- Email: [email protected]
The stochastic effects on the Brazilian Electrical Sector
Energy Economics 49 (2015) 328–335
Contents lists available at ScienceDirect
Energy Economics journal homepage: www.elsevier.com/locate/eneco
The stochastic effects on the Brazilian Electrical Sector Pedro Guilherme Costa Ferreira a,⁎, Fernando Luiz Cyrino Oliveira b, Reinaldo Castro Souza c a b c
Brazilian Institute of Economics (FGVIBRE), R. Barão de Itambi, 60 Botafogo, Rio de Janeiro, Brazil Industrial Engineering Department, Pontifical Catholic University of Rio de Janeiro, R. Marquês de São Vicente, 225 Gávea, Rio de Janeiro, Brazil Electrical Engineering Department, Pontifical Catholic University of Rio de Janeiro, R. Marquês de São Vicente, 225 Gávea, Rio de Janeiro, Brazil
a r t i c l e
i n f o
Article history: Received 9 July 2014 Received in revised form 11 March 2015 Accepted 13 March 2015 Available online 22 March 2015 JEL Classification: O13 Q47 Q28 Q25 C59 O54 Keywords: Brazilian Electrical Sector Expansion planning Operational planning Spot price
a b s t r a c t The size and characteristics of the Brazilian Electrical Sector (BES) are unique. The system includes a large-scale hydrothermal power system with many hydroelectric plants and multiple owners. Due to the historical harnessing of natural resources, the National Interconnected System (NIS) was developed outside of the economic scale of the BES. The central components of the NIS enable energy generated in any part of Brazil to be consumed in distant regions, considering certain technical configurations. This interconnection results in a large-scale complex system and is controlled by robust computational models, used to support the planning and operation of the NIS. This study presents a different vision of the SEB, demonstrating the intrinsic relationship between hydrological stochasticity and the activities executed by the system, which is an important sector of the infrastructure in Brazil. The simulation of energy scenarios is crucial to the optimal manner to operate the sector and to supporting decisions about whether expansion is necessary, thus, avoiding unnecessary costs and/ or losses. These scenarios are an imposing factor in the determination of the spot cost of electrical energy, given that the simulated quantities of water in the reservoirs are one of the determinants for the short-term energy price. © 2015 Elsevier B.V. All rights reserved.
1. Introduction The Brazilian Electrical Sector (BES), which is the ninth largest electrical sector in the world in terms of energy generation, produces approximately 470 TW h (OECD, 2012). In 2011, the generation, distribution and transmission sectors experienced billings of approximately US$6 billion. Electricity reaches over 99% of Brazilian homes. Due to the historical harnessing of natural resources for energy generation, the National Interconnected System (NIS) was developed outside of the economic scale of the BES. The central components of the NIS enable energy generated in any part of Brazil to be consumed by consumers in far away, considering certain technical configurations. This interconnection between regions produces an enhanced utilisation of resources, which results in a large-scale complex system that is controlled by robust computational models that are used to support the NIS' planning and operation. The unification of a nationally interconnected system requires billions of dollars from investments (public and private) and the market structure (monopolies and oligopolies) for hydraulic generation, ⁎ Corresponding author. E-mail addresses: [email protected] (P.G.C. Ferreira), [email protected] (F.L.C. Oliveira), [email protected] (R.C. Souza).
http://dx.doi.org/10.1016/j.eneco.2015.03.004 0140-9883/© 2015 Elsevier B.V. All rights reserved.
distribution and transmission. As a result, several problems and challenges emerge that require decisions on different time scales. These problems involve decisions that are directly related to the three functions of the BES: (i) expansion planning, (ii) operational planning and programming and (iii) determination of energy spot price. These functions are executed by different agencies: the Energy Research Company (EPE), the National Electrical System Operator (ONS) and the Board of Electrical Energy Commercialisation (CCEE). The following questions emerge in this environment: should an investment be made to increase the capacity of the system or would it be better to wait for a time of greater expansion in economic activity? How much thermal and hydroelectric energy should be generated to meet current demand? When is the right time to conserve water and use fossil fuels? What is the required spot price of energy to finance all production factors and ensure that the affordability tariff is paid? Responses to these questions, which pertain to the daily activities of the BES, are not trivial and require extensive planning and synchronised management. Stochasticity exists in the following three functions of the BES. In addition to the introduction, this article is organised as follows: Section 2 contains the history of models adopted by the BES. Section 3 describes the sector from the perspective of the new model by identifying relationships between the three supporting functions and hydrological
P.G.C. Ferreira et al. / Energy Economics 49 (2015) 328–335
stochasticity. Section 4 includes final considerations and relevant contributions. 2. BES historical description The origin of the electrical energy supply crisis in Brazil is related to four main causes: (a) exhaustion of the state-owned market model of finance and structure, which is responsible for the expansion of the sector since the 1960s, (b) failures in planning the transition from the stateowned model to the private model, (c) contractual and regulatory problems and (d) the lack of coordination between governmental agencies (Pires et al., 2002). The collapse of the state-owned model can be attributed to two main reasons: the fiscal crisis of the state, which caused the end of the investment capacity of the union at the required levels for expansion of the system (the companies were predominantly state-owned), and an inadequate regulatory regime, which did not promote efficiency and low generation costs as the tariffs were regulated in the generation, distribution and transmission segments. Other factors also contributed to the fiscal crisis of the state, which began in the 1980s and reduced the quantity of union resources for investments. First, the marginal cost of expansion of the sector increased as new hydroelectric basins were situated further from the consumption centres. For this reason, additional resources had to be invested to produce an equivalent generation capacity. Second, the actual values of tariffs, whose price levels did not reflect increased sector costs, deteriorated. In addition to their stability throughout Brazil, the tariffs were used as an inflationary control tool. This process culminated in the decapitalisation and subsequent default of several sector agents. Last, with the consolidation of the democracy and the advent of monetary stability, social demands imposed the need for the government to develop better criteria for the application of union resources (Pires et al., 2002). Based on these reasons, investments in the state-owned companies were not sufficient for satisfying the intensifying demand in Brazil. Thus, a large number of BES expansion projects were halted and/or did not adhere to previously established schedules, which worsened the financial situation of the sector due to increased costs caused by expanded project timelines. From the regulatory point of view, the lack of stimuli to seek productive efficiency prevented companies from establishing incentives to reduce costs. According to Schaeffer et al.(2003), the tariffs were equalised throughout Brazil in the 1970s to stimulate energy development in certain regions by requiring that surplus and deficit companies compensate each other through transfers, gains and losses from their individual efforts. In 1993, at the beginning of the privatisation process, the tariffs were fixed by the electrical distributing utilities to justify the need for companies to establish appropriate tariffs for their markets and to achieve satisfactory levels of profitability. See more about tariff policy of the Brazilian electric power sector in Santos et al. (2013). Regarding the failures in the transition from the state-owned model to the private model and contrary to what was expected, private companies did not invest in the expansion of generating facilities at the beginning of the privatisations in 1995. According to Fernandes et al.(2005), the most important effect of these privatisations was a rapid return to tariff levels (which had been obsolete until then) to increase the attractiveness of private companies. On February 13, 1995, the privatisation and reformation process of the sector began with Law No. 8987, which is named the “Law of Concessions”. According to Pêgo and Campos Neto(2008), in addition to creating conditions for greater participation of private capital, the new law introduced competition in the construction of new projects through regulation of the concessions bidding system for utilities that were previously exclusive to state and federal utilities. In 1996, the BES Restructuring Project was initiated. The primary objectives of the project are summarised as follows: (i) the need to implement the vertical disintegration of electrical energy companies,
329
that is, to divide them into segments of generation, transmission and distribution; (ii) to incentivise competition in the generation and commercialisation segments and (iii) to regulate the distribution and transmission sectors of electrical energy, which were considered natural monopolies under state regulation (CCEE, 2012). The need for the creation of a new regulatory and inspection agency for all sector relationships was identified (Brazilian Electricity Regulatory Agency — ANEEL). The agency, which began to operate in August 1998, was expected to control the operation of the electrical system, from one operator to the national electrical system, in an integrated manner (ONS) and in an environment for buying and selling electrical energy (Wholesale Electricity Market — MAE); the MAE commenced in 2000 with many restrictions. The authors Goldemberg and Prado (2003), highlight that the failure of the electrical sector reform (free market model) was attributed not only to the lack of external resources or the political resistance encountered by the government but also to failures in strategic management, coordination and planning of the electrical system, which was induced by the adoption of a reform that was grounded in the experiences of other countries and inappropriate for Brazil and its predominantly hydroelectric system. In 2001, due to failures of the reform and a lack of sustainability of the model and water problems (shortage of rainfall), the electrical sector experienced a significant supply crisis that culminated in a national electrical energy rationing plan, which affected all categories of consumers. As the crisis was insufficiently addressed by the government, measurement of the immediate results consisted of the controlling of consumption. The government concentrated on thermoelectric power plant construction projects and reinforced the budget for investments in state-owned companies. In 2003, the financial and distribution problems worsened, which caused the government to implement a programme to organise resources for the distribution utilities via the Brazilian Development Bank (BNDES). In addition to measures of short-term emergencies, a new institutional model (new model) was elaborated for the energy sector to correct the failures that caused the crisis and to focus on tariff affordability, universality of access and the resumption of energy planning (Table 1). The new model defined the creation of an entity that is responsible for long-term energy sector planning, the EPE; an institution assigned to permanently evaluate the safety of the electrical energy supply; the Electrical Sector Monitoring Committee (CMSE); and one to give continuity to activities related to the commercialisation of electric energy in the interconnected system: the Chamber of Commercialisation of Electric Energy (CCEE). Regarding the commercialisation of energy, two environments were instituted for the negotiation of energy trade contract: the Regulated Contracting Environment (ACR), in which the agents of generation and distribution of energy participate, and the Free Contracting Environment (ACL), in which the exporting agents of generation, commercialisation, importation and free consumers of energy participate (CCEE, 2012). The new electrical sector model provided a set of measures to be observed by the agents, such as the requirement to contract all demand by the distributors and free consumers, a new methodology for the calculation of backing1 for the sale of generation and the contracting of hydroelectric and thermoelectric power plants for better equilibrium between the backing and the cost of supply, as well as permanent monitoring of the continuity and security of supply to detect conjunctural imbalances between supply and consumption (CCEE, 2012). In terms of tariff affordability, the model established the purchase of electrical energy through auctions by the distributors in the regulated environment. It is considered the lowest cost criterion with the
1 Decree no. 5163/2004 establishes that selling agents must guarantee the backing of power and energy that is sold to 100% of their contracts. This guarantee must be proportioned per generating enterprise or by third parties through energy or power contracts.
