ABSTRACT
Objective: Build a statistical model to forecast energy consumption in order to estimate the future load in the agricultural sector in São Paulo state.
Theoretical Framework: The basis of the research was the application of time series, which is a set of observations made over time, with the focus of identifying historical patterns that can help explain past events and forecast future events.
Method: The classical method was used, which breaks down a time series into trends and seasonality. The steps were: collecting consumption information in chronological order; plotting the series graph, detecting the consumption trend and its seasonal variations; using the least squares method (trend equation); and using the ratio to moving average method (monthly seasonal indexes).
Results and Discussion: The 2024 energy consumption was determined by the increasing profile of the trend line and monthly seasonal variations with the highest consumption in September, followed by October. The average loads for 2024 were also estimated.
Research Implications: The results and the method may generate insights for managers of utility companies through the understanding and perspective of growth in electricity consumption in the sector.
Originality/Value: It is essential to identify the energy consumption profile, detecting the peak demand period. This is important to ensure the continuous supply by the electricity sector, which must plan its generation, transmission and distribution within safety limits.
Keywords: Statistical Model, Forecast, Future Trend, Energy Consumption, Agricultural Sector.
RESUMO
Objetivo: Construir um modelo estatístico de previsão do consumo de energia com o intuito de estimar a carga futura do setor agropecuário paulista.
Referencial Teórico: A base da pesquisa foi a aplicação das séries temporais, que é um conjunto de observações feitas ao longo do tempo, com o foco de identificar padrões históricos que possam ajudar a explicar eventos passados e a prever eventos futuros.
Método: Utilizou-se o método clássico que fragmenta uma série temporal em tendência e sazonalidade. As etapas foram: coleta das informações de consumo em ordem cronológica; plotagem do gráfico da série, detectando a tendência de consumo e suas variações sazonais; emprego do método dos mínimos quadrados (equação da tendência) e uso do método da razão à média móvel (índices sazonais mensais).
Resultados e Discussão: O consumo de energia de 2024 foi determinado pelo perfil crescente da reta de tendência e variações estacionais mensais com o maior consumo em setembro, seguido de outubro. As cargas médias de 2024 também foram estimadas.
Implicações da Pesquisa: Os resultados e o método poderão gerar insights nos gestores de empresas fornecedoras de energia pela compreensão e perspectiva de crescimento do consumo de eletricidade do setor.
Originalidade/Valor: É imprescindível identificar o perfil de consumo de energia, detectando o período de pico de demanda. A relevância dá-se para a garantia da disponibilidade contínua do fornecimento pelo setor elétrico, que deverá programar sua geração, transmissão e distribuição dentro dos limites de segurança.
Palavras-chave: Modelo Estatístico, Previsão, Tendência Futura, Consumo de Energia, Setor Agropecuário.
RESUMEN
Objetivo: Construir un modelo estadístico para predecir el consumo de energía con el objetivo de estimar la carga futura del sector agrícola del estado de São Paulo.
Marco Teórico: La base de la investigación fue la aplicación de series de tiempo, que es un conjunto de observaciones realizadas a lo largo del tiempo, con el objetivo de identificar patrones históricos que puedan ayudar a explicar eventos pasados y predecir eventos futuros.
Método: Se utilizó el método clásico, que descompone una serie temporal en tendencia y estacionalidad. Los pasos fueron: recolección de información de consumo en orden cronológico; trazar el gráfico de series, detectando la tendencia del consumo y sus variaciones estacionales; uso del método de mínimos cuadrados (ecuación de tendencia) y uso del método de ratio de media móvil (índices estacionales mensuales).
Resultados y Discusión: El consumo de energía en 2024 se determinó por el perfil creciente de la línea de tendencia y las variaciones estacionales mensuales con el mayor consumo en septiembre, seguido de octubre. También se estimaron las cargas medias para 2024.
Implicaciones de la Investigación: Los resultados y el método podrán generar insights para los gestores de empresas de energía eléctrica a través de la comprensión y la perspectiva del crecimiento del consumo de electricidad en el sector.
Originalidad/Valor: Es fundamental identificar el perfil de consumo energético, detectando el periodo de máxima demanda. La relevancia es garantizar la disponibilidad continua de suministro por parte del sector eléctrico, que debe planificar su generación, transmisión y distribución dentro de límites de seguridad.
