to those for which electricity consumption
data is available in the ECD-UY dataset.
(a) Departments of Uruguay
(b) Census sections (Montevideo)
(c) Census segments (Montevideo)
Fig. 2: Maps where the marked areas represents the one that counts with real
electricity consumption data in the ECD-UY subset
The data preparation phase consisted of two steps. First, the electricity con-
sumption of customers without georeferenced data was removed from the subset.
Then, abnormal consumption values were filtered. For this task, the records with
values lower than the 15th percentile and greater than the 85th percentile were
removed.
Proceedings of the III Ibero-American Conference on Smart Cities (ICSC-2020)
Page 145
ISBN 978-9930-541-79-1
10 J. Chavat and S. Nesmachnow
4.4 Implementation
The implementation consists of the next main steps: the load of the datasets,
matrices construction for power demand by appliance and appliance ownership
likelihood, processing of the index score per census area, visualization of results,
and validation of the index scores.
The processing was executed on a personal computer with average processing
power. Code was implemented in a Jupyter notebook using Python language
version 3. For the data loading and the matrices construction, the utility libraries
Pandas and Numpy were used, and the GeoPandas extension was applied for
maps generation. The resulting notebook with its processing results is available
for download at https://bit.ly/3kGUfVO.
5 Results
This section present first the result of the proposed analysis by the three dier-
ent census areas and then the results on the validation of the data using real
consumption records.
5.1 Index scores by census areas
Results of the index by department areas show a dierence up to 65% between
the first and the last position. The department that results with the highest
score is Montevideo, while the one with the lowest score was Cerro Largo. In
general, the departments that present higher scores index are located on the
west and south-west coastlines. A visual inspection of the departments in the
Uruguayan map, starting from Colonia at the most southwest and pointing to
the northeast, shows a progressive increase of the index score. That is observed
at the map presented in Fig. 3. The complete list of departments together with
its resulting index score is shown in the Table 3
Table 3: Index score by departments
score department
4505.8 Montevideo
4139.5 Colonia
4097.9 Salto
4008.6 Maldonado
3964.9 Paysandu
3915.5 Rio Negro
3894.8 Soriano
3821.7 Canelones
3804.3 Artigas
3792.1 San Jose
score department
3725.7 Flores
3695.0 Florida
3559.0 Durazno
3543.8 Lavalleja
3427.0 Rivera
3385.0 Treinta Y Tres
3373.8 Rocha
3322.9 Tacuarembo
2950.7 Cerro Largo
Proceedings of the III Ibero-American Conference on Smart Cities (ICSC-2020)
Page 146
ISBN 978-9930-541-79-1
Analysis of residential electricity consumption by areas in Uruguay 11
Fig. 3: Index scores calculated for departments of Uruguay
Results corresponding to the index score calculated by census section of Mon-
tevideo shows that the highest index score sections are located beside the south-
east coastline of the city. The dierence between the highest and the lowest index
scores is up to 66%. The section with the highest score is number 10, and it cov-
ers the neighbourhoods Carrasco Norte, Buceo, Malvin, Malvin Norte, Punta
Gorda, Union, Las Canteras and Carrasco. In the opposite side, the section
with the lowest score is number 16, it covers the neighbourhoods Tres Ombues,
Victoria, Nuevo Paris, Paso de la Arena, Casab´o, and Pajas Blancas. A visual
inspection on the map shown in Fig. 4 shows that starting from the southeast
coastline and pointing to the northwest, the index score decrease progressively.
Table 4 list the resulting scores by census section, together with the list of cor-
responding neighbourhoods of each section.
Finally, the index calculated by census segments shows an accumulation of
segments with highest scores on the south and southeast area of Montevideo,
while lowest scores segments are located mainly in the outskirts of the city.
Fig. 5 shows a map of census segments in Montevideo, coloured by its index
score. Results also reveal a big dierence in the score among top and bottom
scored segments, diering by more than six times in the most extreme cases.
Table 5 shows a truncated list of the census segments ordered by its resulting
index score.