330
P.G.C. Ferreira et al. / Energy Economics 49 (2015) 328–335
Table 1 Comparison of the energy models. Characteristics
Old model (until 1995)
Free market model (1995 to 2003)
New model (from 2004)
Financing Structure of businesses
Public resources Vertically integrated
Public vs. private relationship
Predominantly state-owned businesses
Market structure
Monopolies — non-existent competition
Consumers Tariff structure
Captive Tariffs regulated in all segments
Public and private resources Generation, distribution, transmission and commercialisation Coexistence between state-owned and private businesses Competition in generation and commercialisation Captive and free Regulated and free environment of contracting
Market
Regulated
Public and private resources Generation, distribution, transmission and commercialisation Opening and emphasis on the privatisation of businesses Competition in generation and commercialisation Captive and free Prices freely negotiated in generation and commercialisation Free
Expansion planning
Coordinating group for the planning of electrical systems 100% of the market Prorated between buyers
Contracting Surpluses/deficits of the energy balance
National Council of Energy Policy (CNPE)
Coexistence between the free and regulated market Energy Research Business
95% of the market (until Dec. 2004) Liquidated on the MAE
100% of the market and reserve Liquidated on the CCEE
Source: (CCEE, 2012).
objective of reducing the electrical energy acquisition costs to be passed on to tariffs of captive consumers. 3. The BES under the new model Under the new model, the federal government defined a set of institutional agents to guarantee proper functioning of the sector aiming three desired objectives: (i) tariff affordability, (ii) supply security and (iii) universality of access (Tolmasquim, 2011). According to the author, the agents can be classified into three levels according to the legal nature of the entity and their institutional competencies: (i) agents that execute government activities, (ii) agents that execute regulatory activities and (iii) entities of private rights that execute special activities. Government activities are conducted by the National Energy Policy Council (CNPE), the Ministry of Mines and Energy (MME) and the Electrical Sector Monitoring Committee (CMSE). Regulatory activities are conducted by ANEEL. Operational entities conducted technical activities including sector expansion planning (EPE), operational planning and programming (ONS) and commercialisation (CCEE) (Fig. 1).
In addition to the inherent characteristics of the hydroelectric– thermal systems, such as the natural monopoly (transmission and distribution), grid industry, elevated capital intensity, maturation of long-term investments, large interconnections and uncertainties (hydrological, increase in demand and price of fuels), the BES exhibits unique characteristics. In addition to occupying the ninth place in terms of global energy generation and a large consumer market, more than 90% of the energy is generated by hydroelectric plants, and nearly all generation, transmission and distribution systems are nationally interconnected. Besides high capital intensity and a long investment maturity period, the time between the decision to build a hydroelectric plant and its effective entry into operation can exceed ten years. Other characteristics consist of sunk costs (or irrecoverable costs) due to the specialised activities connected to the sector (e.g., electrical energy transmission lines). Thus, the importance of the long-term planning conducted by the EPE is rather important. Another fundamental technical attribute of the sector is its physical equilibrium, which requires the coordination of the system due to its strong interdependence. Generation of electrical energy can occur through various technologies with varying costs and social–
Fig. 1. Institutional agents of the BES. Source: (Ministério de Minas e Energia, 2013).
P.G.C. Ferreira et al. / Energy Economics 49 (2015) 328–335
331
The electrical energy generation system in Brazil (approximately 105,000 MW installed) is mainly hydroelectric. In 2009, the participation of this energy source generated approximately 80,000 MW, including the Brazilian portion of the Itaipu Power Plant (7000 MW) The thermoelectric and thermonuclear sectors represent 18% and 2%, respectively, of the generating capacity (EPE, 2012). In a hydroelectric–thermal system, the future of available hydroelectric generation with no generating costs is uncertain; conversely, the known future of high-cost thermal generation and the need to determine the thermoelectric/hydroelectric allocation at each instant is evident. In contrast to purely thermal systems, for the operation of hydrothermal systems,3 in which the operational planning problem
can be simply resolved by establishing a pattern among the plants that minimises fuel costs,4 the decision is coupled over time, that is, a decision made today will have consequences in the future (Terry et al., 1986). For example, if significant hydroelectric dispatch exists prior to a dry period, a risk of high-cost thermal dispatches will exist in the future; conversely, a thermal dispatch prior to a wet period may cause overflows, which results in wasted energy (Fig. 2). Thus, a balanced operation5 of the system involves a compromise between depleting (using water) and not depleting (using thermoelectric plants) the reservoirs. The decision variable is the volume of water stored at the end of the operational period (final volume). This decision has an immediate cost, which is associated with thermal generation (immediate cost function—ICF), and a future cost, which is associated with the expectation of thermal dispatch and is denoted as the future cost function (FCF). The total cost comprises the sum of these costs. The optimal decision is obtained by cancelling the derivative of the total cost relative to the final stored volume (Pereira et al., 1998). As illustrated in Fig. 3, the ICF increases with final stored volume (higher thermoelectric generation). The FCF exhibits the opposite behaviour: with a higher final stock, there is a lower expectation for future fuel usage. The coordination of operational planning for the hydrothermal system can be designed as a large-scale optimisation problem, with temporal and spatial coupling of the dynamic, stochastic, interconnected and non-linear operations. Its solution requires the problem to be decomposed into a chain of coupled models that consider the long (probability of future storage of energy and expected future value of thermal generation), medium (annual contracts for reserves at the end of the contract) and short term (control of flow and hourly dispatch) planning (Lepecki and Kelman, 1985; Pereira and Pinto, 1982). To resolve this problem, the BES adopts a chain of coupled models that consider different planning timeframes: long-term, mediumterm, short-term and daily planning (Maceira et al., 2002). The coupling between models occurs through the FCF of energy operation. Specifically, operational planning is defined as the efforts to delineate system behaviour over a five-year timeframe by promoting the rational use of resources to guarantee quality and safety in satisfying market demand, operational requirements of the hydrothermal system and minimisation of operational costs. In the operational planning, the objective is establishing a shortterm operation of the hydrothermal system. It should provide operation decisions for the generation system, which are feasible for the transmission system and respect the goals established through operational planning. To cope with the models adopted in the SEB energy planning chain, algorithms from the Stochastic Dual Dynamic Programming (SDDP) are employed (Pereira and Pinto, 1985). This methodology uses the Benders decomposition technique (Benders, 1962) to develop optimal strategies for the operation of the interconnected subsystems, whereas the streamflows are treated using a periodic autorregressive model (Ferreira et al., Forthcoming). SDDP is used to determine the operating policy that minimises the expected value of the operational cost over a five-year timeframe. The strategy is confirmed through a simulation process that employs a series of affluent energies representative of each subsystem, which can be historical or synthetic. The hydroelectric generating facility is represented by an equivalent reservoirs approach, in which a set of hydroelectric plants are grouped into a single reservoir that receives, stores and discharges energy (affluent natural energy). BES is represented by four subsystems: southeast/midwest, south, north and
2 Only 3.4% of the electrical production capacity of Brazil is external to the NIS in small isolated systems, which are primarily located within the Amazon region. 3 Planning must also consider the multiple uses of water in the cascades, such as, the maintenance of navigation, irrigation, flood control and sanitation levels.