Palabras clave: Modelo Estadístico, Pronóstico, Tendencia Futura, Consumo de Energía, Sector Agrícola.
1 INTRODUCTION
Idealizing modern society disconnected from any use of electricity is unrealistic. Ever since the light bulb was invented, humanity has enjoyed the benefits of electricity, which has since undergone successive developments up to the present day (Neto & Oliveira, 2021).
For Neto and Oliveira (2021), electricity is fundamental to the functioning of any building and industrial electrical and electronic system, from a residential power socket to a complex factory machine drive and control panel. A large part of people's daily activities (recreational and work-related) require the use of this energy.
According to Fogliatto et al. (2005), energy distributors have a challenging plan when it comes to forecasting their power demands for peak and off-peak times over the three-year period. If a distribution company forecasts and contracts a lower power demand than is actually consumed, a fine will be charged. If the contracted power demand is higher than the amount used, the distributor will bear the losses resulting from this difference.
The research sought to answer the following question: How can we estimate electricity consumption and power demand in rural São Paulo?
Research into forecasting energy consumption and power demand in São Paulo's rural sector is fully justified, as it is a valuable diagnostic tool. It creates a mathematical model that anticipates consumption, offering relevant data for the state, especially considering the upward trend in the use of this form of energy.
The general objective of the research was to build a statistical model that estimates future seasonal electricity consumption in the rural sector in São Paulo.
The specific objectives are:
a) to develop the trend equation for electricity consumption in São Paulo's rural sector;
b) to obtain the monthly seasonal coefficients of the sector's energy consumption;
c) to design the load estimate (future power demand forecast) for the sector.
2 THEORETICAL FRAMEWORK
Electricity demand forecasts are built using qualitative or quantitative techniques, or even a combination of them. Expert considerations are the mainstay of qualitative techniques and are therefore susceptible to predispositions that can damage the credibility of the results. As for quantitative techniques, they consist of analyzing time series, i.e. information that reflects fluctuations in energy demand over a period of time (Fogliatto et al., 2005).
A succession of observations verified at different times and organized sequentially gives rise to a series of information over time. Considering Y as the variable being researched and "t" as the time variable, the series will be represented by the values Y1, Y2, Y3, ..., Yn, at times t1, t2, t3, ..., tn. Consequently, Y is a function of t, i.e. Y=f (t) (Fonseca et al., 2013).
The time series analysis method requires the identification and insertion of time-correlated elements that affect the amounts verified in the series. These elements can be used to help understand and project the values of the time series (Kazmier, 1982).
Assessing changes or patterns of movement in time series is highly valuable in a number of situations, one of which is predicting future movements, and is crucial for both the private and government sectors (Spiegel, 1993).
The patterns used to describe time series are stochastic processes, i.e. processes governed by probabilistic laws (Morettin & Toloi, 2006).
Stevenson (1981) reports that the classic model conceptualizes time series as consisting of four fundamental patterns, namely:
a) long-term trend;
b) cyclical fluctuations;
c) seasonal variations;
d) irregular oscillations.
The concept of "trend" describes a smooth, long-term pattern of values, which can be upward or downward (Figure 1-A), over an extended period of time (Stevenson, 1981).
Cyclical fluctuations show a certain regularity - oscillatory movements up and down - in relation to the prevailing trend (Figure 1-B), lasting for several years (Kazmier, 1982).
In seasonal or seasonal variations, according to Kazmier (1982), it is possible to observe downward and upward movements that are completed in the course of a year and start again every year (Figure 1-C). These oscillations are usually identified on the basis of monthly or quarterly information.
According to Fonseca et al. (2013), understanding seasonal or seasonal time elements is important for estimating the short term. Most methods essentially seek to calculate a seasonal index for each period (such as month, bimonthly, quarterly, etc.) as a proportion of its average.
Irregular oscillations, on the other hand, are disturbances generated by factors that do not recur regularly (Fonseca et al., 2013).
Figure 1 shows three graphs illustrating the characteristic movements of time series. Figure 1-A is known as the secular or long-term trend, Figure 1-B is called the cyclical movement (overlapping the trend line) and Figure 1-C, called the seasonal movement, is also overlapping the trend line.
