Brazil electricity needs in 2030: trends and challenges
BBrazil electricity needs in 2030: trends and challenges
Marcos Paulo Belan¸con a a Universidade Tecnol´ogica Federal do Paran´a (UTFPR), Cˆampus Pato Branco, Brazil
Abstract
The demand for electricity and the need to replace fossil fuels by renewableshave been growing steadily, and this transition will have significant implica-tions to our world that are only beginning to be understood. Brazil is oneimportant example of a big economy where the electricity is already suppliedby renewables, such as hydro, wind and biomass-fired thermal power. In thiswork we investigated the electricity load curves in the last 20 years in Brazil,and four different scenarios for 2030 are proposed in order to evaluate theimpact of increasing renewables in the national grid, at an hourly basis. Theanalysis shows that growing electricity demand and the expected reduction inthe hydropower share will significantly increase the reliability of the nationalgrid, due to higher peak load and also due to the intermittency of Solar andWind. Without any gigawatt scale hydropower projected for the near future,increasing the share of these renewables should push hydropower to operatehundreds of hours every year above typical peak power levels experienced inthe past. In order to avoid or reduce the threat related to this trend one ofour scenarios suggests that solar water heaters could be massively deployedin Brazil, what would positively impact the system reliability by reducingthe electricity demand mostly at peak loads during early evenings.
Keywords:
Solar, Energy in Brazil, Sustainability
1. Introduction
Brazil has more than 200 million inhabitants and is the fifth biggest coun-try in the world by land area. By several parameters, such as gross domesticproduct (GDP) and human development index (HDI) the country representsthe world average and it can be seen as a sample of the entire world.
Email address: [email protected] (Marcos Paulo Belan¸con a ) Preprint submitted to Energy September 24, 2020 a r X i v : . [ phy s i c s . s o c - ph ] S e p he growing need to decarbonize our energy system worldwide and miti-gate climate change [1, 2, 3, 4, 5, 6] makes noteworthy the unique proportionof renewables in Brazil’s primary energy [7], which is three times superior tothe world average (45 % against only 15%). This makes Brazil the “greenest”among the biggest economies of the world: its Carbon intensity is about 0.15 kgCO /U S $ ppp , which is lower than in Europe (0.18), United States (0.29),and China (0.47).Hydropower is by far the most important source of electricity and untila few decades ago this single-source supplied more than 95% of the demand.But since the 2000’s concerns [8] about the challenges of expanding electric-ity production without relying on fossil fuels are growing. In 2001 droughtconditions [9, 10] and delay in generation investments resulted in one of theworst energy crises of the modern times, which sounded the alarm to enhanceenergy security by diversifying its electricity sources.Despite some huge investments in new hydro capacity [11, 12] since the2000’s, in 2014 a new energy crisis was triggered by constrained hydropowerdue to low precipitation levels [13] and minor factors such as mismanagementof the regularization reservoirs [10]. The expansion in nominal hydropowercapacity in the last 10 years was not translated into higher hydropower gen-eration. The production has peaked in 2011 (449 TWh) and remained about10% below this level since then, mainly due to droughts in the southeastof the country and the head loss in several strategic reservoirs in this area,which has as consequence also impacted in the hydropower revenues [14].Other renewables have significantly expanded in Brazil, namely biomass-fired thermal and wind. However, gas-fired thermal power is on the riseand is expected to be the second most important source of electricity inthe coming years, pushing up greenhouse gas (GHG) emissions from powergeneration [15]. This trend is the result of the increasing power demandand the intermittent nature of Wind and Solar, which cannot meet the loadcurves [16].Briefly one may enumerate some relevant questions about the future ofelectricity production in Brazil:1. Is there a fossil-fuel-free pathway to enhance the reliability of the sup-ply?2. How much intermittency of solar and wind can be fulfilled by hy-dropower?3. Among the generation technologies available, which one should be a2riority?To address these questions, in this work we have analyzed the load curvesof the Brazilian national grid, covering the period between 1999 and 2018.Generalized Additive Models were used to better identify the trends. A 24%increase in electricity demand for 2030 was considered, and we have evaluatedfour different scenarios for how this demand could be fulfilled, investigatingits impacts and consequences.
