Effect of Meteorological Variables on Energy Demand in the Northeast and Southeast Regions of Brazil

Abstract

Energy consumption demand has shown successive records during recent months, primarily associated with heat waves in almost all Brazilian states. The effects of climate change induced by global warming and the increasingly frequent occurrence of extreme events, mainly regarding temperature and precipitation, are associated with this increase in demand. In this sense, the impact of meteorological variables on load demand in some substations in the northeast and southeast of Brazil was analyzed, considering the historical series of energy injected into these substations. Fifteen substations were analyzed: three in the state of São Paulo, six in Bahia, three in Pernambuco, and three in Rio Grande do Norte. Initially, essential quality control was carried out on the energy injection data. The SAMeT data sets were used for the variable temperature, and Xavier was used for precipitation and relative humidity to obtain a homogeneous data series. Daily and monthly data were used for a detailed analysis of these variables in energy demand over the northeast and southeast regions of Brazil. Some regions were observed to be sensitive to the maximum temperature variable, while others were sensitive to the average temperature. On the other hand, few cases showed the highest correlation with the precipitation and relative humidity variables, with most cases being considered slight or close to zero. A more refined analysis was based on the type of consumers associated with each substation. These results showed that where consumption is more residential, the highest correlations were associated with maximum temperature in most stations in the northeast and average temperature in the southeast. In regions where consumption is primarily rural, the correlation observed with precipitation and relative humidity was the highest despite being negative. A more detailed analysis shows that rural production is associated with irrigation in these substations, which partly explains consumption, as when rainfall occurs, the demand for irrigation decreases, and thus energy consumption is reduced.

1. Introduction

Within the last few decades, global economic expansion due to industrialization and technological advances associated with an increase in the demographic density of large centers has triggered a considerable increase in energy consumption. In the latest IPCC reports [1], evidence was presented of an increase in the frequency of intense climatic events in different parts of the globe associated with climate change resulting from anthropogenic actions. As in several countries, Brazil has also suffered the impact of these changes, with the occurrence of extremes in precipitation that recently occurred in the state of Rio Grande do Sul [2] and in temperature (heat waves), something frequent in recent years, especially in central-west and southeast regions of Brazil, in addition to the extreme drought of 2023 in the Amazon region [3].

Climate extremes impact various socioeconomic sectors of society in general. In many cases, this impacted the absence of energy supply and, in others, increased consumption, such as in January 2015, when the demand for electrical energy was much higher than planned, culminating in the interruption in energy distribution in ten Brazilian states and the Federal District [4]. In this last situation, one of the factors responsible for the significant load demand was attributed to the excessive use of air conditioning [5]. Some studies show that excessive exposure to air conditioning tends to reduce tolerance to high temperatures [6]. However, in addition to temperature, other meteorological variables can influence thermal comfort, such as relative humidity, wind, and precipitation, and consequently interfere with the demand for injected energy load.

New elements have emerged in recent years in the energy consumption of large cities, such as the increase in sales of electric vehicles, distributed energy, and the entry of renewable energy sources, wind and solar energy, which have impacted energy demand and its prediction [7]. The introduction of more meteorological variables in the energy demand forecast becomes essential for reducing errors associated with climate change, such as significant events and an increase in the occurrence of extreme events such as heat waves and droughts. Ref. [8], analyzed the influence of meteorological variability on Spain’s daily electricity demand and proposed a new index considering large-scale climate variability. Refs. [9,10] analyzed the effects of climate change on energy demand for five energy sources (coal, natural gas, oil, electricity, and renewable energy) across four sectors (commercial, industrial, residential, and transportation) in the United States, and its results highlight the political implications for the development of climate-smart technologies for different energy sources by sectors.

Considering that energy cannot be stored economically, there must be a constant balance between supply and demand. Ref. [11] considered that accurate energy consumption is essential to achieving energy security in the Brazilian Electricity Sector, which is why a forecasting method that considers space-time dynamics was proposed.

