April 2000 Power Play Goals New trends and operations research opportunities in the electricity open market By Laureano F. Escudero and Mario Pereira The pace of deregulation and the introduction of competition into the electric energy industry are accelerating globally. From Kazakhstan to Amsterdam, from California to Singapore, electricity markets based on state-of-the-art e-commerce technology are being deployed at unbelievable speed. As R. Massiello noted in his opening remarks at PICA'99, power pools, power exchanges, independent transmission system operators, and Internet-based bulletin boards (such as OASIS for transmission access reservation in the United States), trading systems, retail sales, billing and metering are all creating explosive growth in new products and business. The main objective of the electricity market deregulation all over the world is to decrease the cost of electricity through competition. This is achieved through radical changes in the market and regulatory structure, such as the "unbundling" of functions (separation of generation, transmission and distribution segments) and the creation of bid-based electricity markets. In order to achieve the best value for their production, electric utilities must undergo a deep organizational restructuring and develop adequate software tools to support both new activities (such as energy bidding through e-commerce platforms) and traditional ones that have been transformed by the liberalization process (such as power unit dispatching in a competitive, non-centralized environment). These new support systems include corporate data warehouses, optimization-based generation schedulers and price-bidding tools. Those tools must also have a fast response time, as timeliness of answers is crucial in the open electricity market. The main technological difference between the electricity business and other commodities is the fact that electricity cannot be stored in large quantities. Production and (uncertain) demand must be matched on a minute-by-minute basis. Also, given an adequate transmission network, any generator can sell electricity to any customer. As a consequence, energy production must be carefully scheduled in order to achieve the highest efficiency; decisions made today will directly determine whether the generation company makes or loses money tomorrow. In a deregulated electricity market there are typically three types of transactions that are related to different time horizons: Physical "spot market" transactions, where energy is traded on a short-term basis (usually on a daily or hourly basis) by a number of producers and consumers. The spot bidding includes the price and quantity of energy that is offered to the market. The spot price is the price that balances the spot supply and demand. Physical bilateral contracts, i.e., agreements between a producer and a for selling energy on a medium- or long-term basis. The agreement establishes quantity and price of the exchanged power for the whole period. Financial transactions ("forwards/futures/options" and variations thereof). As mentioned previously, the overall challenge is to design and develop an efficient software system that integrates decision-support functionalities and e-commerce capabilities. The system should be able to solve large-scale problems with sufficient model realism, such as networks of up to 100 thermal units and 50 water reservoirs distributed in 10 river basins, within the tight time constraints stemming from new open market regulations. In order to achieve this, advanced techniques such as stochastic programming, nonlinear programming with mixed 0-1 variables, controlled Monte Carlo simulation, artificial neural networks and cluster analysis schemes will be required. International Power Trading in Hydrothermal Systems One of the most attractive business opportunities in the new environment is the possibility of exploiting the complementary characteristics of hydroelectric and thermoelectric generation technologies in different countries. Typical examples include Western and Southern Europe and the "Mercosur" countries in South America (Argentina, Brazil, Paraguay and Uruguay). For example, the Brazilian power system is almost 100 percent composed of hydro plants with huge reservoirs that are able to store more than a year's worth of power production. Such systems are designed to supply load even if severe droughts occur. As a consequence, there is surplus energy in wet or even normal hydrological conditions. This leads to an interesting spot price pattern: long periods in which electric cost is very low, punctuated by periods in which this cost increases rapidly. This pattern is illustrated in Figure 1, which shows the spot prices in the Brazilian South-Southeast system from January 1993 until July 1997. Figure 1: Brazilian system spot prices. In contrast, Argentina's power production is predominantly thermal (more than 60 percent of installed capacity). As a consequence, spot prices are much more regular, as illustrated in Figure 2. Figure 2: Argentine system spot prices. Looking at both figures, we see that inexpensive Brazilian power could have been exported to Argentina during 20 months (January 1993 to August 1994), thus avoiding huge fuel costs in that country. For the next four months, where Brazilian prices are high, the flow would be reversed, and Brazil would benefit from Argentine imports. The Brazilian export process would resume in January 1995, and so on. One could argue that the same interconnection benefit could have been achieved in the traditional environment, with state-owned power production in both countries. This is true in principle, but experience has shown that the pace of change is slower in this case. As an illustration, the first 1,000 megawatt interconnection between Brazil and Argentina, planned under the traditional framework, took more than 10 years to implement (it will start operating in 2000) because it involved high-level government negotiations and trade agreements. In contrast, more than 4,000 megawatts of new private interconnections, requiring investments of several hundred