2. Methods
Data on the national system grid load and generation by source over theperiod 1999-2018 were investigated, with hourly resolution. As an example,the data for the year 2018 is shown in figure 1. Besides some analyses directperformed with the raw data, Generalized Additive Models[17] were used tobetter identify trends.In the last two decades, electricity consumption has grown as fast as 4%per year in Brazil, though since 2014 it has averaged only 1%. In this context,we found reasonable to assume a growth rate of 2% for the period 2018-2030,which is similar to the low growth scenario from “Empresa Brasileira dePesquisa Energ´etica” (PNE2030). At such a rate in 2030 Brazil’s annualconsumption of electricity should increase by 137 TWh and reach 689 TWh,against 551.8 TWh in 2018.The next step in our study was to project a load curve demand for 2030.Contour plots of the load power against year and day of the week, as we aregoing to show, indicated that in the last two decades the load peak has beenshifted from early evenings across the entire year to afternoons in Febru-ary/March. As there is no sign that this trend will be reversed in the comingyears, we fixed our targeted load curve for 2030 as 1.24 (689TWh/551.8TWh)times the load curve in 2018.Brazil has today about 85 GW of hydroelectric dams and 12 GW of run-of-river hydropower [18], in such a way that hydro is the only source witha nominal capacity higher than the annual load peak. As no major newgigawatt-scale plant is expected to come online until 2030, we decided toanalyze scenarios where total hydropower production per year do not exceedthe 2018 level, in such way that the new demand expected for 2030 needs tobe supplied by thermal, wind, nuclear and solar power or partially replacedby solar water heaters. In this way, one can evaluate how the expansion3 -
01 2018 -
02 2018 -
03 2018 -
04 2018 -
05 2018 -
06 2018 -
07 2018 -
08 2018 -
09 2018 -
10 2018 -
11 2018 - Date P o w e r ( G W ) LoadHydroWindNuclearSolarThermal
Figure 1: Evolution of power consumption in Brazil with hourly resolution. of these sources will affect the hydropower during an entire year, at hourlybasis.For example, when considering a two-times increase in wind energy pro-duction we used the wind generation curve shown in figure 1 multiplied bytwo, enabling the possibility to access the effects of this source could bring to4he system. Even though such an approach is quite simple, as one year hasabout 8760 hours, this method provides some statistical insight about dailyand seasonal trends when all different sources are combined.
In order to build projections, we need to take into account some funda-mental aspects of Brazil’s electricity mix. While Wind and solar power arewidely known by its intermittency characteristic, thermal power in Brazilwas historically provided by biomass, which is seasonal dependent on theclimate. Though biomass has been surpassed by Gas-fired thermal powerin the last few years [19]. Only two nuclear reactors are running in Braziland the share of nuclear is not expected to change significantly in the nextdecade, since there is only one reactor in the construction phase, which washalted in 2014.In this way, we are analyzing four scenarios, namely 2030x, where x isa,b,c, and d, which are not proposed to compare its feasibility, cost or toconsider the construction of massive infrastructure [20]. Indeed they areused as tools to identify the challenges the national grid is going to faceand the ability of hydropower to keep filling the gap between the supply ofintermittent sources and demand, which is fundamental to the reliability ofthe system.In the scenario 2030a it is assumed that only Wind and Solar will expandin order to supply the additional 137 TWh needed, which results in a veryaggressive scenario of development for these two sources. Scenario 2030b isbased on an expansion of the same two sources but includes also aggressiveconservation of electricity by widely deploying solar water heaters (SWH) inorder to replace electric showerheads [21, 22, 23, 24].To develop the 2030b scenario first we searched for a good estimate ofthe total energy consumption of showerheads in Brazil. Corrˆea da Silva etal. [25] have considered a value of 20 TWh/year in 2004, while the mostrecent estimate by EPE, previously mentioned in this work, was about 31TWh/year for 2017, including losses. In the higher hand, Cruz et al. [24] haveestimated that in 2020 this amount of energy should be near 55 TWh/year.On the other hand, Cardemil et al. [26] have estimated the showerheadload profile of an average day by considering an annual consumption of 33.7TWh, including losses for the year 2012. In this way, we found reasonable toestimate potential savings due to SWH at about 50 TWh per year in 2030.5ext, we looked for an estimation of the hourly consumption of electricitydue to electric showerheads. Based on localized measurements found in theliterature [26, 27] we choose to distribute the daily consumption using threeGaussian curves centered at 7, 12, and 19 hours. The result is described byequation 1, P sh ( t ) = 12 σ (cid:16) e − / t − σ ) + 5 e − / t − σ ) + 60 e − / t − σ ) (cid:17) , (1)where each Gaussian has a weight factor of 35, 5, and 60 percent, while σ = 1 . Source 2018 2030a 2030b 2030c 2030d
Hydro 407.0 (0.8 × )325.6Thermal 80.4 (1.5 × )120.6Wind 46.0 (2.5 × )115.0 (2 × )92.0 (2 × )92.0 (2 × )91.9Nuclear 15.7 (4.1 × )64.2 (1.7 × )26.6Solar 2.7 (26 × )69.6 (16 × )42.9 (16 × )42.9 (26 × )69.6SWH - - 50.5 - 50.5Total 551.8 687.7 688.5 686.5 684.8 Table 1: Production of electricity (in TWh) by source in 2018 and in the four differentprojections made for 2030.