Understanding the influences and evolutions of climatological reference variables and other determinants of injected energy loads, including lags and acclimation, could have important ramifications for expected energy loads in short-term and decadal projections as temperatures rise and other meteorological patterns evolve with climate change. Some studies have developed a new counterfactual modeling approach for Brazil to investigate the impacts of extreme or disruptive historical events on energy consumption, considering the effects of climate change. New proposed methodologies [12] had good results in measuring the impacts of events on daily and hourly energy consumption patterns.

In energy consumption demand, three time horizons are defined. The short-term horizon, typically defined as a few days ahead, is essential for control, power system scheduling, and short-term price forecasting. The medium-term forecast, which ranges from a few days to about a year, informs system operation, maintenance schedules, and the negotiation of future contracts, and the long-term horizon can vary from one year to several decades and guides capacity planning. In short-term demand, the use of meteorological variables is well defined. However, few studies use meteorological variables for forecasting medium- and long-term energy load. Those that do use them focus only on air temperature as the primary climate variable for energy load forecasting. Recent results using linear regression and nonlinear feed-forward neural network methods indicate that models that include future temperature consistently significantly outperform models that exclude temperature across all time horizons [13].

Knowledge of the climatic factors that impact energy consumption is key in projecting consumption and consequent demand. This will support the planning and definition of distribution and commercialization, promote investment planning, and reduce operational costs, considering different classes of consumers.

In this context, to explore the influence of meteorological variables, an analysis will be carried out regarding the injected energy load in several energy substations and the influence of meteorological variables in Brazil’s southeast and northeast regions. It is worth noting that these are preliminary results related to Neoenergia’s research and development (R&D) project, regulated by the National Electric Energy Agency (ANEEL) in partnership with Tempo OK and the Federal University of Alagoas (UFAL).

This study attempts to fill the methodological gap by showing the importance of considering local effects in energy demand forecasting, where, historically, only climatological temperature is considered. In many regions, other meteorological variables can be as important or more critical when included in energy demand forecasting. In this context, this study contributes to the literature on the use of other meteorological variables, in addition to temperature, in energy demand forecasting by showing that there is, in fact, a significant gain when other variables are considered, such as precipitation and relative humidity in certain regions of Brazil, considering the specificities of each location and the specific energy use. Another significant difference between this study and others is the use of measured data on injected energy from fifteen substations in the northeast and southeast of Brazil under the Neoenergia concession.

2. Materials and Methods

Electricity demand varies throughout the day, throughout the week, by season, throughout the year, and for longer periods. This variation can be associated with several factors, climatic and socioeconomic, among others, and the prediction of these variations can be quite complicated, generally due to the different factors that are part of this predictive equation.

In this study, injected energy load data will be analyzed using fifteen Neonenergia substations distributed throughout different regions of the states of São Paulo, Brazil (Andradina, Francisco Morato, and Ubatuba I) in the Eletkro concession region, in Bahia, Brazil (Asa Branca, Barreiras do Norte, Lobato, Rio Guará, Formoso, and América Dourada II) under a concession from Coelba (Companhia de Eletricidade do Estado da Bahia), from Pernambuco, Brazil (Pau Amarelo, Dom Avelar, and Campus), and from Rio Grande do Norte, Brazil (Jiqui, Mossoró III, and Natal) in the concession region of Cosern (Companhia Energética do Rio Grande do Norte), where the spatial distribution of Neoenergia substations are presented in Figure 1 and the details of the substations are presented in

Table 1. Neoenergia substations with concession details, type of consumer (residential and rural), substation acronym, and data availability.