Proceedings of the III Ibero-American Conference on Smart Cities (ICSC-2020)
Page 147
ISBN 978-9930-541-79-1
12 J. Chavat and S. Nesmachnow
Table 4: Index score by census sections
score section neighbourhoods
5444.1 10 Carrasco Norte, Buceo, Malvin, Malvin Norte,
Punta Gorda, Union, Las Canteras, Carrasco
5297.3 18 Punta Carretas, Pocitos, Cordon, Tres Cruces,
Parque Batlle, V. Dolores, Parque Rodo
5284.1 24 Pocitos, Pque. Batlle, Villa Dolores, Buceo
4981.8 14 Prado, Nueva Savona, Reducto, Capurro, Bella Vista
4898.2 23 Tres Cruces, La Blanqueada, Larra˜naga
4795.5 6 Centro (Norte)
4710.5 12 Reducto, Atahualpa, La Figurita, Jacinto Vera,
Larra˜naga, Brazo Oriental, Mercado Modelo, Bolivar
4549.0 15 Cordon, Palermo, Parque Rodo
4432.7 7 Cordon, Palermo
4428.4 4 Centro (Suroeste), Ciudad Vieja (Sureste), Barrio Sur
4414.0 5 Centro (Sur), Barrio Sur
4358.1 21 Pe˜narol, Lavalleja, Conciliacion, Sayago, Nuevo Paris
Paso de las Duranas, Belvedere
4322.9 22 Cerrito, Brazo Oriental, Villa Espa˜nola, Bolivar,
Mercado Modelo, Castro, P. Castellanos
4313.7 8 Aguada
4305.6 19 La Comercial, Villa Mu˜noz, Retiro
4211.6 20 Aires Puros, La Teja, Prado, Nueva Savona, Belvedere,
Nuevo Paris
4162.1 3 Ciudad Vieja (Sur)
4143.9 1 Ciudad Vieja (Noreste), Centro
3907.3 13 Casabo, Pajas Blancas, Paso de la Arena, La Paloma,
Tomkinson, Cerro
3867.4 9 Colon Centro y Noroeste, Colon Sureste, Abayuba,
Lezica, Melilla
3842.9 99 Flor de Maro˜nas, Maro˜nas, Parque Guarani, Union
Ba˜nados de Carrasco, Villa Garcia, Manga Rural,
Punta Rieles, Bella Italia, Las Canteras
3693.1 17 Casavalle, Manga, Las Acacias, Villa Espa˜nola, Piedras
Blancas, Castro, P. Castellanos, Manga, Toledo Chico
3663.9 11 Ituzaingo, Jardines del Hipodromo, Flor de Maro˜nas,
P. Ri e l e s , B e l l a I t a lia, Manga, Toled o C h i c o , M a n g a ,
Piedras Blancas, Villa Garcia, Villa Espa˜nola, Union
3621.2 2 Ciudad Vieja (Norte)
3616.8 16 Tres Ombues, Victoria, Nuevo Paris, Paso de la Arena
Casabo, Pajas Blancas
5.2 Validation of the proposed approach
For validating the proposed approach, the monthly average electricity consump-
tion is calculated and compared with the index score results. Finally, the average
consumption and the index scores are processed by an OLS linear regression to
measure the correlation between real consumption and the index score.
Proceedings of the III Ibero-American Conference on Smart Cities (ICSC-2020)
Page 148
ISBN 978-9930-541-79-1
Analysis of residential electricity consumption by areas in Uruguay 13
Fig. 4: Index score calculated by census sections of Montevideo
Table 5: Truncated list of index scores by census segment
score section segment neighbourhoods
9249.1 10 67 Carrasco
9120.0 10 246 Malv´ın
9020.2 10 75 Carrasco
9013.0 10 74 Carrasco
8687.8 10 64 Carrasco
... ... ... ...
2322.6 9 1 Leizica, Melilla
2235.4 2 3 Ciudad Vieja
2164.1 99 208 Ba˜nados de Carrasco
2141.0 13 9 Cerro
1450.4 13 113 Casab´o, Pajas Blancas
Fig. 6 shows the average monthly electricity consumption of departments
with available consumption data in the ECD-UY dataset. The calculation of
the consumption is based on 242 georeferenced customers unequally distributed
in departments. The comparison of the real consumption and the index score
results shows that only two of the six departments, Montevideo (72 customers)
and Paysand´u (152), are in the same order. These two departments are the ones
with more real data available, and therefore, its calculated average consumption
is more reliable than the rest (which accounts for just six or less customers each).
OLS was performed with average real consumption and index score data.
Results showed no correlation between the values but it should be taken into
account the low number of samples that conform the calculation of average
consumption. It is necessary to repeat the experiments with a more reliable
calculation of real average monthly consumption.