By taking the hourly production curve in 2018 for Thermal, Wind, Nu-clear, and Solar, as well the model for conservation of electricity by SWH,we calculated what it would be the remaining power left to be supplied by6 - - - - - - - - - - - - Time P o w e r ( G W ) Figure 2: Electric showerhead consumption estimated for an entire year in Brazil, withhourly resolution. hydro. To remove any “noise”, clearing the visualization of trends, we fittedthe data for each year with a Generalized Additive Model (GAM) [17]. Thisprocess was performed by using the PyGam library, in Python. A Tensorterm was used with the day of the week and year fraction as features, with200 and 20 splines, respectively. The fitting obtained had a pseudo r higherthan 0.9 in all cases. In the next section, we analyze the hourly load andcompare the results obtained in each of our scenarios.
3. Hourly load analysis
In figure 3 it is shown a contour plot of load power in the Braziliannational grid against weekday and year, with hourly resolution. The well-7
Weekday (0=Monday)
Y e a r
P o w e r ( G W )
Figure 3: Evolution of power consumption in Brazil with hourly resolution. marked load reduction in the second half of 2001 happened when hydropowersupplied more than 90% of Brazil’s electricity needs, and it is well known as adrought triggered crisis [9] as we mentioned before. As the total consumptionper year has increased by about 60% in the period shown in this figure, it isquite difficult to observe effects due to the 2014 crisis. In this way, to remove“noise” and better identify the trends, we have fitted the data for each yearwith a GAM and normalized the power by the maximum load in that year.These results are shown in figure 4.As one can see, the load pattern over a week has changed significantlyduring the last 20 years. In the early 2000s, the load peaks were foundexclusively in evenings and included a seasonal dependence. One possibleexplanation for this trend is the widespread use of showerheads, which are byfar the first source of heat for the shower in Brazil and consume a significant8
Weekday (0=Monday)
A n o
F r a c t i o n o f y e a r p e a k p o w e r
Figure 4: Evolution of power consumption as fraction of annual peak power. amount of electricity in a quite narrow time range [24].After 2001 the country experienced a period of strong economic growthand except by the year 2009, when the country was affected by the globalfinancial crisis, the electricity consumption increased by about 4% every yearuntil 2014. As one can see in figure 4 the load demand pattern has changedfrom early in the night to summers (February/March), mostly in the after-noon. This can be related to the increased share of the services sector inthe Brazilian economy, as well as the widespread adoption of air condition-ing systems. On the other hand, blue regions in figure 4 denote loads near50% of the year’s peak and may indicate a relative reduction in industrialactivity. In this sense, dark blue regions are observed in late evenings troughout the second half of 2001, as well on Mondays and Sundays of periods withconstrained economic activity (2009, 2014 and 2015).9 .1. 2030 Scenarios
The four scenarios mentioned in table 1 were evaluated assuming the 2018generation load for each source, multiplied by a factor, as indicated in thetable 1. Next the generation from wind, solar, thermal, nuclear and SWHwere subtracted from the demand load to estimate what is the load profileleft, which should be fulfilled by hydro power. The results are shown in figure5.
Day of week (0=monday)
Y e a r
P o w e r ( G W )
Figure 5: Simulated hydro power demand for 2030 scenarios and 2018 hydro generation.The scenarios 2030a, 2030b and 2030c have exactly the same amount of hydro powergenerated in 2018, while 2030d has 80%.
