Solar energy policy to boost Brazilian power sector

2.2.1 Auctions in Brazil.

To discuss solar energy auctions, auction mechanisms within Brazil must be assessed. Auctions for power generation contracts began in Brazil after the process of restructuring the electricity sector in 2004. They are part of a new regulatory framework (Dutra and Menezes, 2005), in which the following triple objective was required: efficient contracting mechanisms, security of supply and universal access.

The structure of the electricity market in the regulated environment is as follows:

• the distributors of electricity reveal the forecasted demand and communicate it to the Ministry of Mines and Energy (MME – Ministério de Minas e Energia);

• MME organize auctions for generation capacity; and

• distributors arrange contracts for the forecasted electricity demand with auction winners; winners are selected based on their ability to offer the lowest energy costs.

In the regulated contracting environment between 2004 and April 2018, 67 electric power auctions were carried out in Brazil, resulting in the agreement of 26,544 electricity contracts. Auctions took place through six different modes, namely, existing energy, new energy, structuring, backup energy, tuning and alternative sources. Table 2 explains the main features and presents selected results of the auctions occurring in Brazil in this period. It is important to note that the restructuring of the regulatory framework, of which the auctions are a fundamental part, was a response to the power shortage that the country experienced between 2001 and 2002, the so-called Apagão (blackout), where the electricity supply was unable to meet electricity demand due to insufficient investments in the generation and transmission systems.

Figure 1 shows all the valid auction contracts by energy source, showing supply between 2014 and 2030; there is a noticeably higher contribution of fossil fuels in the early years, which declines in the mid-2020s, with a reduction in coal and the elimination of the oil-based electricity supply. The hydropower contracts both within and outside the Amazon region are of longer duration. Contributions from solar PV remain low.

Five auctions occurred in the lead-up to February 2019, with the potential deployment of utility-scale solar energy; these were divided into two phases, the first with three auctions in 2014 and 2015, and the second with two auctions in 2017 and 2018. Between these two periods, an uncontracted mechanism took place to avoid discouraging future solar energy investments. At that time, the previous contracted project faced a significant challenge, with estimated costs for imported equipment during the plant-planning phase at an exchange rate higher than that agreed during the initial purchase of equipment. The exchange rate risks were not well assessed in the initial phase, causing some of the planned plants to experience delays or in some cases not to be built at all. The response of the power regulator was to cancel the contracts for those plants which could not be built.

Table 3 presents a synthesis of the auctions for utility-scale solar energy occurring in Brazil up to June 2019. Each column represents an auction process; LER is the Portuguese acronym for Reserve Energy Auction and LEN for New Energy Auction. This difference between the categories of an auction is one of the signs of maturity of the source in the Brazilian electricity market; in the first phase, the solar energy was sold to the system as an energy reserve. In the most recent auctions, the distributors bought the solar energy as electricity from other sources. In Table 3, it is still important to highlight the relative stability in the number of qualified proposals: between 315 and 493. It is worth noting that the price of energy in dollars per megawatt-hour reduced sharply, from US$82–90/MWh in 2014 to US$35/MWh in 2018. The reduction in the initial price increased from 17.9% to 62.2% between the first and the fifth auctions; this represents the linkage – from the one side, the expectation by the government of the cost of the technology, and from the other side, the expectation by the market of profits and risks.

2.2.2 Net metering regulation in Brazil.

In April 2012, the regulatory agency, Agência Nacional de Energia Elétrica (ANEEL) published the normative resolution RN 482/2012 introducing a system of net metering in the Brazilian power system, as a mechanism for incentivizing distributed generation (ANEEL, 2012). This system uses a bidirectional counter installed by the distribution companies, allowing consumers who produce power to inject their surplus energy into the grid and balance it with consumption at times of lower production. Energy bills, however, never reach zero because of a minimum payment awarded to the distribution companies for the costs of electricity availability. This minimum established by RN 414/2010 is very important in defining the size of installed rooftop systems (Vilaça Gomes et al., 2018).

One challenge faced by this mechanism was the effect of value-added tax (VAT or ICMS as the Portuguese acronym, for tax on the circulation of products and services) on the electricity provided to the grid through rooftop systems. This bottleneck effect is cited by both Vilaça Gomes et al. (2018) and Ferreira et al. (2018). The solution to this problem involves ANEEL, the 27 states of the country receiving the VAT and the National Financial Policy Council (CONFAZ). In 2015, these boards published Agreement 16 allowing the states to apply VAT to the balance of energy; 23 of the 27 states adhere to this Agreement (CONFAZ, 2015).