Proceedings of the III Ibero-American Conference on Smart Cities (ICSC-2020)
Page 149
ISBN 978-9930-541-79-1
14 J. Chavat and S. Nesmachnow
Fig. 5: Index score calculated by census segments of Montevideo
Fig. 6: Monthly average consumption from real data in ECD-UY dataset
Regarding the validation of the index by census section/census segment, the
available real data is even more limited. For census sections, the available data
is from 72 customers distributed in 12 sections, meaning an average ratio of 6
customers per section. For census segments, the 72 customers are distributed in
38 segments and its average ratio is lower than 2. Both cases are not statistically
representative, and therefore, a full validation is not possible in this case.
6 Conclusions and future work
This article presented an electricity consumption analysis based on household
appliance ownership data processed from national surveys in Uruguay. The pro-
posed approach consists of building an index to score the potential electricity con-
sumption in three dierent census areas: departments, sections, and segments.
Proceedings of the III Ibero-American Conference on Smart Cities (ICSC-2020)
Page 150
ISBN 978-9930-541-79-1
Analysis of residential electricity consumption by areas in Uruguay 15
Appliance information was collected from several sources and processed to
build the proposed index. Results showed that departments located the South-
eastern coastline have the highest index scores while departments in the north
and northeast have the lowest ones. Regarding census sections and segments, the
index was performed only for Montevideo and results show the highest scores
in the south/southeast coastlines and lowest scores in the outskirts of the city.
Census segments show great dierences, up to six times, between extreme values.
Validation of the index was limited by the lack of more real data.
The main lines for future work are related to study the relationship between
the index and socioeconomic variables provided by the ECH, such as household
incoming, education level, number of kids, among others. Also, updating the
index to consider data from last and previous ECH versions, to study the index
evolution along time.
References
1. Anderson, B., Lin, S., Newing, A., Bahaj, A.B., James, P.: Electricity consumption
and household characteristics: Implications for census-taking in a smart metered
future. Computers, Environment and Urban Systems 63, 58–67 (2017)
2. Chavat, J., Graneri, J., Alvez, G., Nesmachnow, S.: ECD-UY: Detailed household
electricity consumption dataset of uruguay. Scientific Data (2020), (submitted)
3. Chavat, J., Nesmachnow, S., Graneri, J.: Non-intrusive energy disaggregation by
detecting similarities in consumption patterns. Revista Facultad de Ingenier´ıa Uni-
versidad de Antioquia (2020)
4. Ch´evez, P., Barbero, D., Martini, I., Discoli, C.: Application of the k-means clus-
tering method for the detection and analysis of areas of homogeneous residential
electricity consumption at the Great La Plata region, Buenos Aires, Argentina.
Sustainable Cities and Society 32, 115–129 (2017)
5. Ford, R.: Reducing domestic energy consumption through behaviour modification.
Ph.D. thesis, Oxford University (2009)
6. International Energy Agency: World Energy Outlook 2015. White paper (2015)
7. Larcher, D., Tarascon, J.: Towards greener and more sustainable batteries for elec-
trical energy storage. Nature Chemistry 7(1), 19–29 (2015)
8. Laureiro, P.: Determinantes del consumo de energ´ıa el´ectrica del sector residencial
en Uruguay. Serie Documentos de investigaci´on estudiantil, DIE 05/18 FCS, Udela
(2018)
9. Luj´an, E., Otero, A., Valenzuela, S., Mocskos, E., Steenel, L., Nesmachnow, S.: An
integrated platform for smart energy management: the CC-SEM project. Revista
Facultad de Ingenier´ıa Universidad de Antioquia (2019)
10. Massobrio, R., Nesmachnow, S., Tchernykh, A., Avetisyan, A., Radchenko, G.:
Towards a cloud computing paradigm for big data analysis in smart cities. Pro-
gramming and Computer Software 44(3), 181–189 (2018)
11. McLoughlin, F., Duy, A., Conlon, M.: Characterising domestic electricity con-
sumption patterns by dwelling and occupant socio-economic variables: An Irish
case study. Energy and Buildings 48(July 2009), 240–248 (2012)
12. Orsi, E., Nesmachnow, S.: Smart home energy planning using IoT and the cloud.
In: IEEE URUCON (2017)
13. Villareal, M., Moreira, J.: Household consumption of electricity in Brazil between
1985 and 2013. Energy Policy 96, 251–259 (2016)
Proceedings of the III Ibero-American Conference on Smart Cities (ICSC-2020)
Page 151
ISBN 978-9930-541-79-1
... Beyond the main differences, it should be noted that Chavat and Nesmachnow's works 23,24 used variables for the construction of their model that can be catalogued as relevant ones. These variables were: information about appliance ownership, georeferentiation, and the number of rooms per dwelling. ...