3 METHODOLOGY
3.1 METHODOLOGY PHASES
The first phase of the methodology defined the central question or research problem. The problem to be solved was to obtain estimates of electricity consumption and electricity demand in the agricultural sector in the state of São Paulo.
The second phase of the methodology was to determine the variables and time components. The relevant variables in the project were power demand (electrical load) and electricity consumption.
As for the time components, the monthly period was chosen as the time interval, since the consumption information collected from the Secretariat for the Environment, Infrastructure and Logistics (SEMIL) is in this unit of measurement.
Depending on the needs of decision-makers, consumption and demand for electrical power can be estimated at peak times: an interval of three consecutive hours on working days, from 6pm to 9pm in the concession area of the Energisa Sul-Sudeste distributor (ESS), as well as at off-peak times: other consecutive hours complementary to peak times on working days, in the ESS between 9pm and 6pm, in addition to the full day on weekends and national holidays (ANEEL, 2021). However, in order to achieve this goal, it is necessary to obtain hourly and daily data on electricity consumption.
Data collection was the third phase of the research. Energy consumption data, in Megawatt-hour (MWh), is input information for the proposed prediction model.
3.2 EXPLORING THE TIME SERIES
In the fourth phase, mathematical procedures for estimating energy consumption in the rural sector of São Paulo were carried out, as well as the plotting of this forecast profile.
As described by Fonseca et al. (2013), the method of searching for a time series involves decomposing the series into its four characteristic movements. The response (dependent) variable Y will be determined by the components T (trend), C (cyclical), S (seasonal) and I (irregular).
According to Stevenson (1981), there are two variations of the classic time series model: multiplicative and additive. In the additive model, C, S and I are also effective parameters, while in the multiplicative model, C, S and I are expressed as percentages of the trend.
Typically, the multiplicative model, Equation (1), is more used, above all, because it more accurately represents reality (Stevenson, 1981).
Y = T . S . C . I (1)
where:
Y = dependent/researched variable;
T = trend component;
S = seasonal component;
C = cyclic component; and
I = irregular component.
3.2.1 Trend determination
All research that contains long-term planning requires trend analysis (Fonseca et al., 2013).
According to Kazmier (1982), the least squares method is the most regularly used to determine the trend component of a time series by generating a straight or curved equation that best fits.
Equation (2) is the equation of the straight line that represents the trend component of the time series, adapted from Stevenson (1981).
Tt = α + β. t (2)
where:
α = amount (intercept of the straight line), in MWh, of Tt for the zero instant (t = 0);
β = angular coefficient of the line (slope of the line);
t = number of time period.
According to Stevenson (1981), the coefficients α and β are calculated, respectively, by Equations (3) and (4).
... (3)
... (4)
where:
CEE = electrical energy consumption (MWh);
n = number of observations (periods).
3.2.2 Determining seasonal variation
Seasonal variations are those that occur regularly within a year. To work with seasonal patterns, it is essential, first, to identify them and stipulate their scope (Stevenson, 1981).
To calculate the seasonal factor, it is necessary to estimate how the information in a time series changes from one month to another over the course of a normal year. A set of numbers that reflect the relative change in a variable for each month of the year is called a seasonality index (Spiegel, 1993).
According to Kazmier (1982), the most commonly used technique to calculate seasonality indices is the ratio to moving average method.
Fonseca et al. (2013) suggest the following steps that demonstrate the method:
a) first, the 12-month moving average is calculated;
b) then, a two-month moving average must be calculated to center the 12-month average between the months. Thus, a 12-month centered moving average is obtained;
c) then, the values must be calculated as percentages of the 12-month centered moving average;
d) finally, the average of the percentages is determined for each sub-period, resulting in seasonal indices.
3.2.3 Determination of cyclic and random fluctuations
When calculating the cyclical factors, as well as the random ones represented by the product "C.I", the result of this multiplication will have a negligible percentage variation. Therefore, for prediction purposes, this product is considered to be equal to one (1) (Spiegel, 1993).
Based on the above, Equation (1) will be modified and compacted, becoming Equation (5) for the multiplicative model.
Y = T . S (5)
where:
Y = dependent/researched variable;
T = trend component;
S = seasonal component.
By multiplying the values resulting from the trend equation (T), Equation (2), for each month, by their corresponding seasonal coefficients (S), monthly estimates will be obtained for all months of the year researched, and the results of the dependent variable (Y) will have irrelevant deviations (differences) due to the use of Equation (5).