Stations	Concession	Consumer	Acronym	Period
Asa Branca, BA	Neoenergia Coelba	81% Res	ASB	Mar. 2019 to Nov. 2023
Barreiras Norte, BA	Neoenergia Coelba	79% Res	BRN	Jan. 2011 to Nov. 2023
Lobato, BA	Neoenergia Coelba	79% Res	LBT	Jan. 2016 to Nov. 2023
Rio Guará, BA	Neoenergia Coelba	73% Rur	RGA	Nov. 2015 to Nov. 2023
Formoso, BA	Neoenergia Coelba	65% Rur	FRM	Nov. 2015 to Nov. 2023
América Dourada II, BA	Neoenergia Coelba	46% Rur	ADD	Jan. 2017 to Nov. 2023
Andradina, SP	Neoenergia Elektro	66% Res	ANR	Aug. 2010 to Nov. 2023
Francisco Morato, SP	Neoenergia Elektro	87% Res	FRM	Aug. 2010 to Nov. 2023
Ubatuba I, SP	Neoenergia Elektro	74% Res	UBA2	Aug. 2010 to Nov. 2023
Pau Amarelo, PE	Neoenergia PE	85% Res	PAM	Dec. 2015 to Nov. 2023
Don Avelar, PE	Neoenergia PE	84% Res	DAV	Dec. 2015 to Nov. 2023
Campus, PE	Neoenergia PE	81% Res	CPS	Dec. 2015 to Nov. 2023
Jiqui, RN	Neoenergia Cosern	77% Res	JQI	Jan. 2016 to Nov. 2023
Mossoró III, RN	Neoenergia Cosern	74% Res	MST	Jan. 2016 to Nov. 2023
Natal RN	Neoenergia Cosern	74% Res	NTU	Jan. 2014 to Dec. 2018

.

Table 1 presents the details of the substations used in the present study, the concessionaire, the percentage of consumer type (residential or rural), the acronym, and the availability period of injected energy load data for each substation.

The first phase of the study consisted of primary quality control of energy load data from the fifteen substations, where inconsistent data were removed. To avoid possible trends associated with data observed from surface meteorological stations, it was proposed to use data from the South American Mapping of Temperature (SAMeT) [14], which are regularly spaced in space and time. SAMeT combines temperature data from the ERA5 reanalysis [15] with data from surface meteorological stations (automatic and conventional), METeorological Aerodrome Report (METAR), and data from regional centers across South America, with a horizontal resolution of 5 km. The SAMet data were interpolated for the location of each substation, using the average temperature for Elektro’s concession regions (SP) and the maximum temperature for the other regions. The choice to use average temperature in the southeast region is associated with the fact that the region is more susceptible to the entry of cold air masses coming from the south of Brazil, while the northeast region is more influenced by high temperatures, with a more tropical climate, without a direct influence of cold air masses, which results in a smaller thermal amplitude, compared to the southeast region of Brazil. For precipitation and relative humidity data, the dataset generated by [16] was used, which provides the most complete gridded analysis of surface meteorological variables in Brazil based on all available observation networks from federal, state, municipal, and independent networks. The dataset consists of precipitation and relative humidity data collected between 1961 and 2020, subjected to rigorous quality control. The data were then interpolated following the best results obtained through cross-validation between the inverse distance weighting (IDW) and angular distance weighting (ADW) methods to generate a high-resolution 0.1° × 0.1° grid, incorporating data on topographic relief and using observed data from 11,473 rain gauges and 1252 meteorological stations.

In the data analysis, a quantitative check was performed daily, using Pearson’s correlation coefficient (Equation (1)) to measure the strength and direction of the relationship between the two variables [17,18].

$$r={\displaystyle \frac{{\sum }_{i=1}^n\left[(\mathit{xi}-\overline{x})(\mathit{yi}-\overline{y})\right]}{\sqrt{{\sum }_{i=1}^n(\mathit{xi}-\overline{x}\stackrel{2}{)}}\sqrt{{\sum }_{i=1}^n(\mathit{yi}-\overline{y}\stackrel{2}{)}}}}$$

(1)

where xi is the value of the energy injected from each Neoenergia substation, and yi is the variable used in the correlation, which can be temperature, precipitation, or relative humidity. The values with the bar are the averages of the values used.