Article
Given the massive deployment of smart meters at international level, it is necessary to develop methodologies to extract knowledge from the data that they can provide. To this end, it is necessary to associate energy, socio-demographic and/or technical-constructive data, because this is the only way to identify profiles with their corresponding relevant variables or drivers. The usual problem is that socio-technical information about users is limited or non-existent, as it is costly to collect. Consequently, this work presents as a novelty the use of census information to characterize groups of urban segments with similar daily electricity load curves, which avoids the need to collect socio-technical information through specific surveys or direct measurements. In this way, relevant variables are identified in the determination of consumption patterns in the study case (Montevideo-Uruguay) and they are used to infer the daily behavior of those sectors of the city that don’t have this information.
... In turn, data gathered in ECD-UY is also very valuable for the electricity company, in order to study and analyze electricity consumption patterns of citizens, relating the consumption with relevant socio-demographic data and indicators 27,28 , the design of personalized electricity billing plans for different segments of the population, and the study of specific interventions to influence on the users' behavior to achieve a rational utilization of www.nature.com/scientificdata www.nature.com/scientificdata/ the electric resources, among others relevant issues related to the intelligent utilization of electricity in modern smart cities. ...
Article
Full-text available
This article introduces a dataset containing electricity consumption records of residential households in Uruguay (mostly in Montevideo). The dataset is conceived to analyze customer behavior and detect patterns of energy consumption that can help to improve the service. The dataset is conformed by three subsets that cover total household consumption, electric water heater consumption, and by-appliance electricity consumption, with sample intervals from one to fifteen minutes. The datetime ranges of the recorded consumptions vary depending on the subset, from some weeks long to some years long. The data was collected by the Uruguayan electricity company (UTE) and studied by Universidad de la República. The presented dataset is a valuable input for researchers in the study of energy consumption patterns, energy disaggregation, the design of energy billing plans, among other relevant issues related to the intelligent utilization of energy in modern smart cities.
Chapter
This article presents a study to characterize the electricity consumption in residential buildings in Uruguay. Understanding residential electricity consumption is a relevant concept to identify factors that influence electricity usage, and allows developing specific and custom energy efficiency policies. The study focuses on two home appliances: air conditioner and water heater, which represents a large share of the electricity consumption of Uruguayan households. A data-analysis approach is applied to process several data sources and compute relevant indicators. Statistical methods are applied to study the relationships between different relevant variables, including appliance ownership, average income of households, and temperature, and the residential electricity consumption. A specific application of the data analysis is presented: a regression model to determine the consumption patterns of water heaters in households. Results show that the proposed approach is able to compute good values for precision, recall and F1-score and an excellent value for accuracy (0.92). These results are very promising for conducting an economic analysis that takes into account the investment cost of remotely controlling water heaters and the benefits derived from managing their demand.
Article
Full-text available
This article introduces a dataset containing electricity consumption records of residential households in Uruguay (mostly in Montevideo). The dataset is conceived to analyze customer behavior and detect patterns of energy consumption that can help to improve the service. The dataset is conformed by three subsets that cover total household consumption, electric water heater consumption, and by-appliance electricity consumption, with sample intervals from one to fifteen minutes. The datetime ranges of the recorded consumptions vary depending on the subset, from some weeks long to some years long. The data was collected by the Uruguayan electricity company (UTE) and studied by Universidad de la República. The presented dataset is a valuable input for researchers in the study of energy consumption patterns, energy disaggregation, the design of energy billing plans, among other relevant issues related to the intelligent utilization of energy in modern smart cities.
Article
Full-text available
Breaking down the aggregated energy consumption into a detailed consumption per appliance is a crucial tool for energy efficiency in residential buildings. Non-intrusive load monitoring allows implementing this strategy using just a smart energy meter without installing extra hardware. The obtained information is critical to provide an accurate characterization of energy consumption in order to avoid an overload of the electric system, and also to elaborate special tariffs to reduce the electricity cost for users. This article presents an approach for energy consumption disaggregation in households, based on detecting similar consumption patterns from previously recorded labelled datasets. The experimental evaluation of the proposed method is performed over four different problem instances that model real household scenarios using data from an energy consumption repository. Experimental results are compared with twobuilt-in algorithms provided by the nilmtk framework (combinatorial optimization and factorial hidden Markov model). The proposed algorithm was able to achieve accurate results regarding standard prediction metrics. The accuracy was not affected in a significant manner by the presence of ambiguity between the energy consumption of different appliances or by the difference of consumption between training and test appliances.