ANEEL’s RN 687/2015 supported incentives for a distributed system using net metering, extending the compensation deadline to 60 months, increasing the limits of installed capacity and the creation of three new figures (ANEEL, 2015). One figure is remote consumption when production and consumption can occur at different sites (produced by the same owner/supplier) within the region of the same distribution company. The other two possibilities are multiple consumption units and shared generation: the first case involves one producer and several consumers and the second has multiple generators that can compensate each other in terms of production and consumption before to trading with the grid (Sugawara and Nikaido, 2014).

The number of PV facilities grew rapidly in 2017, despite national economic crises, and the number of installed units increased by 168% and 141% in 2018. Figure 2 brings a comparison between the GDP evolution and the PV distributed installations, the figure shows that the increase in solar distributed generation is high despite the crisis. On 3 June 2019, there were 78,936 PV distributed generators registered on the ANEEL database, considering all sectors and all generation modes (site consumption, remote consumption, shared generation and multiple consumption). There were 58,510 residential PV panels connected to the grid, with the majority (51,138) located at the consumption site. There were some remote PV units for self-consumption (7,197), shared generation (149) and multiple consumption (26) (ANEEL, 2019a).

Future projections by ANEEL (2017) point to the continuity of this exponential growth, predicting that 886,700 micro generators will be connected to the network in 2024, 91% of which will be residential and the remaining 9% in commercial buildings (ANEEL, 2017). In 2018, the commercial sector represented 17% of generators and accounted for 43% of PV contributions from distributed generation (ANEEL, 2019a).

3.2.1 Energy system model to suggest solar targets.

Energy modelling is widely used to explore how different policies may affect energy systems, namely, in terms of the expansion of renewables, emissions reductions and costs; modelling, therefore, has hEigh importance for energy planning. In this paper, energy system modelling will be used to identify future solar targets for Brazil, derived from cost-effective criteria and taking into account the characteristics of the energy systems. The expected development of the technical–economic characteristics of the energy technologies is also considered, namely, through cost-curves.

A TIMES model was used, which is a dynamic linear optimization bottom-up model generator, developed by the Energy Technology Systems Analysis Program (ETSAP) of the International Energy Agency (IEA). The ultimate objective of the TIMES model is the meeting of energy service demands at minimum cost across the total system (i.e. net surplus maximization), subject to technological, physical and policy constraints. The model computes the energy demand/supply equilibrium by making simultaneous decisions regarding equipment investment, primary energy supply and energy trade.

Energy-economic-environmental models have been used widely for the development of energy and climate policies (Fortes et al., 2014). Governments have adopted tools of this type in relation to the security of energy supply, mitigation of climate change and/or exchange costs (Østergaard, 2015). In general, these models have been used to help policymakers with decisions strongly linked to climate change (Simoes et al., 2015).

The cost-effective potential of the Brazilian solar energy system was assessed through a new TIMES-MARKAL based model named TIMES_BR_light, developed and calibrated for the Brazilian case. TIMES_BR_light refers to one region – the whole country and takes 192 timeslices (24 h of a typical day of each season, accounting for a weekday and a weekend day). The energy system consists of the primary sector, with mining and international trade, the transformation process (eight types) and final consumption by sector, with 202 processes. The final energy demand for the residential sector and other economic sectors (industry, commerce, public sector, transport and agriculture) are estimated exogenously to the model.

This modelling exercise provides an optimal technological deployment solution with the minimization of total system costs in line with the research time horizon. The most cost-effective solution to the power supply is provided to accomplish the projected energy demand. The results of the amount of electricity generated by source (PV centralized, PV distributed and CSP) in this optimal solution are the basis of the solar targets assessed in this paper.

The assessment considers a scenario of a limitation in emissions in the power sector, equivalent to the reduction in total emissions in the NDCs in the Paris Agreement, i.e. a reduction of 43% in 2030 compared with 2005. It is worth noting that this reduction is more ambitious than the projected reduction reported in the official NDCs for the power sector (emissions in 2030 of 20 Mton CO_2e instead of 73 Mton CO_2eq) (EPE, 2016a). The goal of this paper is to explore the cost-effectiveness of solar energy contributing to solving the current Brazilian “triple challenge.” Solar is likely to be a bolder contributor to national emissions reductions, particularly in comparison with other power sources.

3.2.2 Incentives for solar deployment.

A specific type of incentive mechanism is suggested for each technological solar target option (PV centralized, PV distributed and CSP) because of the following reasons. For centralized PV, it is argued that the best solution is the continuance of auctions on a regular basis, with a substantial increment of the capacity contracted in each auction. The previous auction results showed that private investment is available in the market and that competitiveness caused a significant reduction in the cost of electricity.