3. Results

Considering the availability of meteorological data in the dataset used in this work and the variables that have the greatest influence on load demand, temperature, precipitation, and relative humidity, these variables will be analyzed separately in the following topics.

3.1. Temperature

Initially, the injected energy load data obtained by Neonergia will be analyzed and compared with the temperatures of each substation. Neoenergia considers in its projections the average temperature data for the stations located in the southeast region of Brazil, in the state of São Paulo (Andradina, Francisco Morato, and Ubatuba I), and the maximum temperature for the other stations located in the northeast region of Brazil, in the states of Bahia (Asa Branca, Barreiras Norte, Lobato, Rio Guará, Formoso, and América Dourada II), Pernambuco (Pau Amarelo, Don Avelar, and Campus), and Rio Grande do Norte (Jiqui, Mossoró III, and Natal). Figure 2 presents the time series of the fifteen Neoenergia substations, considering the observed energy injected load and the temperature associated with each substation obtained from the SAMeT dataset [14].

The joint analysis of the time series of all substations (Figure 2) shows a certain regularity with the annual cycles of maximum and minimum temperature values. The analysis of the injected energy load series shows a similar behavior, with some substations presenting a very well-defined annual cycle and similar to the annual temperature cycle, with maximum and minimum values, which reinforces regular consumption, as is the case with Don Avelar substation (PE) where consumption is 84% residential, and strongly associated with temperature. However, this behavior is unclear in some substations, such as the Rio Guará (RGA) substation and Formoso (FRM) in the Neoenergia COELBA concession region. These substations, in particular, present an opposite relationship between maximum temperature peaks and maximum injected energy loads. A more careful analysis shows that these substations have a different type of consumer (Table 1), with 73% and 65% being predominantly rural consumption, respectively, which may be associated with other climatological variables besides temperature.

Analysis of the graphs in Figure 2 shows a strong relationship between the energy injected load series and the temperature series for most of the substations analyzed, especially for the stations of Asa Branca, Lobato, América Dourada II, Andradina, Ubatuba I, Pau Amarelo, Don Avelar, and Jiqui. It should be noted that for the Andradina (Figure 2g), Francisco Morato (Figure 2h), and Ubatuba I (Figure 2i) substations, average temperature data are used, and maximum temperature data for the other substations that are in the northeast region. In some substations, the relationship is inverse, as shown above, where the maximum injected energy load coincides with the minimum temperature values, as can be observed in Rio Guará (Figure 2d). However, in other substations, the relationship is not very clear, such as in Formoso (Figure 2e), and they will be analyzed below. The highest daily correlations with injected energy and maximum temperature occurred for the Don Avelar (0.69) and Jiqui (0.59) substations, and considering the average temperature, the highest correlations were observed in Ubatuba I and Andradina (0.56) and (0.54), respectively.

All analyses consider daily averages of injected energy load and temperature. When the monthly average is analyzed, there is an increase in the correlation between these variables, as can be seen in Figure 3. In practically all substations, there is an increase in the correlation between the injected energy load and temperature when the monthly averages are considered. This is because the monthly average reduces the most sensitive temperature variations, smoothing daily extremes and strengthening the relationship between the variables. The Ubatuba I substation presents a moderate-to-high correlation value considering daily correlation values. When we consider the correlation with monthly data, this value becomes more significant (Figure 3). However, this is a substation where most residences are summer residences, with a peak in use in the summer, especially at the end and beginning of the year (Figure 2l), which is very common on the coast of the state of São Paulo, where most of the residences are used only for summer vacations, which explains the increase in consumption, associated with the temporary increase in the general population. One of the only substations that showed a reduction in the monthly correlation was Rio Guará (Figure 2d), which will be the subject of new analyses. In some cases where the correlation is light-to-moderate, the analysis of the series shows that there was a considerable increase in load variation during the analysis period, as is the case of Barreiras Norte (Figure 2b), as well as a reduction in consumption, such as what happened in Natal (Figure 2p), which in the last case can be explained by the growing increase in the number of users with residential solar generation in this region [19] and also can be associated with other factors, particularly extra climatic factors, such as population variation and socioeconomic factors. Another notable substation is Pau Amarelo—PAM (Figure 2j), located in the interior of the state of Pernambuco, with 85% residential use. The correlation between injected energy load and maximum temperature is 0.47 (Figure 3) when considering the entire analysis period. However, in a more detailed analysis, if the first years of 2016–2018 are removed and only 2019 onwards is considered when the injected energy load was stabilized, the correlation rises to 0.72, showing a strong correlation between the two variables. A similar situation occurs at other stations, such as Don Avelar—DAV (Figure 2l), also in Pernambuco, with 83.55% residential consumption, where if the first years (2016–2017) are disregarded until the adjustment in energy injection load occurs, the correlation rises from 0.69 to 0.78. These results can only be obtained when daily data are used in the analysis.