Article
Full-text available
This article describes the Cloud Computing for Smart Energy Management (CC-SEM) project, a research effort focused on building an integrated platform for smart monitoring, controlling, and planning energy consumption and generation in urban scenarios. The project integrates cutting-edge technologies (Big Data analysis, computational intelligence, Internet of Things, High-Performance Computing and Cloud Computing), specific hardware for energy monitoring/controlling built within the project and efficient communication protocols. The proposed platform considers the point of view of both citizens and administrators, providing a set of tools for controlling home devices (for end-users), lanning/simulating scenarios of energy generation (for energy companies and administrators), and proposes advances in communication infrastructure for transmitting the generated data.
Article
Full-text available
In this paper, we present a Big Data analysis paradigm related to smart cities using cloud computing infrastructures. The proposed architecture follows the MapReduce parallel model implemented using the Hadoop framework. We analyse two case studies: a quality-of-service assessment of public transportation system using historical bus location data, and a passenger-mobility estimation using ticket sales data from smartcards. Both case studies use real data from the transportation system of Montevideo, Uruguay. The experimental evaluation demonstrates that the proposed model allows processing large volumes of data efficiently.
Article
Full-text available
This paper assesses the feasibility of determining key household characteristics based on temporal load profiles of household electricity demand. It is known that household characteristics, behaviours and routines drive a number of features of household electricity loads in ways which are currently not fully understood. The roll out of domestic smart meters in the UK and elsewhere could enable better understanding through the collection of high temporal resolution electricity monitoring data at the household level. Such data affords tremendous potential to invert the established relationship between household characteristics and temporal load profiles. Rather than use household characteristics as a predictor of loads, observed electricity load profiles, or indicators based on them, could instead be used to impute household characteristics. These micro level imputed characteristics could then be aggregated at the small area level to produce ‘census-like’ small area indicators. This work briefly reviews the nature of current and future census taking in the UK before outlining the household characteristics that are to be found in the UK census and which are also known to influence electricity load profiles. It then presents descriptive analysis of two smart meter-like datasets of half-hourly domestic electricity consumption before reporting on the results from a multilevel modelling-based analysis of the same data. The work concludes that a number of household characteristics of the kind to be found in UK census-derived small area statistics may be predicted from particular load profile indicators. A discussion of the steps required to test and validate this approach and the wider implications for census taking is also provided.
Article
Full-text available
This article describes the electricity consumption in Brazilian residences between 1985 and 2013 through linear regressions. The explanatory variables considered were the number of households, effective consumption of families as a proxy for family income, and electricity tariff for households. To deal with the power generation crisis of 2001 we have introduced a dummy variable in the form of a step function. With such explanatory variables, we were able to account for the reduction of household electricity consumption caused by the policies conducted in 2001 and their permanent consequences. The regression presented coefficient of determination of 0.9892, and the several statistic tests conducted assured the existence of long-term relation between the electricity consumption in residences and the explanatory variables. The obtained elasticities for the household consumption of electricity with respect to number of residences, family income and residential tariff of electricity were 1.534±0.095, 0.189±0.049, and −0.230±0.060, respectively. These results allowed understanding the evolution over time of the household consumption of electricity in Brazil. They suggest that the electric sector in Brazil should pursue an active policy to manage demand of residential electricity using tariffs as a means to control it.
Article
This research aims to detect areas of homogeneous residential electric consumption in the Great La Plata, Buenos Aires, Argentina. This study will identify main socio-demographic factors that impact on the electricity demand and the geographical location of these areas of homogeneous electric consumption. Results were analyzed from groups obtained from the six bimestrial electric average consumption of each of the 1010 census radius which constitutes the Great La Plata, using the K-means clustering method.
Article
Ever-growing energy needs and depleting fossil-fuel resources demand the pursuit of sustainable energy alternatives, including both renewable energy sources and sustainable storage technologies. It is therefore essential to incorporate material abundance, eco-efficient synthetic processes and life-cycle analysis into the design of new electrochemical storage systems. At present, a few existing technologies address these issues, but in each case, fundamental and technological hurdles remain to be overcome. Here we provide an overview of the current state of energy storage from a sustainability perspective. We introduce the notion of sustainability through discussion of the energy and environmental costs of state-of-the-art lithium-ion batteries, considering elemental abundance, toxicity, synthetic methods and scalability. With the same themes in mind, we also highlight current and future electrochemical storage systems beyond lithium-ion batteries. The complexity and importance of recycling battery materials is also discussed.