Solar thermal power on a large scale is still a costly technology, and therefore a competitive mechanism is suggested, where tenders are used to select the best offers in terms of investment price, technical quality and location. The winners agree on contracts directly with consumers in a free market environment and the government pays the difference between the higher technological costs and the alternative costs, for example, gas. This approach is similar to FIT but with a competitive approach. The rationality behind this mechanism is the increase of technological diversity in electricity deployment using the huge endogenous potential in an innovative, efficient, storable and dispatchable renewable source.

For distributed PV electricity generation, the enlargement of an existing green loan programme run by an official financial institution is suggested, given that grid, parity has already been achieved and that the technology requires a degree of support on a temporary basis to cope with the upfront costs. This solution could provide small investors who can help to increase the security system, without GHG emissions and at the national industrial scale the national industry.

These incentive mechanisms are analyzed in terms of cost, ability to adhere to stakeholder requirements and regulatory feasibility within the context of Brazilian energy policy.

3.3.1 Definition of the most cost-effective levels of deployment of solar energy.

This section presents the most cost-efficient levels for solar energy deployment, taking the CO₂ emissions cap as explained previously. Utility-scale solar PV generation will likely be 273 PJ in 2030 and 808 PJ in 2040 from an optimization perspective, accounting for 18% of the total electricity production in 2040, which demonstrates the competitiveness of solar energy.

Figure 3 shows the total electricity generated by energy source; the majority of fossil fuels are phased out, except natural gas, which has a modest increase. Wind power increases until 2030 then stabilizes because of the maximum potential reached. Bioelectricity increases and solar power too. This last source presents the strongest growth.

The results of the model indicate that the cost-effective solar options are 92T Wh in 2030 (76 TWh PV-US and 15 TWh of PV-distributed) and 324 TWh in 2040 (224 TWh of PV-US, 111TWh of CSP and 40 TWh of PV-distributed). The solar technology that appears with the strongest and earliest is utility-scale PV, due to the lower investment cost and greater potential. With the expected increase in the demand for electricity, CSP technology would be used in 2040. The distributed options are not as expensive as CSP, but have a limited potential considered after the literature review for the construction model; for this reason, the participation of this technology is limited, even with a greater electricity demand in 2040.

3.3.2 Policy incentive mechanism assessment: auctions for utility-scale photovoltaic.

Considering the optimization model results, to achieve a cap of 43% emissions reduction in the Brazilian power sector (compared with emissions in 2005), solar utility-scale PV may contribute significantly, as shown in Figure 2. However, to reach the most cost-efficient deployment of the technology, an incentive mechanism is suggested, the acceleration auction.

Considering the rate of auctions similar to 2018 (i.e. 3,500 GWh of additional electricity per annum, assuming that operation begins four years after the bid is won), 36.7 TWh of electricity in 2030 and 66.5 TWh in 2040 might be expected. These values are insufficient to meet the cost-effective potential, which means there is room for a more ambitious scheme.

The generation proposed here is an acceleration in auctions. In 2030, there should be at least 38,771 GWh more than the amount currently contracted, with a sustained rate of new auctions continuing to generate 3,500 GWh/year. This means that an additional 4,846 GWh per year in auctions in the period 2019-2026 and an additional 10,505 GWh per year between 2027 and 2036 would be required. These values would be substantially higher than the contracted 2003 GWh cited in the 27th LEN ( Table 3). However, the total amount of viable projects presented in the 25th and 27th LEN would be 80,103 GWh, so the current market potential could be said to be under-exploited (EPE, 2017c).

Giving clear signs to the market, the acceleration could be carried out in two phases, the first one between 2019 and 2026, with a yearly addition of 5,585.57 GWh of new contracts above the continuity of the actions in the same rhythm as the 2018 auction.

In a second phase, the acceleration would be greater, and the new yearly contracts would need to be 11,320.70 GWh above the rhythm of the 2018 auctions in the period between 2027 and 2036. The contracts considered here would made four years before the begging of the technological operation; then with this scheme, in 2040 the more cost-effective potential of the technology would be deployed.

3.3.3 Policy incentive mechanism assessment: competitive tenders for concentrated solar power.

Examples of feasible free-market tender options might follow the French model, with three clear contract deadlines: 2025, 2030 and 2035. The use of direct power purchase agreements for thermal solar power between consumers and generators would be considered. Competitive tender processes should be used to select the best technical – economical projects. Such projects should be eligible for financial support from the government or from the system as a whole by a sectoral funds review, for example, to compensate for higher technological costs. Regarding these results, a scenario of 186 plants of 100 MW each could be considered.