million U.S. dollars have been proposed in the past 18 months. In order to evaluate the benefit of an interconnection, it is necessary to schedule the system generation and calculate spot prices. The basic objective of hydrothermal scheduling is to ensure an economic and reliable supply of predicted load along the planning period. The economic objective is to minimize fuel costs in thermal plants. The reliability objective is to avoid supply interruptions, including those due to failures in generating units or in the transmission system, and rationing due to the depletion of hydro energy stocks in system reservoirs. Systems with a substantial component of hydroelectric generation can use the free stored hydro energy in the system reservoirs to meet demand, thus avoiding fuel expenses with the thermal units. However, the availability of hydro energy is limited by the reservoir storage capacities. This introduces a relationship between the operative decision in a given stage and the future consequences of this decision. In other words, if the stored hydro energy is used today, and a drought occurs, it may be necessary to use expensive thermal electric generation in the future, or even interrupt the energy supply. If, on the other hand, reservoir levels are kept high through a more intensive use of thermal generation, and high inflows occur in the future, there can be a spillage in the system, which represents a waste of energy and, in consequence, an increase in operation cost. This situation is illustrated in Figure 3. Figure 3: Decision process for hydrothermal systems. Therefore, hydro system operation is coupled in time, that is, an operative decision today affects operating costs and may have environmental consequences in the future. The optimal solution to the hydrothermal planning problem is to establish a balance at an inter-regional level between the immediate benefit from using the water now and the future benefit from storing it. This benefit is measured in terms of expected fuel saving from displaced thermal generation, the cost of unattended energy demand and the cost or the penalization, at least, for not satisfying the water demand for other purposes (irrigation, urban, industrial and ecological) in the future. In order to know whether we should use the stocks of hydro energy, it is necessary to simulate system operation in the future, and assess the impact of this decision in terms of operating costs and societal impact in the geographical area as a whole. The simulation horizon depends on the system storage capacity. If the capacity is small, as in the Argentine system, the impact of a decision is diluted in a few months. If the capacity is substantial, as in the Brazilian system, the simulation horizon will be around five years. This dynamic decision problem is made more complex by the variability of inflows to reservoirs, which fluctuate seasonally, regionally and from year to year. As inflow comes from rainfall, forecasts are generally inaccurate. Because of this inflow uncertainty, simulation studies have to represent a large number of hydrological scenarios (dry, medium, wet years, etc.) in order to evaluate the impact of an operating decision. The objectives of economic operation and supply reliability in hydrothermal systems are conflicting. In order to determine an operating policy, it is therefore necessary to define the tradeoff factor between the two objectives. This can be done by establishing a minimum acceptable level for supply reliability. As discussed above, one of the main decisions to be made is the quantification of the hydro energy to produce now, instead of saving the water for future power generation and other water uses. Because it is impossible to have prior knowledge of future inflows, this tradeoff can only be expressed on a probabilistic basis. Therefore, instead of a generation schedule for each plant, that is, a sequence of operating decisions for each time stage, we need to calculate an operating strategy, i.e., a decision for each possible system state at each stage in a so-called scenario tree. For example, it is intuitive that less preventive thermal generation would be needed if hydro storage levels were high than if they were low. In other words, the optimal amount of thermal generation depends on the reservoir storage state. Decision Support Systems in Electricity Open Markets In order to solve the stochastic scheduling problem discussed above, we need a decision support system (DSS) that globally addresses the open market energy generation allocation, while sharing a common database, as well as modeling and algorithmic approaches. This DSS should include the following functionalities: scenario generation, medium-term energy bilateral sales contract allocation, water value function, unit commitment, energy bidding. Most of the main parameters in the problem are uncertain by nature, namely, water exogenous inflow, regulated energy demand, bilateral sales contracted energy, price, penalization and other specifications, spot market demand and price, fuel procurement cost and availability, and generation units availability. One can treat the uncertainty with stochastic programming via scenario analysis, using a scenario tree to deal with the parameters. The first set of periods, so-called implementable time periods, requires a unique solution without subordinating it to any scenario, but considering all of them. The second set of periods requires a solution for each time period under each scenario group by considering all scenarios, but without subordinating it to any of them. This type of module can be developed by using schemes based on artificial neural networks. Controlled Monte Carlo simulation and clustering approaches can be used as well. The scheduling problem consists of: determining the generation level of each power unit group for each time period along the planning horizon, so that the candidate energy bilateral sales contracts are allocated in their allowed time window, the risk of committing power capacity that could be profitable to keep unused (below a given threshold), the risk of keeping unused