3.2.4 Estimation of average electrical power
The fifth phase of the proposed method was the estimation of the average electrical power demand (DMPE) for each month "m", which is determined through a multiplicative calculation based on the forecast value of electricity consumption for that month.
As the unit of measurement for electricity consumption is given in MWh and its amount is cumulative over the month, electrical power is determined in megawatt (MW), dividing the consumption of month "m", determined by Equation (5), by the total number of hours in month "m" according to Equation (6), adapted from Fogliatto et al. (2005).
In improving forecasts of power demand and energy consumption, government technical experts could be heard as they have privileged information about possible market changes, as well as future expansions. These experts would be responsible for validating or not validating seasonality factors.
... (6)
where:
DMPEm = average electrical power demand (MW) for month "m";
Ym = electricity consumption (MWh) of month "m";
tdm = number of days in month "m".
Next, the results achieved with the application of the proposed method are determined and evaluated.
4 RESULTS AND DISCUSSIONS
There was difficulty in obtaining the information, as in the digital file Balanço Energético do Estado de São Paulo (BEESP), made available by SEMIL, data from the agricultural sector are included in the class called "others". This class also includes the consumption of public lighting, public power, public service and the consumption of companies responsible for the generation, transmission and distribution of energy.
The difficulty mentioned previously was resolved through requests via FALA.SP.GOV.BR (https://fala.sp.gov.br/).
The information collected from BEESP 2019 and 2023 produced historical series for the target sector of the study from January 2000 to December 2023, with the tabulation of 288 consumptions.
The values referring to electricity expenses in 2020 and 2021 may have been affected by the COVID-19 pandemic.
A time series is reproduced by a line graph, in which each point marked on it symbolizes a pair of observed values for the dependent and independent variables. Generally, a time interval is displayed on the horizontal axis, while the amounts of the researched variable are described on the vertical axis (Kazmier, 1982).
To better visualize seasonal electricity consumption (month by month) in the rural sector of São Paulo, a graph was plotted in orange with a trend line in red, as shown in Figure 2.
4.1 APPLICATION OF THE PROPOSED METHOD
The proposed method (multiplicative time series model) was used to estimate energy consumption and average power in the rural sector of São Paulo for the 12 months of 2024.
In this work, the response (dependent) variable Y was the estimated electricity consumption (ECEE) of the rural sector in São Paulo.
First, the trend, seasonality, cyclical and irregular/random components will be determined.
4.1.1 Trend determination
Trend estimation was carried out using the least squares method. This method determines the appropriate equation for a trend line.
To facilitate the calculations of coefficients α and β, Table 1 was created. The months were coded in increasing natural numbers - column t (period) - starting with the unit value (1), with the aim of simplifying the calculations, as well how to provide a value for α when "t" is equal to zero (t = 0).
With the energy consumption of 288 months and the calculation of the sums of the factors present in the columns of Table 1, the coefficients β equal to 429.106 were determined by Equation (4) and α equal to 177848.739, with Equation ( 3 ).
Replacing the amounts of the coefficients α and β in Equation (2), Equation (7) was obtained. Based on this linear equation, it is possible to obtain the trend T amounts.
Tt = 177848,739 + 429,106 . t (7)
4.1.2 Obtaining seasonality coefficients
The moving average ratio method was the technique used to calculate the seasonal indices. The values calculated at each stage are listed in Table 2 and explained below.
Initially, for the historical series information, 12-month moving totals were computed (Table 2).
The 1st amount of the rolling total, represented by 2,364,892 MWh in the 3rd column, is the result of adding the consumption from January to December 2000 (2nd column). The 2nd moving total, denoted by 2,351,750 MWh, corresponds to the sum of consumption from February 2000 to January 2001, and so on, until the sum of consumption from January to December 2023 equals 2,852,790 MWh.
Next, considering that the moving average must be centered on each month, rather than between months, the moving totals of 12 adjacent months are combined (adding their amounts, 2 by 2, from the 3rd column) to generate the centralized two-year moving totals (4th column).
It is important to understand that this type of total does not exactly represent two years of data, but two overlapping 12-month periods.