4 Additional factors that can exacerbate the complexity of the problem include energy losses, transmission limitations and initial operation costs (Pereira et al., 1998). 5 The system is balanced when sufficient energy can be generated to satisfy demand with a low risk of failure (Kelman, 2001).
environmental impacts. For example, the raw material of hydroelectric plants is the random inflow based on the rainfall regime, which results in heightened complexity for short (one month ahead) and medium (sixty months, or five years, ahead) term planning processes. Such task is performed by the ONS. Despite being an efficient system for planning supply and demand for electrical energy, differences exist between the energy effectively produced by the market and effectively consumed by the market. To resolve this problem, a short-term energy market, whose final objective is solving the disparities between supply and demand, is handled by the CCEE. The NIS was established due to the historical process of harnessing natural resources for energy generation. The NIS, which contains central components, enables energy generation in any part of Brazil to be consumed by consumers in distant regions,2 with certain technical configurations. The interconnection between the regions produces an enhanced utilisation of resources. This interconnection is derived from the notion that Brazil operates a hydroelectric system composed of water reservoirs and hydroelectric plants, which are planned such that they can take advantage of the pluviometric diversity in the existing basins. This measure ensures an important energetic gain for the Brazilian system. The coordinated operation of the NIS by ONS adheres to the general guidelines of the CMSE, which incorporate the objective of meeting the electrical energy consumption requirements for the system to ensure continuity of supply in the most reliable manner and with reduced operational costs. Thus, the rational use of resources must be planned with the objective of satisfying the current requirements of the system to determine the needs for expansion (EPE) with consequent investments. The considerations of the BES specificities in relation to the expansion planning, operational planning and the valuation of spot prices for energy, demonstrate the importance of analysing the sector. These decisions are primarily rendered in an uncertain environment and require systematic decision-making processes, especially with regard to future perspectives. The next three subsections address the expansion planning, the operational planning, the unique determination of the spot price of energy and the identification of characteristics unique to each segment; these areas are related to the stochasticity inherent in the sector. First, the operational planning of the system is addressed, as this activity is directly related to the optimisation model (known as the NEWAVE by the BES) that supports the three activities addressed in this study. Secondly, the expansion planning and determination of spot prices of energy are addressed to demonstrate how stochasticity affects these activities. 3.1. Operational planning of the BES
332
P.G.C. Ferreira et al. / Energy Economics 49 (2015) 328–335
the optimal dispatch from the hydrothermal system, which considers the energetic, hydraulic and electrical aspects of the system. The short and extremely short timeframes (daily programming) allow for the consideration of known or deterministic flow values (Maceira et al., 2002). Last, as the considered time interval decreases, uncertainties related to flows/affluent natural energy (ANE)7 decrease and reach a deterministic extrapolation, and the level of details for representation of the system increases such that characteristics of the individualised systems are considered. 3.2. Planning of electrical system expansion
Fig. 2. Problem of operational decisions. Source: (ONS, 2012).
northeast, which represent the geographic regions. The thermoelectric cost is calculated from the operational costs of thermoelectric plants and the minimum and maximum generations. The NEWAVE model provides a FCF coupled with the short-term model at the end of the planning timeframe (Fig. 4). The objective is to reduce the computational effort required by the optimisation models; the models used for medium-term operational planning employ the grouping of plants into equivalent reservoirs of energy. Each NIS' subsystem (southeast/midwest, south, north and northeast6) is represented by an equivalent reservoir composed of the hydroelectric plants of the region. However, to confirm whether the operational policy obtained from the strategic decision model is feasible, the solution obtained for the equivalent reservoirs of energy must be divided into individual hydroelectric plants, that is, whether the plants comprising the equivalent system will be able to supply the quantity of hydroelectricity generated for the system calculated by the strategic decision making model (shortterm planning) must be analysed. The weekly (and daily) dispatch proposal is subsequently adjusted in the context of daily planning. This study focuses on the stochastic module that supports the NEWAVE planning model. In general terms, this model optimises water use and is dependent on stochasticity, represented by the scenarios generated from time series models, considered in the affluent energies module. These scenarios are generated from a periodic autoregressive model, from Box& Jenkins family, and the simulations are conducted with Monte Carlo Methods and Bootstrap, see Hipel and McLeod (1994) and Cyrino Oliveira et al. (2015). The short-term aspects of quality and satisfying demand involve fulfilling the goals for generation as defined in medium-term planning and verification of operational feasibility, in terms of restrictions of generation and transmission equipment. In this planning timeframe, the objective of the DECOMP model is to minimise the expected value of the total cost of the system, as detailed in the subsequent sections. When the study lead time is reduced to the short term and extremely short term (DECOMP and DESSEM-PAT, respectively), the system representation is refined to the hydroelectric plants by communicating their operational characteristics and hydraulic and energetic restrictions. For operational planning, the timeframe encompasses two hours with discretisation into hours or half hours (Fig. 4). The objective is to obtain
6 Currently, there is a discussion in the SEB to increase the number of subsystems by considering for incorporation into the NIS, for example, the basins of the Madeira, Tapajos and Xingu Rivers. However, only the four traditional subsystems are considered in this study.
According to Tolmasquim (2011), the relationship between the future and immediate use of available financial capital for expansion is similar to the operational planning involving a compromise between the immediate and future use of water for expansion planning. Agents need to decide between investing in the present and assuming the risk of idleness in the system due to the lower than expected growth of demand and postponing investment and risk rationing, as illustrated in Fig. 5. According to the logic of water use in the operation, the decision to expand system capacity decreases the stock of capital, with an immediate cost (known cost of expansion) and a future cost (estimated cost of deficit). Thus, the decision to “save” has a low immediate cost and a high future cost due to an increase in fuel consumption and rationing. Conversely, the decision to “invest” has a high immediate cost and a low future cost (Fig. 6)8 (Tolmasquim, 2011). Similar to the operation process, several alternatives for expansion exist, given a predicted demand in the case of planning. Each alternative is equivalent to a level of reliability R (x-axis), where the cost of expansion is dependent on demand and is unique to the level of reliability R desired by the planner. Similar to the operation, the solution to the planning process is given by optimisation techniques, with the level of reliability (R) as the decision variable and total cost (TC—sum of the costs of expansion and operation) as the objective function. The level of reliability (R*) and the expansion plan are optimal minimum total cost point (Tolmasquim, 2011). In the optimal plan, the derivative of the expansion cost (EC) with respect to the demand (D) is denoted by the marginal cost of expansion (MCE), which is associated with the cost of incremental load with an expansion of capacity. The derivative of the operational cost (OC) with respect to the demand is denoted by the marginal cost of operation (MCO), which represents the cost of meeting the incremental load without expansion of capacity. Thus, once the demand expands over time, the expansion of the SEB occurs when the MCO module is equivalent to the MCE. By expanding the system, the MCO decreases in relation to the MCE, and a new cycle begins. Similar to the operational planning, the NEWAVE model for expansion planning is used to support decision making. The calculation of the MCO is performed through simulations of the system operation, whereas the calculation of the MCE is estimated by the results of the auctions of new energy and by the belief that the winning bid for the enterprise will be the most expensive bid of the auction, which demonstrates the agents' disposition to invest and constitutes a suitable approximation. 7 In general terms, ANE is the quantity of electricity that can be generated by hydroelectric facilities with the water retrieved by the hydroelectric plants. This energy is estimated by assuming that the level of the reservoirs comprises an average level of 65% of their total capacity and assuming an operational policy. Note that this value may vary according to the operational policy (Terry et al., 1986). 8 Although the analysis is presented in a binary manner (to invest or not to invest), the expansion decision in practice comprises various levels of investment, which characterise alternative time sequences of generation and transmission enterprises.
P.G.C. Ferreira et al. / Energy Economics 49 (2015) 328–335
333
Fig. 3. Operational planning criteria. Source: (Pereira et al., 1998).
Note the dichotomy between the calculations of the marginal costs. The MCO is calculated through the NEWAVE model, in which the main decision variable is based on simulated ANEs, whereas the MCE is estimated by combining market decisions, such as the weighted average cost of capital (WACC), the Brazil risk, interest and the NEWAVE model parameters (Tolmasquim, 2011). Finally, given the time for maturation of the investments in the electrical sector and the significant consequences of rationing, generation planning needs to satisfy the reliability criteria defined by the CNPE, according to which the annual risk of deficit must not exceed 5% in any subsystem. Furthermore, to meet the economic viability criteria, the MCO and the MCE must be equivalent.