3.2. Precipitation

Among the meteorological variables that are most sensitive to variations in energy consumption is precipitation, which, despite being a discrete variable, greatly influences the thermal comfort of the general population. The time series of daily precipitation presents a certain consistency, with well-defined periods of dry season and rainy season. However, in some substations, although the rainy season is well-defined, there is no clear dry season, as precipitation occurs practically throughout the year, especially in the stations located in the southeast, namely Andradina (Figure 4g), Francisco Morato (Figure 4h), and Ubatuba I (Figure 4i). On the other hand, in some substations, the period with precipitation is very short, characterizing a typical semi-arid rainfall regime, such as the América Dourada II (Figure 4f) and Don Avelar (Figure 4l) substations.

The main results observed from the comparison between the injected energy load and precipitation data make it clear that it is impossible to verify a direct relationship with precipitation, as can be verified with temperature. However, for the substations highlighted in the previous session, Rio Guará (Figure 4d), Formoso (Figure 4e), and América Dourada II (Figure 4f), it is possible to verify that the correlations are significant and inverse, −0.17, −0.35, and −0.40, respectively, indicating that in situations in which the highest volumes of precipitation occur are associated with the lowest values of injected energy load, since these substations have their largest consumers registered as rural. The high demand for irrigation in these regions may partly explain this increase in energy injection, which is associated with a long period of drought. Reinforcing that the América Dourada II substation (Figure 4f) has almost 50% rural consumption. The correlation of the energy load injected from this substation with the maximum temperature can also be considered moderate (0.36), where irrigation occurs during the late spring period and continues until the beginning of autumn [20]. The average typical consumption is around 50,000 MW, reaching values above 100,000 MW during the most significant injection of energy load. Another interesting factor at this substation is associated with a decrease in the energy load injected during recent years, which may be related to an increase in precipitation totals, especially at the end of 2021 and during 2022 (Figure 4f).

The Don Avelar substation (Figure 4l) presents a low-to-moderate value of inverse correlation between precipitation and injected energy load (−0.16). Still, it presents a high correlation with maximum temperature (Figure 3), being a substation with 84% residential consumers (Table 1). This can be associated with low precipitation values in the substation region, which relates the occurrence of precipitation with a reduction in temperature maximums, being a region with low precipitation values with a climatological average of around 500 mm per year.

3.3. Relative Humidity

The analyses associated with the precipitation series and injected energy load showed a significant correspondence for the substations with predominantly rural consumption (Table 1). In this session, only the Rio Guará, Formoso, and América Dourada II substations, which showed the highest correlation with precipitation, will be analyzed.

Figure 5 presents the time series of daily injected energy load and relative humidity associated with the substations of Rio Guará (Figure 5a), Formoso (Figure 5b), and América Dourada II (Figure 5c), located in the region of concession from Neoenergia, COELBA, for the period from January 2016 to July 2020, January 2011 to July 2020, and January 2017 to July 2020, respectively, according to the availability of relative humidity data from the data series by [16].