Therefore, as the amounts of the moving totals for each 12 months are positioned between the rows of the Table (3rd column), since it is a moving total from an even number of months, the result is always between two months. Note that the 1st result (2,364,892 MWh) in the 3rd column of Table 2 is between the 6th and 7th months (between June and July 2000).
One way to solve this problem is to calculate the sum, two by two, of the amounts in the 3rd column, obtaining central values, which was called the two-year centered moving total (4th column).
In this way, the 4th column was determined, with its 1st amount being equal to 4,716,642 MWh, the result of adding 2,364,892 MWh and 2,351,750 MWh. The other values in the 4th column followed this line of reasoning.
The 5th column (centered moving average) was then assembled by dividing the amounts of the centered moving total (4th column) by 24 (months).
At the end of Table 2, the moving average ratio (last column) was determined by dividing each of the actual monthly consumptions (2nd column) by the centered moving average (5th column) corresponding to that month, expressing each result as a percentage.
The values of the moving average ratio of the historical series, calculated and expressed in Table 2, were ordered month by month, giving rise to Table 3, which shows the seasonality of these indices in percentage terms for the period between July 2000 and June 2023.
After grouping the relative values of similar periods (Table 3), the seasonal averages for each period are calculated.
According to Stevenson (1981), it is common to determine the "modified average" for each month. This percentage average is calculated after removing the highest and lowest values from each monthly group.
In view of the above, we calculated the so-called modified average for all the months after removing the two extreme values (the maximum and the minimum) from each monthly group. The modified averages for each month are transcribed in the 2nd column of Table 4.
Stevenson (1981) states that the addition of the modified averages must be equal to the number of periods; consequently, for a period of 12 months, the sum of the relative seasonals must be equal to 12.
Kazmier (1982) mentions that if the total of the modified averages is different from 12 for a monthly seasonality, each of the 12 relative modified averages needs to be multiplied by an adjustment factor, which should be determined by Equation (8).
... (8)
As the total modified means, according to Table 4 (2nd column), was a percentage value equal to 1197.5%, and not 1200%, Equation (8) was applied to determine the adjustment factor. The adjustment factor resulted in a value of R$1.0021.
This factor was multiplied by the 12 modified monthly averages to obtain the adjusted averages (3rd column), which are the final monthly seasonal indices described in Table 4.
The total adjusted average was 1200% (Table 4), revealing good calibration of the 12 monthly indices.
Finally, the monthly seasonal indices in percentage are detailed in the 3rd column (adjusted averages) of Table 4.
4.1.3 Estimated energy consumption for 2024
First, the consumption trend for the months to 2024 was determined using Equation (7). Since the historical series surveyed consists of 288 months, the value of the variable "t" in the equation will be equal to 289 for January 2024, and so on, according to the amounts of "t" and the trends in MWh listed month by month in Table 5.
The seasonality percentages listed in Table 5 were then transcribed from the last "Adjusted Averages" column in Table 4.
Finally, with the monthly trend information and seasonal indices, consumption for the months of 2024 was estimated using Equation (5), as described in Table 5.
With the estimated monthly amounts of electricity consumption in the rural sector of São Paulo (last column of Table 5), it is possible to observe that the highest consumption will be in September and October, respectively.
4.1.4 Average electricity demand forecast for 2024
Regarding the forecast of the average electrical power demand (DMPEm) for the months of 2024, Equation (6) was used, with the consumption results listed in Table 5.
The 12 forecasts of average electrical power demands are recorded in Table 6, considering 29 days for the month of February as it is a leap year (2024).
According to Table 6, the maximum predicted power (502.46MW) will occur in the month of September, a month that will require greater attention with the supply of energy to the rural sector of São Paulo.
The future values estimated with the simulation of energy consumption and average electrical power for the months of 2024 may be compared in the future with information from the BEESP digital file, allowing an assessment of whether the results achieved were consistent with the real values consumed by the sector.
5 CONCLUSION
If the information obtained from SEMIL were hourly (available 24 hours a day) and daily (seven days a week), it would be possible to determine, in consultation with electricity distributors, the hourly power demand profile and daily, as well as the period of greatest energy demand (peak consumption) by the agricultural sector in São Paulo, including its behavior at peak and off-peak times.
The research results were constructed using an increasing linear trend with influences, month by month, of seasonality indices, resulting in monthly values predicted for 2024 higher than the actual values for 2023.