Fig. 4. Chain of models of the BES. Source: (ONS, 2012).
In summary, the stochasticity of the flows is intrinsically connected to the expansion planning of the SEB as the decision to expand the system is intimately connected to the MCO, which is calculated based on the NEWAVE model and utilises the synthetic energy series as the main stochastic variable. 3.3. Commercialisation of electrical energy In Brazil, unlike most developed countries (e.g. Dutch electricity market) the energy commercialization market is extremely regulated and the energy spot price is evaluated by computational models. The new model divided the Brazilian energy market into two commercialisation environments: the regulated contracting environment (RCE), which aims to satisfy the needs of captive consumers and is represented by term contracts derived from the energy pool of the market, and the Free Contracting Environment (FCE), involving companies with larger consumption volumes, in which bilateral contracts are freely negotiated according to specific rules and procedures of commercialisation (Castro and Leite, 2010). However, due to the physical attributes of electrical energy that require the instantaneous equilibrium between supply and demand, the ex-ante predicted supply is not always equivalent to the observed demand, which results in a required instantaneous equilibrium at two points: supply and financial accounting. Regarding the first point, ONS, which is responsible for the system operations planning system operations, centralises the dispatch from the hydroelectric plants by grouping the generation and transmission enterprises to petition additional effective management of resources and consequently minimise the cost of energy. Despite all of its contracted energy, a generator may not have to supply energy to the system in this case due to the decisions of the operator. The second point pertains to a function of the CCEE, which accounts for differences between the energy that was produced or consumed and the energy that was contracted. Positive or negative differences are liquidated on the spot market and valued at the liquidation price of the differences (LPD), which is determined weekly by each load level and for each submarket based on the MCO of the subsystem. The LPD is limited by minimum and maximum prices. In this market, the price does not conform to the economic relationship between supply and demand of the agents. Rather, it is determined by a set of computational models operated by the ONS and the CCEE. Expectations regarding future electricity consumption and the future
334
P.G.C. Ferreira et al. / Energy Economics 49 (2015) 328–335
Fig. 5. Investment decision. Source: (ONS, 2012).
ANE regime have a deciding role in the use of energy accumulated in the hydroelectric reservoirs and, as a consequence, also on the price of energy in the spot market. The minimum expected cost for a given timeframe must consider different flow scenarios that lead to different operational decisions. The NEWAVE model calculates the optimal system operation policy by considering present and future costs. The prevalence of hydroelectric plants in the Brazilian generating facilities solves the problem of the randomness of streamflows. This is crucial to the optimisation of the total operational cost in the timeframe considered, as future cost is a function of random future streamflows; immediate cost is a function of current dispatches from the thermal and hydroelectric plants, of which the latter is considered to be zero. Note, however, that this marginal cost will only be zero (corresponding to the cost of water for hydroelectric generation) when full reservoirs exist in the present and the future, that is, from the series of flows with favourable hydrology. Thus, considering previous information on operational planning, expansion planning and planning for the determination of the spot price of energy, strong relationships between stochasticity and the three functions of the sector in the BES becomes evident, that is, the
stochastic scenarios of energy are crucial to the calculation of the optimal manner to operate the sector and supporting decision making on whether expansion is necessary, thus avoiding unnecessary costs and/or losses. These series are a predominant factor in the calculation of the spot price of electrical energy, given that they are determined by the simulated/predicted quantity of water in the reservoirs. 4. Final remarks This study presented a different view of the SEB by demonstrating the intrinsic relationship between hydrological stochasticity and the activities executed by the system, which is an important sector of the infrastructure in Brazil. The synthetic series of energy/flow rates are crucial in the calculation of the optimal manner to operate the sector and for supporting decisions about whether expansion is necessary, thus, avoiding unnecessary costs and/or losses. These electrical energy series are an imposing factor in the determination of the spot cost of electrical energy, given that the simulated/predicted quantities of water in the reservoirs are one of the determinants of the short-term price of energy.
Fig. 6. Expansion planning criteria. Source: (Tolmasquim, 2011).
P.G.C. Ferreira et al. / Energy Economics 49 (2015) 328–335
The importance of this study exceeds its initial objective by creating numerous possibilities and initiating discussions about the manner in which operational planning, expansion planning and determination of the spot price of electrical energy are conducted. A potential outcome of this study is the creation of a research agenda on stochastic modelling for the BES. A discussion of potential improvements for the model adopted by the sector and the appropriateness of the current model is critical, as well as the effects of climate changes that have become prevalent in recent years, such as changes in the rainfall regime, which can generate significant losses in the BES and consequently harm the Brazilian economy. Acknowledgements The authors would like to thank CNPq (research project 443595/ 2014-3), FAPERJ (research project E-26/111.621/2014), which are governmental agencies that have given support to this research. The authors also would like to thank the R&D program of the Brazilian Electricity Regulatory Agency (ANEEL) (PD-0387-0113/2013) for the financial support. References Benders, J.F., 1962. Partitioning procedures for solving mixed variables programming problems. Numer. Math. 4, 238–252. Castro, N.J., Leite, A.L., 2010. Preço spot de eletricidade: teoria e evidências do caso brasileiro. IV Encontro de Economia Catarinense, Criciúma. IV Encontro de Economia Catarinense. CCEE, 2012. Arquivo 2012. (Acesso em 15 de March de 2012, disponível em Câmara de Comercialização de Energia Elétrica: http://www.ccee.org.br). Cyrino Oliveira, F.L., Souza, R.C., Marcato, A.L., 2015. A time series model for building scenarios trees applied to stochastic optimisation. Int. J. Electr. Power Energy Syst. 67, 315–323. EPE, 2012. 2012 Report. Retrieved February 15, 2012, from Empresa de Pesquisa Energética. http://www.epe.gov.br/.
335
Fernandes, E., Fonseca, M., Alonso, P., 2005. Natural gas in Brazil's energy matrix: demand for 1995–2010 and usage factors. Energy Policy 33, 365–386. Ferreira, P.G.C., Souza, R.C., Marcato, A.L., 2015g. The PAR(p) interconfigurations model used by the Brazilian Electric Sector. Int. J. Electr. Power Energy Syst. (Forthcoming). Goldemberg, J., Prado, L.T., 2003. Reforma e crise do setor elétrico no período FHC. Tempo Social 15. 2. USP. São Paulo. Hipel, K.W., McLeod, A.I., 1994. Time Series Modelling of Water Resources and Environmental Systems. Elsevier, Amsterdam, The Netherlands. Kelman, J., 2001. Relatório da Comissão de Análise do Sistema Hidrotérmico de Energia Elétrica (Brasília, Brazil). Lepecki, J., Kelman, J., 1985. Brazilian hydroelectric system. Water Int. 10 (4). Maceira, M.E., Terry, L.A., Costa, F.S., Damázio, J.M., Melo, A.C., 2002. Chain of Optimization models forSetting the Energy Dispatch and Spot Price in the Brazilian System. 14th PSCC. Ministério de Minas e Energia, 2013. Retrieved 2013, from http://www.mme.gov.br/ mme. OECD, 2012. 2011 Report (Retrieved February 15, 2012, from Organisation for Economic Co-operation and Development: http://www.oecd.org/home/). ONS, 2012. http: www.ons.org.br (Retrieved Agosto 2012). Pêgo, B., Campos Neto, C.A., 2008. O PAC e o Setor Elétrico: Desafios para o abastecimento do mercado brasileiro (2007–2010). Texto para Discussão n° 1329. IPEA. Pereira, M.V., Pinto, L.M., 1982. Decomposition Approach to the Economic Dispatch of Hydrothermal Systems. WM. Pereira, M.V., Pinto, L.M., 1985. Stochastic optimization of multireservoir hydroelectric system: a decomposition approach. Water Resour. Res. 21 (6). Pereira, M.V., Campodónico, N., Kelman, R., 1998. Long-term hydro scheduling based on stochastic models. EPSOM'98. Pires, J.C., Giambiagi, F., Sales, A.F., 2002. As perspectivas do setor elétrico após o Racionamento. Texto para Discussão n° 97. BNDES. Santos, G.F., Haddad, E., Hewings, G.J., 2013. Energy policy and regional inequalities in the Brazilian economy. Energy Econ. 36, 241–255. Schaeffer, R., Cohen, C., Almeida, M., Achão, C., Cima, F., 2003. Energia e pobreza: problemas de desenvolvimento energético e grupos sociais marginais em áreas rurais e urbanas do Brasil. CEPAL (Recursos Naturais e Infraestrutura), Chile (2003, p. 79). Terry, L.A., Pereira, M.V., Neto, T.A., Silva, L.F., Sales, P.R., 1986. Coordinating the energy generation of the Brazilian National Hydrothermal Electrical Generating System. Interfaces 16. Tolmasquim, M.T., 2011. Novo Modelo do Setor Elétrico Brasileiro. Editora Synergia, Brasilia, Brazil.