The contour plot should be carefully interpreted. These are the plots of10he GAM obtained by fitting the data. The red color, for example, representsa higher chance to have this hydropower demand. So, one may say that inscenario 2030a the peaks are more concentrated in early evenings duringsummer, while in 2030b it is more distributed in the afternoon once thisscenario has lower solar photovoltaics, which is likely to peak in this period.Brazil has today about 85 GW of hydroelectric dams and 12 GW ofrun-of-river hydropower [18]. Even though our scenarios do not consideran increase in annual hydropower production, in all of them hydro needs toprovide higher power for more time, mainly at the end of the summer (Febru-ary/March). For example, in 2018 the Hydropower generation in Brazil deliv-ered more than 70 GW for only 12 hours, while this number reaches 429, 289and 360 in the scenarios 2030a, 2030b and 2030c, respectively. This result in-dicates an increased risk for the national grid, once the maximum hydropoweravailable is often constrained due head loss in the reservoirs, which has al-ready played a central role in the last energy crisis in the country. As Huntet al [10] have discussed, in 2012 the operator of the national grid have takensome “optimistic decisions”, like “to spill some of the water in the FurnasReservoir wasting hydroelectric potential, to increase peak generation in theGrande and Paran´a Rivers, and reduce thermoelectric generation”. This kindof management has worsened the energy crisis experienced in 2014-2015. Inthe scenario 2030d hydropower should supply more than 70 GW by only 10hours. In such projection hydropower would supply only 80% of what is hasgenerated in 2018, and the results indicate that such aggressive conservationshould be necessary in order to keep about the same level of power demandfrom hydro experienced in 2018.On other hand, during winters hydropower could operate most of thetime at reduced power compared to 2018, though another problem emergesas reduce hydropower below 25 GW may not be practical because someminimum water flow is required. Additionally one may expect that higherpenetration of intermittent power sources will increase the rate at whichhydropower increases and decreases. To quantify the trends mentioned abovewe show in figure 6 histograms of expected power from hydro (number ofhours operating at each power) and power variation (number of times thateach variation in hydropower would be required) for our scenarios comparedto 2018 data.As one can see, in all scenarios there is a significant shift in the histograms.In the first three of them, as we mentioned before, the number of hoursin which hydropower operates above its 2018 maximum (70 GW) increases11 N u m b e r o f H o u r s Power (GW)
10 5 0 5 10050010001500
Load variation (GW/hour)
Figure 6: Histograms of load (left column) and load variation (right column) to be suppliedby hydro in our scenarios compared to 2018 data. significantly. On the other hand the number of hours operating below the2018 minimum hydropower (20 GW) also increases. At right in figure 6 onecan see that in 2018 the hydropower production barely oscillated more than 5GW in one hour, while in the scenario 2030a it could be required sometimesto increase this production by 11 GW/hour. The scenario less aggressiveconcerning such ramps is the 2030c, where we have less solar and wind andmore nuclear than in the scenario 2030a.12 . Discussion
As Brazil does not have any gigawatt-scale hydropower plant in the con-struction phase, we found it reasonable to evaluate scenarios, where invest-ment will be made in other sources, such as Wind, Solar, Nuclear Thermal,or electricity consumption is avoided due conservation introduced by solarwater heaters. When hourly load and production by these sources are takeninto account, the outcome achieved in each scenario and its impact in themanagement of hydropower could be evaluated.In 2018 the highest production of hydropower was concentrated in after-noons between February and March. A massive investment in solar powerresulting in 26 times more peak power than the 2018 level (2030a and 2030d)would result in shifting hydropower peak for early evenings, and the lowesthydropower demand should be experienced during winters, on Saturday andSunday mornings, at levels that may not be practical. Often the precipita-tion levels in many regions may be quite high until June or July, in such away that this time is likely to be the period where reservoirs are full. In otherwords, they would be full when lower power is demanded, increasing the riskto have water spilled and to have a significant head loss of the reservoirs inthe following summer.More conservative scenarios such as 2030b and 2030c, where solar poweris increased by a factor of 16 would make hydro to operate at its highestpower levels in afternoons and evenings, but the effect on Saturday andSunday mornings during winter are attenuated. It is especially interestingto compare these two scenarios once both have the same amount of Windand Solar, but the first uses SWH while the second considers an increase innuclear power.Martins et al. [28] have investigated also the financial and economicalaspects of SWH and payback time for a small scale system of only 4 yearswas obtained, which is compatible with the value estimated by Naspolini etal. [23]. On the other hand, Cruz et al. [24] have calculated that for about17.9% of the Brazilian households it does make sense from the economicalpoint of view to invest in SWH, while only 6% of the residences have an SWHsystem in 2018. However, such economic benefits are often underestimatedbecause in practice the price of electricity at peak hours is subsidized inBrazil [29, 23]. In this context SWH could be an important alternative toreduce the demand of electricity and reduce the need to expand thermal ornuclear power. 13inally one may conclude that increasing the share of intermittent renew-ables in the grid are very likely to affect dramatically the hydropower systemin Brazil. An increase in the peak hydropower, longer periods demandingmore than 70 GW or less than 25 GW of hydropower are very likely, and itseems likely to impose a challenge to the Brazilian national grid. To enhancethe reliability of the electricity supply without increasing the storage capac-ity seems not possible, though we estimated that hydropower can keep fillingthe gap between supply and demand until certain increased penetration ofSolar and Wind. Among the technologies readily available for large scaledeployment, SWH seems a feasible one that could significantly contribute tothe reliability of the Brazilian national grid.