Eligibility for participation in the competitive tender scheme would depend on the existence of a supply contract between the generator and the consumer or distribution company. The advantage for the generator of participating in this tender would be complementary revenue, in addition to electricity payments. Mechanism costs would be estimated in line with a 55% reduction in CSP electricity costs, equivalent to the forecasted reduction of investment costs between 2016 and 2040. Assuming electricity costs from CSP of US$90.00/MWh or R$335.00/MWh and electricity costs from natural gas of R$213.00, as seen in the latest auction (27th LEN), this would result in a difference of R$122.00/MWh or 35% in relation with the cost of CSP.

In the initial phase, with a deadline application in 2025, this represents R$939m per year, an amount similar to the national coal subsidy paid in 2018 for electricity produced using national coal (R$771m in 2018) (CCEE, 2018b).

The logical model of the three phases for the adoption of this incentive mechanism suggested here is that the early adopters would have the greater advantage, but low participation would be expected; in the next phase, the advantages would be lesser but higher deployment would be expected. The generators would receive the value of the power contract in the free market, plus the difference between the cost of energy produced in the CSP plant and the cost of energy in the most recent natural gas auction. The three phases are different in terms of incentive to new CSP generators: in the first phase, the incentive would decrease and the CSP generator would receive the total difference plus 10% of the difference or 110% of the difference. In the second phase, there would be just the difference, and in the third phase, there would be 90% of the difference. A second difference of the three phases would be the ambition of the mechanism in each one: the first phase would see a small participation of CSP generators; the second one, on the other hand, would assume a high potential of investments; and the third phase would be almost all residual investments.

Compared to the French model, this proposed new mechanism presents some differences, namely, only solar thermal power projects would be allowed, without size or sectoral differentiation. The criteria to calculate the subsidy might follow different options. In the French case, this calculation considers the number of hours when the spot price is positive or zero, multiplied by the applicant’s electricity price (in €/MWh), minus the benchmarked electricity price of all solar electricity producers. There is still a specific point of participative investments with a bonus of 3€/MWh.

The proposed design of a new incentive mechanism such as competitive tenders allows public policy not just to define which technology would receive the incentive and the associated timelines but also to decide where the load would be installed. It is argued that the northeast Brazilian submarket would be the most appropriate region for the development of this incentive mechanism because it has the highest irradiation and because there are some congestion costs. Besides that, this is one of the poorest parts of the country, one of the most populated semiarid regions in the world, and the economic basis of agriculture and cattle-raising tends to suffer greatly from climate change (Orsato et al., 2017).

3.3.4 Policy incentive mechanism assessment: green loans for distributing photovoltaic.

An official bank in Brazil called Banco do Nordeste launched a loan programme in 2016 with a reduced interest rate and long payback period (12 years), named Fundo Constitucional de Financiamento do Nordeste (FNE Solar). Between May 2016 and October 2018, it funded 770 projects at a total of R$124m, for non-household customers (Banco do Nordeste, 2018). The same programme is now open for residential customers.

The potential for solar distributed generation derived from the optimization energy model for Brazil is estimated as 16,667 GWh in 2030 and 44,722 GWh in 2040. Finding the means of realizing this potential could be a way to alleviate poverty in the northeast region of Brazil. This idea is aligned with the arguments provided by Nobre et al. (2019) that it is possible to change the paradigm and use in a better way the most abundant resource of the region: the sun.

As information about the technical characteristics of the projects was not found, to estimate the total of PV-distributed installed capacity and electricity generated, the price of distributed installations in the commercial sector was used, of R$4.31/Wp in June 2018 (GREENER, 2018). With total cost and cost per watt, the estimated installation supported by the FNE-Sol program is 28.87 MW of installed capacity currently in operation. Using a 20% availability factor, 51 GWh of annual electricity production was found at distribution level in the 770 existing projects, which means on average a cost of R$2.44m/GWh in the 770 existing projects.

To accomplish the cost-effective potential in 2030, the additional electricity generated annually should be 1,385 GWh, requiring R$3.4bn available for a loan each year until 2030. For 2040, a total of R$6.85bn of financial support would be required.

Considering that it is a loan, the cost of the mechanism must be calculated based on the difference in the interest rate. This loan has a lower interest rate (1.67% yearly of fixed-rate) than the average interest rate in the country (6.5% yearly). If solely this difference is considered between both the aforementioned interest rates and transaction costs are not considered, then the cost of the mechanism in 2016 was about R$1.9m and in 2040 the annual cost would be R$3.8bn. The loans are provided by a public regional development bank with public resources for private companies.

Banco do Nordeste affirmed that the demand rate for loans in the FNE Sol program was below the available resource for the scheme and should the demand for such loans increase, the bank could redirect resources, from other programmes into this one (Banco do Nordeste, 2019). In the future, with market growth, other financial options for distributed generation investment might be designed.