Contents lists available at ScienceDirect
Energy Economics journal homepage: www.elsevier.com/locate/eneco
The stochastic effects on the Brazilian Electrical Sector Pedro Guilherme Costa Ferreira a,⁎, Fernando Luiz Cyrino Oliveira b, Reinaldo Castro Souza c a b c
Brazilian Institute of Economics (FGVIBRE), R. Barão de Itambi, 60 Botafogo, Rio de Janeiro, Brazil Industrial Engineering Department, Pontifical Catholic University of Rio de Janeiro, R. Marquês de São Vicente, 225 Gávea, Rio de Janeiro, Brazil Electrical Engineering Department, Pontifical Catholic University of Rio de Janeiro, R. Marquês de São Vicente, 225 Gávea, Rio de Janeiro, Brazil
a r t i c l e
i n f o
Article history: Received 9 July 2014 Received in revised form 11 March 2015 Accepted 13 March 2015 Available online 22 March 2015 JEL Classification: O13 Q47 Q28 Q25 C59 O54 Keywords: Brazilian Electrical Sector Expansion planning Operational planning Spot price
a b s t r a c t The size and characteristics of the Brazilian Electrical Sector (BES) are unique. The system includes a large-scale hydrothermal power system with many hydroelectric plants and multiple owners. Due to the historical harnessing of natural resources, the National Interconnected System (NIS) was developed outside of the economic scale of the BES. The central components of the NIS enable energy generated in any part of Brazil to be consumed in distant regions, considering certain technical configurations. This interconnection results in a large-scale complex system and is controlled by robust computational models, used to support the planning and operation of the NIS. This study presents a different vision of the SEB, demonstrating the intrinsic relationship between hydrological stochasticity and the activities executed by the system, which is an important sector of the infrastructure in Brazil. The simulation of energy scenarios is crucial to the optimal manner to operate the sector and to supporting decisions about whether expansion is necessary, thus, avoiding unnecessary costs and/ or losses. These scenarios are an imposing factor in the determination of the spot cost of electrical energy, given that the simulated quantities of water in the reservoirs are one of the determinants for the short-term energy price. © 2015 Elsevier B.V. All rights reserved.
1. Introduction The Brazilian Electrical Sector (BES), which is the ninth largest electrical sector in the world in terms of energy generation, produces approximately 470 TW h (OECD, 2012). In 2011, the generation, distribution and transmission sectors experienced billings of approximately US$6 billion. Electricity reaches over 99% of Brazilian homes. Due to the historical harnessing of natural resources for energy generation, the National Interconnected System (NIS) was developed outside of the economic scale of the BES. The central components of the NIS enable energy generated in any part of Brazil to be consumed by consumers in far away, considering certain technical configurations. This interconnection between regions produces an enhanced utilisation of resources, which results in a large-scale complex system that is controlled by robust computational models that are used to support the NIS' planning and operation. The unification of a nationally interconnected system requires billions of dollars from investments (public and private) and the market structure (monopolies and oligopolies) for hydraulic generation, ⁎ Corresponding author. E-mail addresses: [email protected] (P.G.C. Ferreira), [email protected] (F.L.C. Oliveira), [email protected] (R.C. Souza).
http://dx.doi.org/10.1016/j.eneco.2015.03.004 0140-9883/© 2015 Elsevier B.V. All rights reserved.
distribution and transmission. As a result, several problems and challenges emerge that require decisions on different time scales. These problems involve decisions that are directly related to the three functions of the BES: (i) expansion planning, (ii) operational planning and programming and (iii) determination of energy spot price. These functions are executed by different agencies: the Energy Research Company (EPE), the National Electrical System Operator (ONS) and the Board of Electrical Energy Commercialisation (CCEE). The following questions emerge in this environment: should an investment be made to increase the capacity of the system or would it be better to wait for a time of greater expansion in economic activity? How much thermal and hydroelectric energy should be generated to meet current demand? When is the right time to conserve water and use fossil fuels? What is the required spot price of energy to finance all production factors and ensure that the affordability tariff is paid? Responses to these questions, which pertain to the daily activities of the BES, are not trivial and require extensive planning and synchronised management. Stochasticity exists in the following three functions of the BES. In addition to the introduction, this article is organised as follows: Section 2 contains the history of models adopted by the BES. Section 3 describes the sector from the perspective of the new model by identifying relationships between the three supporting functions and hydrological
P.G.C. Ferreira et al. / Energy Economics 49 (2015) 328–335
stochasticity. Section 4 includes final considerations and relevant contributions. 2. BES historical description The origin of the electrical energy supply crisis in Brazil is related to four main causes: (a) exhaustion of the state-owned market model of finance and structure, which is responsible for the expansion of the sector since the 1960s, (b) failures in planning the transition from the stateowned model to the private model, (c) contractual and regulatory problems and (d) the lack of coordination between governmental agencies (Pires et al., 2002). The collapse of the state-owned model can be attributed to two main reasons: the fiscal crisis of the state, which caused the end of the investment capacity of the union at the required levels for expansion of the system (the companies were predominantly state-owned), and an inadequate regulatory regime, which did not promote efficiency and low generation costs as the tariffs were regulated in the generation, distribution and transmission segments. Other factors also contributed to the fiscal crisis of the state, which began in the 1980s and reduced the quantity of union resources for investments. First, the marginal cost of expansion of the sector increased as new hydroelectric basins were situated further from the consumption centres. For this reason, additional resources had to be invested to produce an equivalent generation capacity. Second, the actual values of tariffs, whose price levels did not reflect increased sector costs, deteriorated. In addition to their stability throughout Brazil, the tariffs were used as an inflationary control tool. This process culminated in the decapitalisation and subsequent default of several sector agents. Last, with the consolidation of the democracy and the advent of monetary stability, social demands imposed the need for the government to develop better criteria for the application of union resources (Pires et al., 2002). Based on these reasons, investments in the state-owned companies were not sufficient for satisfying the intensifying demand in Brazil. Thus, a large number of BES expansion projects were halted and/or did not adhere to previously established schedules, which worsened the financial situation of the sector due to increased costs caused by expanded project timelines. From the regulatory point of view, the lack of stimuli to seek productive efficiency prevented companies from establishing incentives to reduce costs. According to Schaeffer et al.(2003), the tariffs were equalised throughout Brazil in the 1970s to stimulate energy development in certain regions by requiring that surplus and deficit companies compensate each other through transfers, gains and losses from their individual efforts. In 1993, at the beginning of the privatisation process, the tariffs were fixed by the electrical distributing utilities to justify the need for companies to establish appropriate tariffs for their markets and to achieve satisfactory levels of profitability. See more about tariff policy of the Brazilian electric power sector in Santos et al. (2013). Regarding the failures in the transition from the state-owned model to the private model and contrary to what was expected, private companies did not invest in the expansion of generating facilities at the beginning of the privatisations in 1995. According to Fernandes et al.(2005), the most important effect of these privatisations was a rapid return to tariff levels (which had been obsolete until then) to increase the attractiveness of private companies. On February 13, 1995, the privatisation and reformation process of the sector began with Law No. 8987, which is named the “Law of Concessions”. According to Pêgo and Campos Neto(2008), in addition to creating conditions for greater participation of private capital, the new law introduced competition in the construction of new projects through regulation of the concessions bidding system for utilities that were previously exclusive to state and federal utilities. In 1996, the BES Restructuring Project was initiated. The primary objectives of the project are summarised as follows: (i) the need to implement the vertical disintegration of electrical energy companies,
329
that is, to divide them into segments of generation, transmission and distribution; (ii) to incentivise competition in the generation and commercialisation segments and (iii) to regulate the distribution and transmission sectors of electrical energy, which were considered natural monopolies under state regulation (CCEE, 2012). The need for the creation of a new regulatory and inspection agency for all sector relationships was identified (Brazilian Electricity Regulatory Agency — ANEEL). The agency, which began to operate in August 1998, was expected to control the operation of the electrical system, from one operator to the national electrical system, in an integrated manner (ONS) and in an environment for buying and selling electrical energy (Wholesale Electricity Market — MAE); the MAE commenced in 2000 with many restrictions. The authors Goldemberg and Prado (2003), highlight that the failure of the electrical sector reform (free market model) was attributed not only to the lack of external resources or the political resistance encountered by the government but also to failures in strategic management, coordination and planning of the electrical system, which was induced by the adoption of a reform that was grounded in the experiences of other countries and inappropriate for Brazil and its predominantly hydroelectric system. In 2001, due to failures of the reform and a lack of sustainability of the model and water problems (shortage of rainfall), the electrical sector experienced a significant supply crisis that culminated in a national electrical energy rationing plan, which affected all categories of consumers. As the crisis was insufficiently addressed by the government, measurement of the immediate results consisted of the controlling of consumption. The government concentrated on thermoelectric power plant construction projects and reinforced the budget for investments in state-owned companies. In 2003, the financial and distribution problems worsened, which caused the government to implement a programme to organise resources for the distribution utilities via the Brazilian Development Bank (BNDES). In addition to measures of short-term emergencies, a new institutional model (new model) was elaborated for the energy sector to correct the failures that caused the crisis and to focus on tariff affordability, universality of access and the resumption of energy planning (Table 1). The new model defined the creation of an entity that is responsible for long-term energy sector planning, the EPE; an institution assigned to permanently evaluate the safety of the electrical energy supply; the Electrical Sector Monitoring Committee (CMSE); and one to give continuity to activities related to the commercialisation of electric energy in the interconnected system: the Chamber of Commercialisation of Electric Energy (CCEE). Regarding the commercialisation of energy, two environments were instituted for the negotiation of energy trade contract: the Regulated Contracting Environment (ACR), in which the agents of generation and distribution of energy participate, and the Free Contracting Environment (ACL), in which the exporting agents of generation, commercialisation, importation and free consumers of energy participate (CCEE, 2012). The new electrical sector model provided a set of measures to be observed by the agents, such as the requirement to contract all demand by the distributors and free consumers, a new methodology for the calculation of backing1 for the sale of generation and the contracting of hydroelectric and thermoelectric power plants for better equilibrium between the backing and the cost of supply, as well as permanent monitoring of the continuity and security of supply to detect conjunctural imbalances between supply and consumption (CCEE, 2012). In terms of tariff affordability, the model established the purchase of electrical energy through auctions by the distributors in the regulated environment. It is considered the lowest cost criterion with the
1 Decree no. 5163/2004 establishes that selling agents must guarantee the backing of power and energy that is sold to 100% of their contracts. This guarantee must be proportioned per generating enterprise or by third parties through energy or power contracts.