power capacity that later could be committed under non-favorable conditions or even remaining unused (also below a given threshold), satisfying a set of technical constraints related to the hydro and thermal generation subsystems, the distribution network, the energy demand and the spinning reserve in an aggregated mode given the planning horizon length and the inherent uncertainties, and maximizing the global profit. The profit is obtained from the difference between the expected income derived from the energy sales contracts that are allocated and the fuel procurement cost of the thermal power units, their operation and maintenance costs and other penalizations. The risk due to committing a power capacity consists of both the risk of losing opportunities for serving future bilateral sales contracts at a better energy price and/or better energy delivery conditions, and the risk of losing opportunities for committing power capacity in a profitable way in the spot market. One of the schemes for risk reducing is the possibility of hedging the contract portfolio's expected profit against the parameters' variability by using financial instruments. The problem is stochastic in nature since the exogenous water inflow to the reservoirs, the water demand for non-hydroelectric uses along the river basins, the fuel procurement costs and availability, and the thermal power units' availability are only known with uncertainty for the medium-term time frame (one to two years). However, above all of these types of uncertainty sits the uncertainty due to the future demand and spot prices, as well as the uncertainty on the future candidate sales contracts (quantity, price and time periods) in the open market. So, a new approach based on scenario sample analysis should be used. The advantage is that it provides solutions depending on the availability of information over time and, then, to state what sort of energy sales contract allocation decisions are to be taken to minimize the regret of wrong decisions along the scenario tree. Stochastic programming approaches for producing robust solutions along the scenario tree should be used on splitting variables and mathematical representations to allow decomposition approaches. The water value function at the end of the short-term horizon (i.e., implementable time horizon) is a crucial piece of information for obtaining the appropriate power unit commitment for complying with the energy bilateral sales contracts and the appropriate energy bidding in the spot market. The function is obtained once the energy bilateral sales contract portfolio has been allocated along the mid- to long-term time horizon and the remaining energy demand has been forecasted for the same time horizon. The uncertainty of the main parameters of the problem, as noted earlier, must be considered and can be dealt with via scenario analysis. The explicit calculation of the water value (i.e., its marginal cost) function requires the computation of the optimal hydrothermal generation coordination strategy for all combinations of possible reservoir states. Note that for each combination state of the water levels at the end of a given time stage, the system should solve the generation coordination problem for the next time stage under uncertainty in the main parameters. As a consequence, the problem quickly becomes computationally intractable even by considering aggregation in the energy generation resources characteristics. The exponential increase of computational effort with the number of reservoirs prevents the explicit solution of the optimal energy generation coordination strategy. Therefore, innovative schemes can be used for obtaining the water value function without explicitly considering all combinations of the reservoir states. Such schemes may require solving successive stochastic programming models in a back-to-front approach until the water value function for the last time period of the implementable time horizon is obtained. These types of schemes can be based on stochastic dual dynamic programming. The short-term (1-2 weeks) dispatching problem consists of determining a very deaggregated power unit commitment with a difficult objective function to minimize. It includes the current cost of using thermal power units (dispatch costs, operations and maintenance costs and fuel consumption costs), the unserved demand penalizations, and the future expected cost as a function of the water stored in the reservoirs at the end of the short-term time horizon. All the data are known. The problem also entails finding the energy generation level of every power unit (thermal or hydro) in every time period, so that the energy demand is supplied at minimum cost and all the constraints related to the hydro and thermal subsystems as well as the global constraints are satisfied. The constraints related to thermal power units include the maximum power generation capacity, the units' start-up and shut-down duration, the units' technical minimum power and its duration, the bounds on the ramp-down and ramp-up rates for the thermal units, the thermal capacity states, the fuel consumption function of the thermal power, the fuel procurement and consumption upper bounds (e.g., due to environmental constraints) and quotas per fuel type, and the fuel stock's upper and lower bounds. The hydro power units' constraints are the water conservation constraints in all reservoirs of the river basins under control of the electric utility, the lower and upper bounds of the reservoirs' water stored volume and water flow through the water subsystem channels, the hydropower function of the individual reservoir's height and turbine discharges, and the upper and lower bounds of the turbine discharges and spillages. The system's global constraints are related to the appropriate spinning reserve margins and the energy demand satisfaction per time period and geographical area (where the distribution network topology and capacity should be considered). The demand to satisfy is the energy bilateral sales contract (that has been previously decided to be served along the short-term horizon) as well as