Despite the population increase and its growing needs for food and renewable energy produced in the countryside, national and/or global economic downturns are likely to occur, which would negatively impact its consumption of electrical energy and, consequently, its demand for power.
However, the entire electrical supply system - generation, transmission and distribution of electricity - must be capable and have safety margins to meet customer requests; hence the importance of studying consumption and load demand forecasts.
It is important to emphasize that time series analysis is an art as much as a science, and that data processing alone does not solve all problems. Combined with the researcher's common sense, judgment, experience, and skill, this mathematical analysis can be extremely valuable in predicting shortand long-term events.
ACKNOWLEDGMENTS
This work was carried out with the support of the Coordination for the Improvement of Higher Education Personnel - Brazil (CAPES) - Funding Code 001.
I would like to thank the Federal Institute of São Paulo (IFSP) for granting me paid leave for my Stricto Sensu qualification.
REFERENCES
Agência Nacional de Energia Elétrica (ANEEL). (2021). Resolução Normativa (REN) ANEEL N 1.000, de 7 de dezembro de 2021. Consultado em 25 de janeiro de 2023. https://www2.aneel.gov.br/cedoc/ren20211000.pdf.
Fogliatto, F. S., Ribeiro, J. L. D., Werner, L., Lemos, F. O., & Brum, M. P. (2005). Previsão de demanda de energia elétrica - Método e aplicação. Encontro Nacional De Engenharia De Produção, Porto Alegre. Anais [...] p. 3100-3107. Consultado em 7 de novembro de 2022. https://producaoonline.org.br/rpo/article/view/385/457.
Fonseca, J. S. da, Martins, G. de A., & Toledo, G. L. (2013). Estatística aplicada. Editora Atlas.
Kazmier, L. J. (1982). Estatística aplicada à economia e administração. Editora Pearson Makron Books.
Morettin, P. A., & Toloi, C. M. C. (2006). Análise de séries temporais. Editora Blucher.
Neto, A., & Oliveira, Y. de (2021). Instalação residencial aplicada à IoT: aprenda de forma descomplicada. Editora Alta Books.
Spiegel, M. R. (1993). Estatística. Editora Makron Books.
Stevenson, W. J. (1981). Estatística aplicada à administração. Editora Harper & Row do Brasil.
You have requested "on-the-fly" machine translation of selected content from our databases. This functionality is provided solely for your convenience and is in no way intended to replace human translation. Show full disclaimer
Neither ProQuest nor its licensors make any representations or warranties with respect to the translations. The translations are automatically generated "AS IS" and "AS AVAILABLE" and are not retained in our systems. PROQUEST AND ITS LICENSORS SPECIFICALLY DISCLAIM ANY AND ALL EXPRESS OR IMPLIED WARRANTIES, INCLUDING WITHOUT LIMITATION, ANY WARRANTIES FOR AVAILABILITY, ACCURACY, TIMELINESS, COMPLETENESS, NON-INFRINGMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Your use of the translations is subject to all use restrictions contained in your Electronic Products License Agreement and by using the translation functionality you agree to forgo any and all claims against ProQuest or its licensors for your use of the translation functionality and any output derived there from. Hide full disclaimer
© 2024. This work is published under https://rgsa.emnuvens.com.br/rgsa/about/editorialPolicies#openAccessPolicy (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Abstract
Objective: Build a statistical model to forecast energy consumption in order to estimate the future load in the agricultural sector in São Paulo state. Theoretical Framework: The basis of the research was the application of time series, which is a set of observations made over time, with the focus of identifying historical patterns that can help explain past events and forecast future events. Method: The classical method was used, which breaks down a time series into trends and seasonality. The steps were: collecting consumption information in chronological order; plotting the series graph, detecting the consumption trend and its seasonal variations; using the least squares method (trend equation); and using the ratio to moving average method (monthly seasonal indexes). Results and Discussion: The 2024 energy consumption was determined by the increasing profile of the trend line and monthly seasonal variations with the highest consumption in September, followed by October. The average loads for 2024 were also estimated. Research Implications: The results and the method may generate insights for managers of utility companies through the understanding and perspective of growth in electricity consumption in the sector. Originality/Value: It is essential to identify the energy consumption profile, detecting the peak demand period. This is important to ensure the continuous supply by the electricity sector, which must plan its generation, transmission and distribution within safety limits.