330
P.G.C. Ferreira et al. / Energy Economics 49 (2015) 328–335
Table 1 Comparison of the energy models. Characteristics
Old model (until 1995)
Free market model (1995 to 2003)
New model (from 2004)
Financing Structure of businesses
Public resources Vertically integrated
Public vs. private relationship
Predominantly state-owned businesses
Market structure
Monopolies — non-existent competition
Consumers Tariff structure
Captive Tariffs regulated in all segments
Public and private resources Generation, distribution, transmission and commercialisation Coexistence between state-owned and private businesses Competition in generation and commercialisation Captive and free Regulated and free environment of contracting
Market
Regulated
Public and private resources Generation, distribution, transmission and commercialisation Opening and emphasis on the privatisation of businesses Competition in generation and commercialisation Captive and free Prices freely negotiated in generation and commercialisation Free
Expansion planning
Coordinating group for the planning of electrical systems 100% of the market Prorated between buyers
Contracting Surpluses/deficits of the energy balance
National Council of Energy Policy (CNPE)
Coexistence between the free and regulated market Energy Research Business
95% of the market (until Dec. 2004) Liquidated on the MAE
100% of the market and reserve Liquidated on the CCEE
Source: (CCEE, 2012).
objective of reducing the electrical energy acquisition costs to be passed on to tariffs of captive consumers. 3. The BES under the new model Under the new model, the federal government defined a set of institutional agents to guarantee proper functioning of the sector aiming three desired objectives: (i) tariff affordability, (ii) supply security and (iii) universality of access (Tolmasquim, 2011). According to the author, the agents can be classified into three levels according to the legal nature of the entity and their institutional competencies: (i) agents that execute government activities, (ii) agents that execute regulatory activities and (iii) entities of private rights that execute special activities. Government activities are conducted by the National Energy Policy Council (CNPE), the Ministry of Mines and Energy (MME) and the Electrical Sector Monitoring Committee (CMSE). Regulatory activities are conducted by ANEEL. Operational entities conducted technical activities including sector expansion planning (EPE), operational planning and programming (ONS) and commercialisation (CCEE) (Fig. 1).
In addition to the inherent characteristics of the hydroelectric– thermal systems, such as the natural monopoly (transmission and distribution), grid industry, elevated capital intensity, maturation of long-term investments, large interconnections and uncertainties (hydrological, increase in demand and price of fuels), the BES exhibits unique characteristics. In addition to occupying the ninth place in terms of global energy generation and a large consumer market, more than 90% of the energy is generated by hydroelectric plants, and nearly all generation, transmission and distribution systems are nationally interconnected. Besides high capital intensity and a long investment maturity period, the time between the decision to build a hydroelectric plant and its effective entry into operation can exceed ten years. Other characteristics consist of sunk costs (or irrecoverable costs) due to the specialised activities connected to the sector (e.g., electrical energy transmission lines). Thus, the importance of the long-term planning conducted by the EPE is rather important. Another fundamental technical attribute of the sector is its physical equilibrium, which requires the coordination of the system due to its strong interdependence. Generation of electrical energy can occur through various technologies with varying costs and social–
Fig. 1. Institutional agents of the BES. Source: (Ministério de Minas e Energia, 2013).
P.G.C. Ferreira et al. / Energy Economics 49 (2015) 328–335
331
The electrical energy generation system in Brazil (approximately 105,000 MW installed) is mainly hydroelectric. In 2009, the participation of this energy source generated approximately 80,000 MW, including the Brazilian portion of the Itaipu Power Plant (7000 MW) The thermoelectric and thermonuclear sectors represent 18% and 2%, respectively, of the generating capacity (EPE, 2012). In a hydroelectric–thermal system, the future of available hydroelectric generation with no generating costs is uncertain; conversely, the known future of high-cost thermal generation and the need to determine the thermoelectric/hydroelectric allocation at each instant is evident. In contrast to purely thermal systems, for the operation of hydrothermal systems,3 in which the operational planning problem
can be simply resolved by establishing a pattern among the plants that minimises fuel costs,4 the decision is coupled over time, that is, a decision made today will have consequences in the future (Terry et al., 1986). For example, if significant hydroelectric dispatch exists prior to a dry period, a risk of high-cost thermal dispatches will exist in the future; conversely, a thermal dispatch prior to a wet period may cause overflows, which results in wasted energy (Fig. 2). Thus, a balanced operation5 of the system involves a compromise between depleting (using water) and not depleting (using thermoelectric plants) the reservoirs. The decision variable is the volume of water stored at the end of the operational period (final volume). This decision has an immediate cost, which is associated with thermal generation (immediate cost function—ICF), and a future cost, which is associated with the expectation of thermal dispatch and is denoted as the future cost function (FCF). The total cost comprises the sum of these costs. The optimal decision is obtained by cancelling the derivative of the total cost relative to the final stored volume (Pereira et al., 1998). As illustrated in Fig. 3, the ICF increases with final stored volume (higher thermoelectric generation). The FCF exhibits the opposite behaviour: with a higher final stock, there is a lower expectation for future fuel usage. The coordination of operational planning for the hydrothermal system can be designed as a large-scale optimisation problem, with temporal and spatial coupling of the dynamic, stochastic, interconnected and non-linear operations. Its solution requires the problem to be decomposed into a chain of coupled models that consider the long (probability of future storage of energy and expected future value of thermal generation), medium (annual contracts for reserves at the end of the contract) and short term (control of flow and hourly dispatch) planning (Lepecki and Kelman, 1985; Pereira and Pinto, 1982). To resolve this problem, the BES adopts a chain of coupled models that consider different planning timeframes: long-term, mediumterm, short-term and daily planning (Maceira et al., 2002). The coupling between models occurs through the FCF of energy operation. Specifically, operational planning is defined as the efforts to delineate system behaviour over a five-year timeframe by promoting the rational use of resources to guarantee quality and safety in satisfying market demand, operational requirements of the hydrothermal system and minimisation of operational costs. In the operational planning, the objective is establishing a shortterm operation of the hydrothermal system. It should provide operation decisions for the generation system, which are feasible for the transmission system and respect the goals established through operational planning. To cope with the models adopted in the SEB energy planning chain, algorithms from the Stochastic Dual Dynamic Programming (SDDP) are employed (Pereira and Pinto, 1985). This methodology uses the Benders decomposition technique (Benders, 1962) to develop optimal strategies for the operation of the interconnected subsystems, whereas the streamflows are treated using a periodic autorregressive model (Ferreira et al., Forthcoming). SDDP is used to determine the operating policy that minimises the expected value of the operational cost over a five-year timeframe. The strategy is confirmed through a simulation process that employs a series of affluent energies representative of each subsystem, which can be historical or synthetic. The hydroelectric generating facility is represented by an equivalent reservoirs approach, in which a set of hydroelectric plants are grouped into a single reservoir that receives, stores and discharges energy (affluent natural energy). BES is represented by four subsystems: southeast/midwest, south, north and
2 Only 3.4% of the electrical production capacity of Brazil is external to the NIS in small isolated systems, which are primarily located within the Amazon region. 3 Planning must also consider the multiple uses of water in the cascades, such as, the maintenance of navigation, irrigation, flood control and sanitation levels.