other types of load to accept. The traditional difficulty of the problem due to the deaggregated consideration of the generation resources (where, given the economical impact of the solution, only the optimal solution is accepted) is increased by adding the water value function to the objective function to optimize, and the competitive nature of the process these days. A scheme based on nonlinear programming with mixed 0-1 variables could be used. The trading problem in the short-term timeframe involves bidding for energy, offering in price and quantity out of the power-generation capacity by considering the demand constraints to satisfy the energy bilateral sales contracts portfolio and other committed energy load for each time period. The objective function to maximize is the net profit to be obtained by the energy generation of every power unit in every time period, where the profit is given by the targeted spot price times the pool offered energy minus the cost function that is also used in the unit commitment task. A parametric nonlinear optimization approach with mixed 0-1 variables and game theory scheme could be used. Good decision-making relies upon readily available data (or facts) and on good quality analytic information. Information system specialists have now accepted that operational or transactional databases that support the organization are not necessarily suitable for decision-making. There is a strong move towards creating analytic databases or data warehouses for the global organization including fuel suppliers, thermal units and reservoirs, main customers, spot market (i.e., pool) operator, independent system operator, etc. We may call it the decision database as it supports decision-making models. A transactional database typically has information about power units' production rates, costs, reservoirs, topology and water levels and sales contracts (energy price, quantity, load per type of hour and days). An external database could include transmission system topology, potential customers' characteristics and suppliers, spot market prices and demand. Analysis of data items lead to forecast demands, power unit performance efficiency and breakdown rate. External analytic information could be competitor intelligence, industry sector performance indices, economic forecasts and consumer behavior trends among others. Getting the multi-site information in time at the appropriate decision-making installations can only be accomplished by using quick and reliable simulation or optimizing algorithms implemented in electronic commerce environments. So, starting from the system requirements, a new integrated simulation and optimization algorithmic approach could be developed for the optimization of the open market generation allocation processes in the energy sector. References, Bibliography and Further Reading Alvarez, M., C.M. Cuevas, L.F. Escudero, J.L. de la Fuente, C. Garcia and F.J. Prieto, "Network planning under uncertainty with an application to hydropower generation," TOP Trabajos de Investigacion Operativa, Vol. 2 (1994), 25-58. Bianto, S. and M.V.F. Pereira, "Decentralized Planning of Hydroelectric Power Systems," Paper 44 SM 522-3 PWRS IEEE PES Summer Meeting, San Francisco, Calif., 1994. Escudero, L.F., "On using multi-stage linking constraints for stochastic optimization as a decision-making aid," Revista de la Academia de Ciencias, 92 (1998), 371-376. Escudero, L.F., "A robust modeling approach for water resources system planning under uncertainty," Annals of Operations Research (1999, accepted for publication). Escudero, L.F., J.L. de la Fuente, G. Garcia and F.J. Prieto, "Hydropower generation management under uncertainty via scenario analysis and parallel computation," IEEE Transactions on Power Systems, Vol. 11 (1996), 683-689. Escudero, L.F., J.L. de la Fuente, C. Garcia and F.J. Prieto, "A parallel computing approach for solving multistage stochastic networks," Annals of Operations Research, 90 (1999), 131-160. Escudero, L.F., I. Paradinas and F.J. Prieto, "Generation expansion planning under uncertainty in demand, economic environment, generation availability and book life," in: Invited Speakers Sessions Proceedings of the Stockholm Power Tech (IEEE, Stockholm, Sweden, 1995). 226-233. Escudero, L.F., I. Paradinas, J. Salmeron and M. Sanchez, "SEGEM: A simulation approach for Electric Generation Management," IEEE Transactions on Power Systems, Vol. 13, 3 (1998), 738-748. Gorestin, B.G., N.M. Campodonico, J.P. Costa and M.V.F. Pereira, "Power System Expansion planning under uncertainty," IEEE Transactions on Power Systems, Vol. 8 (1993), 129-136. Pereira, M.V.F., "Optimal stochastic operations scheduling of large hydroelectric systems," International Journal of Electric Power and Energy Systems, Vol. 11 (1989), 161-169. Pereira, M.V.F., N.M. Campodonico, B.G. Gorestin and J.P. Costa, "Application of Stochastic Optimization to power system planning and operations," Invited Speakers' Sessions Proceedings of the Stockholm Power Tech (IEEE, Stockholm, Sweden, 1995). Pereira, M.V.F. and N.M. Campodonico, "Stochastic hydrothermal scheduling in a competitive environment," IEEE Transactions on Power Systems (1999, to appear). Pereira, M.V.F. and L.M.V.G. Pinto, "Multistage stochastic optimization applied to energy planning," Mathematical Programming, 52 (1991). Pereira, M.V.F. and L.M.V.G. Pinto, "Stochastic optimization of a multireservoir hydroelectric system, a decomposition approach," Water Resources Research, Vol. 21 (1985), 779-792. Laureano F. Escudero is the director of Sistemas de Apoyo a la Decisión (SAD) in Madrid, Spain. Mario Pereira is the president of Power Systems Research Inc. (PSRI) in Rio de Janeiro, Brazil. Table of Contents OR/MS Today Home Page OR/MS Today copyright © 2000 by the Institute for Operations Research and the Management Sciences. All rights reserved. Lionheart Publishing, Inc. 506 Roswell Street, Suite 220, Marietta, GA 30060, USA Phone: 770-431-0867 | Fax: 770-432-6969 E-mail: lpi@lionhrtpub.com URL: http://www.lionhrtpub.com Web Site © Copyright 1999, 2000 by Lionheart Publishing, Inc. All rights reserved.