The analysis of Figure 5 clearly shows a strong relationship between the energy load injected in the period and the relative humidity, which is an inverse relationship where the reduction in the availability of relative humidity is associated with the increase in the energy load. As presented in the previous session, the Rio Guará substation has a majority profile of rural consumers (73.15%), with a high demand for irrigation [20], which explains the increase in demand for injected energy when relative humidity is below 50% (Figure 5a). A similar situation can be observed at the Formoso (Figure 5b) and América Dourada II (Figure 5c) substations since the América Dourada II substation presents a very similar pattern—increase (decrease) in injected energy load and decrease (increase) in relative humidity—with that of Rio Guará (Figure 5a), with a strong relationship between the injected energy and relative humidity, in an inverse pattern, when one increases the other decreases, showing the sensitivity of these substations to the drop in relative humidity.

The results become more evident when the correlation between meteorological variables and the energy load injected into the substations is calculated (Figure 6). The correlation with the precipitation series showed signs of a strong inverse relationship with the injected energy load. The high inverse correlation of injected energy load and specific humidity (Figure 6) reinforces the inverse relationship, which occurs mainly with the substations that have the highest rural consumption, Rio Guará (RGA), Formoso (FRM), and América Dourada II (ADD), associated with energy consumption and irrigation, with inverse correlation values of −0.70, −0.65, and −0.63, respectively, single values with high correlation. It is important to highlight that all substations showed a negative correlation with specific humidity (Figure 6) and a positive correlation with temperature (Figure 3).

Analyses relating to the energy load series injected into the various substations show that in addition to temperature, variables such as precipitation and relative humidity must be considered in future energy load forecasts, especially when primarily rural consumers are considered.

4. Conclusions

In this study, we analyze how injected energy loads are associated with meteorological variables in several areas of the southeast and northeast regions of Brazil, with particular interest in how injected energy demand is related to temperature, precipitation, and relative humidity and its relationship with different types of consumers, considering daily injected energy load measured data in fifteen Neoenergia substations.

The results analyzed show an increase in energy demand in practically all substations, which can be attributed to several factors, such as an increase in population, socioeconomic factors, and factors associated with the climate. The understanding and relationship of these variations with meteorological variables is the main object of this study, and along the way, it became clear the high correlation of the injected energy load with the maximum temperature in Don Avelar and Jiqui in the northeast and with the average temperature in Ubatuba and Andradina in southeast Brazil, with a stronger relationship appearing when we consider the monthly average temperature and injected energy load. These relationships were primarily observed among residential consumers. When rural consumers are analyzed, this relationship becomes weaker, and the precipitation and relative humidity variables become more effective in explaining the relationship with energy demand. The explanation for this change in relationship patterns with meteorological variables is associated with the irrigation component, which is responsible for the most significant demands in this region with a predominance of rural consumers. When relative humidity falls below 50%, there is an increase in load demand at these substations, in Rio Guará, Formoso, and América Dourada II, which occurs when the irrigation systems come into operation, in one of the driest regions of Brazil [21].

Overall, this work contributes to understanding the relationship between energy demand and meteorological variables and different types of consumers. It can also be used to inform load forecasts on a time scale of months up to a decade, where other types of consumers must be considered. Studies of longer decennial energy demand projections typically use temperature data and socioeconomic factors (e.g., variation in electricity price and population change), in which case relative humidity should be included, especially when consumers are considered rural areas, with daily or monthly data, to predict the main effects of climate change.

With climate change representing an increasingly greater challenge for electricity consumption planning and society in general, the results presented here, as well as possible extensions, can assist in the accuracy of electricity demand projections, especially when considering both types of consumers, residential and rural. Since climate projections suggest that there will be higher temperatures, altered precipitation patterns, and, consequently, relative humidity, as well as more frequent, intense, and longer heat waves, projections of meteorological variables are likely to be of fundamental importance for energy demand projections.

Through this study, evidence was found that regional energy consumption in Brazil is spatially dependent, presenting a spatial pattern of dissimilarity between regions, with the need to include other meteorological variables in energy demand projections, such as relative humidity and precipitation, in addition to temperature.