4 Additional factors that can exacerbate the complexity of the problem include energy losses, transmission limitations and initial operation costs (Pereira et al., 1998). 5 The system is balanced when sufficient energy can be generated to satisfy demand with a low risk of failure (Kelman, 2001).
environmental impacts. For example, the raw material of hydroelectric plants is the random inflow based on the rainfall regime, which results in heightened complexity for short (one month ahead) and medium (sixty months, or five years, ahead) term planning processes. Such task is performed by the ONS. Despite being an efficient system for planning supply and demand for electrical energy, differences exist between the energy effectively produced by the market and effectively consumed by the market. To resolve this problem, a short-term energy market, whose final objective is solving the disparities between supply and demand, is handled by the CCEE. The NIS was established due to the historical process of harnessing natural resources for energy generation. The NIS, which contains central components, enables energy generation in any part of Brazil to be consumed by consumers in distant regions,2 with certain technical configurations. The interconnection between the regions produces an enhanced utilisation of resources. This interconnection is derived from the notion that Brazil operates a hydroelectric system composed of water reservoirs and hydroelectric plants, which are planned such that they can take advantage of the pluviometric diversity in the existing basins. This measure ensures an important energetic gain for the Brazilian system. The coordinated operation of the NIS by ONS adheres to the general guidelines of the CMSE, which incorporate the objective of meeting the electrical energy consumption requirements for the system to ensure continuity of supply in the most reliable manner and with reduced operational costs. Thus, the rational use of resources must be planned with the objective of satisfying the current requirements of the system to determine the needs for expansion (EPE) with consequent investments. The considerations of the BES specificities in relation to the expansion planning, operational planning and the valuation of spot prices for energy, demonstrate the importance of analysing the sector. These decisions are primarily rendered in an uncertain environment and require systematic decision-making processes, especially with regard to future perspectives. The next three subsections address the expansion planning, the operational planning, the unique determination of the spot price of energy and the identification of characteristics unique to each segment; these areas are related to the stochasticity inherent in the sector. First, the operational planning of the system is addressed, as this activity is directly related to the optimisation model (known as the NEWAVE by the BES) that supports the three activities addressed in this study. Secondly, the expansion planning and determination of spot prices of energy are addressed to demonstrate how stochasticity affects these activities. 3.1. Operational planning of the BES
332
P.G.C. Ferreira et al. / Energy Economics 49 (2015) 328–335
the optimal dispatch from the hydrothermal system, which considers the energetic, hydraulic and electrical aspects of the system. The short and extremely short timeframes (daily programming) allow for the consideration of known or deterministic flow values (Maceira et al., 2002). Last, as the considered time interval decreases, uncertainties related to flows/affluent natural energy (ANE)7 decrease and reach a deterministic extrapolation, and the level of details for representation of the system increases such that characteristics of the individualised systems are considered. 3.2. Planning of electrical system expansion
Fig. 2. Problem of operational decisions. Source: (ONS, 2012).
northeast, which represent the geographic regions. The thermoelectric cost is calculated from the operational costs of thermoelectric plants and the minimum and maximum generations. The NEWAVE model provides a FCF coupled with the short-term model at the end of the planning timeframe (Fig. 4). The objective is to reduce the computational effort required by the optimisation models; the models used for medium-term operational planning employ the grouping of plants into equivalent reservoirs of energy. Each NIS' subsystem (southeast/midwest, south, north and northeast6) is represented by an equivalent reservoir composed of the hydroelectric plants of the region. However, to confirm whether the operational policy obtained from the strategic decision model is feasible, the solution obtained for the equivalent reservoirs of energy must be divided into individual hydroelectric plants, that is, whether the plants comprising the equivalent system will be able to supply the quantity of hydroelectricity generated for the system calculated by the strategic decision making model (shortterm planning) must be analysed. The weekly (and daily) dispatch proposal is subsequently adjusted in the context of daily planning. This study focuses on the stochastic module that supports the NEWAVE planning model. In general terms, this model optimises water use and is dependent on stochasticity, represented by the scenarios generated from time series models, considered in the affluent energies module. These scenarios are generated from a periodic autoregressive model, from Box& Jenkins family, and the simulations are conducted with Monte Carlo Methods and Bootstrap, see Hipel and McLeod (1994) and Cyrino Oliveira et al. (2015). The short-term aspects of quality and satisfying demand involve fulfilling the goals for generation as defined in medium-term planning and verification of operational feasibility, in terms of restrictions of generation and transmission equipment. In this planning timeframe, the objective of the DECOMP model is to minimise the expected value of the total cost of the system, as detailed in the subsequent sections. When the study lead time is reduced to the short term and extremely short term (DECOMP and DESSEM-PAT, respectively), the system representation is refined to the hydroelectric plants by communicating their operational characteristics and hydraulic and energetic restrictions. For operational planning, the timeframe encompasses two hours with discretisation into hours or half hours (Fig. 4). The objective is to obtain
6 Currently, there is a discussion in the SEB to increase the number of subsystems by considering for incorporation into the NIS, for example, the basins of the Madeira, Tapajos and Xingu Rivers. However, only the four traditional subsystems are considered in this study.
According to Tolmasquim (2011), the relationship between the future and immediate use of available financial capital for expansion is similar to the operational planning involving a compromise between the immediate and future use of water for expansion planning. Agents need to decide between investing in the present and assuming the risk of idleness in the system due to the lower than expected growth of demand and postponing investment and risk rationing, as illustrated in Fig. 5. According to the logic of water use in the operation, the decision to expand system capacity decreases the stock of capital, with an immediate cost (known cost of expansion) and a future cost (estimated cost of deficit). Thus, the decision to “save” has a low immediate cost and a high future cost due to an increase in fuel consumption and rationing. Conversely, the decision to “invest” has a high immediate cost and a low future cost (Fig. 6)8 (Tolmasquim, 2011). Similar to the operation process, several alternatives for expansion exist, given a predicted demand in the case of planning. Each alternative is equivalent to a level of reliability R (x-axis), where the cost of expansion is dependent on demand and is unique to the level of reliability R desired by the planner. Similar to the operation, the solution to the planning process is given by optimisation techniques, with the level of reliability (R) as the decision variable and total cost (TC—sum of the costs of expansion and operation) as the objective function. The level of reliability (R*) and the expansion plan are optimal minimum total cost point (Tolmasquim, 2011). In the optimal plan, the derivative of the expansion cost (EC) with respect to the demand (D) is denoted by the marginal cost of expansion (MCE), which is associated with the cost of incremental load with an expansion of capacity. The derivative of the operational cost (OC) with respect to the demand is denoted by the marginal cost of operation (MCO), which represents the cost of meeting the incremental load without expansion of capacity. Thus, once the demand expands over time, the expansion of the SEB occurs when the MCO module is equivalent to the MCE. By expanding the system, the MCO decreases in relation to the MCE, and a new cycle begins. Similar to the operational planning, the NEWAVE model for expansion planning is used to support decision making. The calculation of the MCO is performed through simulations of the system operation, whereas the calculation of the MCE is estimated by the results of the auctions of new energy and by the belief that the winning bid for the enterprise will be the most expensive bid of the auction, which demonstrates the agents' disposition to invest and constitutes a suitable approximation. 7 In general terms, ANE is the quantity of electricity that can be generated by hydroelectric facilities with the water retrieved by the hydroelectric plants. This energy is estimated by assuming that the level of the reservoirs comprises an average level of 65% of their total capacity and assuming an operational policy. Note that this value may vary according to the operational policy (Terry et al., 1986). 8 Although the analysis is presented in a binary manner (to invest or not to invest), the expansion decision in practice comprises various levels of investment, which characterise alternative time sequences of generation and transmission enterprises.
P.G.C. Ferreira et al. / Energy Economics 49 (2015) 328–335
333
Fig. 3. Operational planning criteria. Source: (Pereira et al., 1998).
Note the dichotomy between the calculations of the marginal costs. The MCO is calculated through the NEWAVE model, in which the main decision variable is based on simulated ANEs, whereas the MCE is estimated by combining market decisions, such as the weighted average cost of capital (WACC), the Brazil risk, interest and the NEWAVE model parameters (Tolmasquim, 2011). Finally, given the time for maturation of the investments in the electrical sector and the significant consequences of rationing, generation planning needs to satisfy the reliability criteria defined by the CNPE, according to which the annual risk of deficit must not exceed 5% in any subsystem. Furthermore, to meet the economic viability criteria, the MCO and the MCE must be equivalent.
Fig. 4. Chain of models of the BES. Source: (ONS, 2012).
In summary, the stochasticity of the flows is intrinsically connected to the expansion planning of the SEB as the decision to expand the system is intimately connected to the MCO, which is calculated based on the NEWAVE model and utilises the synthetic energy series as the main stochastic variable. 3.3. Commercialisation of electrical energy In Brazil, unlike most developed countries (e.g. Dutch electricity market) the energy commercialization market is extremely regulated and the energy spot price is evaluated by computational models. The new model divided the Brazilian energy market into two commercialisation environments: the regulated contracting environment (RCE), which aims to satisfy the needs of captive consumers and is represented by term contracts derived from the energy pool of the market, and the Free Contracting Environment (FCE), involving companies with larger consumption volumes, in which bilateral contracts are freely negotiated according to specific rules and procedures of commercialisation (Castro and Leite, 2010). However, due to the physical attributes of electrical energy that require the instantaneous equilibrium between supply and demand, the ex-ante predicted supply is not always equivalent to the observed demand, which results in a required instantaneous equilibrium at two points: supply and financial accounting. Regarding the first point, ONS, which is responsible for the system operations planning system operations, centralises the dispatch from the hydroelectric plants by grouping the generation and transmission enterprises to petition additional effective management of resources and consequently minimise the cost of energy. Despite all of its contracted energy, a generator may not have to supply energy to the system in this case due to the decisions of the operator. The second point pertains to a function of the CCEE, which accounts for differences between the energy that was produced or consumed and the energy that was contracted. Positive or negative differences are liquidated on the spot market and valued at the liquidation price of the differences (LPD), which is determined weekly by each load level and for each submarket based on the MCO of the subsystem. The LPD is limited by minimum and maximum prices. In this market, the price does not conform to the economic relationship between supply and demand of the agents. Rather, it is determined by a set of computational models operated by the ONS and the CCEE. Expectations regarding future electricity consumption and the future
334
P.G.C. Ferreira et al. / Energy Economics 49 (2015) 328–335
Fig. 5. Investment decision. Source: (ONS, 2012).
ANE regime have a deciding role in the use of energy accumulated in the hydroelectric reservoirs and, as a consequence, also on the price of energy in the spot market. The minimum expected cost for a given timeframe must consider different flow scenarios that lead to different operational decisions. The NEWAVE model calculates the optimal system operation policy by considering present and future costs. The prevalence of hydroelectric plants in the Brazilian generating facilities solves the problem of the randomness of streamflows. This is crucial to the optimisation of the total operational cost in the timeframe considered, as future cost is a function of random future streamflows; immediate cost is a function of current dispatches from the thermal and hydroelectric plants, of which the latter is considered to be zero. Note, however, that this marginal cost will only be zero (corresponding to the cost of water for hydroelectric generation) when full reservoirs exist in the present and the future, that is, from the series of flows with favourable hydrology. Thus, considering previous information on operational planning, expansion planning and planning for the determination of the spot price of energy, strong relationships between stochasticity and the three functions of the sector in the BES becomes evident, that is, the
stochastic scenarios of energy are crucial to the calculation of the optimal manner to operate the sector and supporting decision making on whether expansion is necessary, thus avoiding unnecessary costs and/or losses. These series are a predominant factor in the calculation of the spot price of electrical energy, given that they are determined by the simulated/predicted quantity of water in the reservoirs. 4. Final remarks This study presented a different view of the SEB by demonstrating the intrinsic relationship between hydrological stochasticity and the activities executed by the system, which is an important sector of the infrastructure in Brazil. The synthetic series of energy/flow rates are crucial in the calculation of the optimal manner to operate the sector and for supporting decisions about whether expansion is necessary, thus, avoiding unnecessary costs and/or losses. These electrical energy series are an imposing factor in the determination of the spot cost of electrical energy, given that the simulated/predicted quantities of water in the reservoirs are one of the determinants of the short-term price of energy.
Fig. 6. Expansion planning criteria. Source: (Tolmasquim, 2011).
P.G.C. Ferreira et al. / Energy Economics 49 (2015) 328–335
The importance of this study exceeds its initial objective by creating numerous possibilities and initiating discussions about the manner in which operational planning, expansion planning and determination of the spot price of electrical energy are conducted. A potential outcome of this study is the creation of a research agenda on stochastic modelling for the BES. A discussion of potential improvements for the model adopted by the sector and the appropriateness of the current model is critical, as well as the effects of climate changes that have become prevalent in recent years, such as changes in the rainfall regime, which can generate significant losses in the BES and consequently harm the Brazilian economy. Acknowledgements The authors would like to thank CNPq (research project 443595/ 2014-3), FAPERJ (research project E-26/111.621/2014), which are governmental agencies that have given support to this research. The authors also would like to thank the R&D program of the Brazilian Electricity Regulatory Agency (ANEEL) (PD-0387-0113/2013) for the financial support. References Benders, J.F., 1962. Partitioning procedures for solving mixed variables programming problems. Numer. Math. 4, 238–252. Castro, N.J., Leite, A.L., 2010. Preço spot de eletricidade: teoria e evidências do caso brasileiro. IV Encontro de Economia Catarinense, Criciúma. IV Encontro de Economia Catarinense. CCEE, 2012. Arquivo 2012. (Acesso em 15 de March de 2012, disponível em Câmara de Comercialização de Energia Elétrica: http://www.ccee.org.br). Cyrino Oliveira, F.L., Souza, R.C., Marcato, A.L., 2015. A time series model for building scenarios trees applied to stochastic optimisation. Int. J. Electr. Power Energy Syst. 67, 315–323. EPE, 2012. 2012 Report. Retrieved February 15, 2012, from Empresa de Pesquisa Energética. http://www.epe.gov.br/.
335
Fernandes, E., Fonseca, M., Alonso, P., 2005. Natural gas in Brazil's energy matrix: demand for 1995–2010 and usage factors. Energy Policy 33, 365–386. Ferreira, P.G.C., Souza, R.C., Marcato, A.L., 2015g. The PAR(p) interconfigurations model used by the Brazilian Electric Sector. Int. J. Electr. Power Energy Syst. (Forthcoming). Goldemberg, J., Prado, L.T., 2003. Reforma e crise do setor elétrico no período FHC. Tempo Social 15. 2. USP. São Paulo. Hipel, K.W., McLeod, A.I., 1994. Time Series Modelling of Water Resources and Environmental Systems. Elsevier, Amsterdam, The Netherlands. Kelman, J., 2001. Relatório da Comissão de Análise do Sistema Hidrotérmico de Energia Elétrica (Brasília, Brazil). Lepecki, J., Kelman, J., 1985. Brazilian hydroelectric system. Water Int. 10 (4). Maceira, M.E., Terry, L.A., Costa, F.S., Damázio, J.M., Melo, A.C., 2002. Chain of Optimization models forSetting the Energy Dispatch and Spot Price in the Brazilian System. 14th PSCC. Ministério de Minas e Energia, 2013. Retrieved 2013, from http://www.mme.gov.br/ mme. OECD, 2012. 2011 Report (Retrieved February 15, 2012, from Organisation for Economic Co-operation and Development: http://www.oecd.org/home/). ONS, 2012. http: www.ons.org.br (Retrieved Agosto 2012). Pêgo, B., Campos Neto, C.A., 2008. O PAC e o Setor Elétrico: Desafios para o abastecimento do mercado brasileiro (2007–2010). Texto para Discussão n° 1329. IPEA. Pereira, M.V., Pinto, L.M., 1982. Decomposition Approach to the Economic Dispatch of Hydrothermal Systems. WM. Pereira, M.V., Pinto, L.M., 1985. Stochastic optimization of multireservoir hydroelectric system: a decomposition approach. Water Resour. Res. 21 (6). Pereira, M.V., Campodónico, N., Kelman, R., 1998. Long-term hydro scheduling based on stochastic models. EPSOM'98. Pires, J.C., Giambiagi, F., Sales, A.F., 2002. As perspectivas do setor elétrico após o Racionamento. Texto para Discussão n° 97. BNDES. Santos, G.F., Haddad, E., Hewings, G.J., 2013. Energy policy and regional inequalities in the Brazilian economy. Energy Econ. 36, 241–255. Schaeffer, R., Cohen, C., Almeida, M., Achão, C., Cima, F., 2003. Energia e pobreza: problemas de desenvolvimento energético e grupos sociais marginais em áreas rurais e urbanas do Brasil. CEPAL (Recursos Naturais e Infraestrutura), Chile (2003, p. 79). Terry, L.A., Pereira, M.V., Neto, T.A., Silva, L.F., Sales, P.R., 1986. Coordinating the energy generation of the Brazilian National Hydrothermal Electrical Generating System. Interfaces 16. Tolmasquim, M.T., 2011. Novo Modelo do Setor Elétrico Brasileiro. Editora Synergia, Brasilia, Brazil.