The EIA Short-Term Regional Electricity Model and Its Application in Power System Analysis

Transcription

1 The EIA Short-Term Regional Electricity Model and Its Application in Power System Analysis Phillip Tseng, Senior Analyst United States Energy Information Administration 1 Independence Avenue, Washington, DC 2585 Telephone: (22) Phillip.Tseng@EIA.DOE.Gov Abstract The United States Energy Information Administration (EIA) developed a 13-region shortterm electricity model for its monthly projections of electric power generation from fossil fuel power plants. The model has explicit representation of hourly/monthly demand and supply for each region. Data inputs and model parameters are derived from the 25 data on generation, plant capacity, and sales published by EIA as well as data from power control areas. The contributions of this paper are: (1) to provide forecasts and (2) to provide simulation results addressing potential impacts on gas market and trade, which have implications on efficiency, reliability, and investment. The modeling framework and model outputs provide insights into the economics of dispatching power from gas plants, trade flows, and the utilization pattern of existing transmission capacity. First, natural gas power plants are important to six of the thirteen regions, especially in peak demand seasons. Simulation results demonstrate that the model can capture the effects of load changes on generation from gas power plants. As a result, gas demand in the power generation sector can be assessed with the rest of the natural gas market and provide a more comprehensive understanding of the overall demand and supply situations. Second, hourly trade flows shed light on hourly utilization rate of transmission lines, which are essential to investment decisions in improving the reliability of the power system. Simulation results also demonstrate the benefits of inter-regional trade in improving overall generation efficiency and supply reliability. Finally, model results on hourly capacity utilization patterns provide indicators to the importance of fuel availability during peak demand hours in regions relying on imported coal, natural gas, and oil. In addition, the implied hourly utilization rate of transmission lines can help decision makers evaluate optimal investment choices in an evolving and complex energy market. 1. Introduction The United States Energy Information Administration (EIA) publishes a Monthly Short Term Energy Outlook. The publication, as its name suggests, covers EIA s view of short term energy markets and the likely development in demand and supply of oil, gas, and electric power. One of the challenges in this publication is making consistent projections of natural gas use in the electric power sector. It is important because an increase in power sector demand for natural gas can be a burden to the already tight U.S. gas market. Therefore, EIA needs to improve its The views and opinions of the author expressed herein do not necessarily represent those of the EIA or the U.S. Government. The author would like to thank Dr. Howard Gruenspecht for his encouragement and support; Dr. Tyler Hodge made significant contributions to the analysis of generation data; Dr. Howard Bradsher-Fredrick and Dr. Robert Trost provided valuable comments and suggestions to an earlier draft of the paper.

2 capability in assessing the impacts of higher demand for power on gas usage, overall demand for natural gas, the amount of gas injected into storage, and gas prices. In addition, EIA also uses the model to conduct in-depth analysis of the electricity market. Policy makers may want to know the reliability of supply in a region facing transmission constraints. Contingency planners may want to know about potential bottlenecks in trade flow and supply of power if there are unplanned outages. Investors may want to know the best investment strategy in meeting demand by assessing demand growth, infra-structure constraints, and investment opportunities in DG technologies, demand side management, new power plants, or transmission lines. Answers to the questions of where to invest, what to invest, and how much to invest, however, lie in empirical data on regional demand, generation, and power flow. In meeting these challenges, EIA has developed a 13-region short-term regional electricity dispatching model for the projection of power generation from coal, natural gas, and oil fired power plants. The model divides the 13 regions into three trading blocs: Eastern, Western, and Texas. The Eastern bloc has nine regions, the Western bloc has 3 regions, and Texas is a region by itself. The model has explicit demand and supply representations developed from historical data and it is structured to serve both as a short term forecasting tool and an analytical framework for scenario analysis. Section 2 of this paper discusses the data sources and modeling methodology. Section 3 reports the performance of the model. Section 4 analyzes hourly trade flows and the economics of dispatching decisions. Section 5 summarizes the findings and discusses optimal investment decisions in improving power system reliability. 2. Data Sources and Modeling Methodology The EIA short term regional electricity model is designed to perform two functions: forecast short term electric power generation from coal, gas, residual fuel, and diesel fuel power plants and simulate the effects of changes in demand and generating capacity on regional generation and trade. These two functions impose very rigid conditions on the structure of the model. As a short term forecasting tool, the model must use historical data to capture demand and generation patterns. As a simulation tool, the model must have demand and supply structure for modelers to adjust input assumptions. Differences in regional characteristics and availability of data also dictate the regional design of the model. The model is built on three major data sets collected and published by the U.S. Energy Information Administration. These include annual generation capacity, monthly net electricity generation, and monthly electricity sales. Following is a brief description of the data sources used to construct the model. EIA-826: Monthly electric utility sales and revenues reported by states. Sales data is based on billing cycle, which may not coincide with the calendar month. EIA-96: Monthly data on generation and fuel use at the generator or plant level by states. EIA-86: Annual generating capacity by technology and by states

3 EIA relies on a private consulting firm to collect power control area data and convert the monthly sales data to 24-hour average daily load curves. Figure 1 shows the relationship of data inputs and the demand and supply representation. Regional supply curves for coal, natural gas, residual fuel, and diesel fuel are derived from generation capacity and monthly generation data. These regional supply curves are aggregated to represent total supply in a trading bloc. Sales data are adjusted from billing month to calendar month. The hourly power control area data are then used to convert the monthly sales data to a 24-hour load curve to represent average daily load for each month. Regional demand curves are aggregated to represent total demand in a trading bloc. The model solves for market clearing generation for the trading bloc and computes regional generations from coal, gas, residual fuel and diesel fuel. 2.1 Regions The EIA regional electricity dispatching model divides the lower-48 states into three trading blocs: Eastern, Western, and Texas. Figure 2 shows regions. This division uses nine census regions plus four states: California, Florida, New York, and Texas. The selection of trading blocs and regions takes into consideration data availability, effects of trade within each trading bloc, and ease of use and maintenance for regular monthly dissemination. The general model structure can accommodate more regions for contingency analysis provided they are aggregated up from the state level. The Eastern bloc includes nine regions: New England, Mid-Atlantic minus New York, East North Central, West North Central, South Atlantic minus Florida, East South Central, West South Central minus Texas, Florida, New York. The Western bloc includes three regions: Mountain, Pacific minus California, California. Texas, separated out from West South Central, is a region by itself in the regional electricity dispatching model. 2.2 Model Structure and solution algorithm This section describes the estimation of supply curves, load curves, and the solution algorithm for the electricity generation model Regional supply curves The most challenging task in modeling supply of electricity by fuel type is the identification of regional supply curves. Ideally, a regional supply curve should show the relationship between the price and the quantity supplied. A plant with the lowest variable cost should be dispatched first; as demand increases, the next lowest cost plant will supply power. In reality, dispatching decisions may depend on the least system cost. Spatial considerations and institutional factors play an important role in determining dispatching orders and plant utilization rates. These factors include costs and accessibility of transmission lines, supply contracts, distance to load centers, transmission and distribution losses, and availability of fuels. Ideally, a

4 regional electricity model should include all the cost components. However, the level of efforts to build such a detailed model would require very detailed data analysis of load center, plant locations, and transmission capacity. These data are not readily available and will be very costly to collect and process. An easier alternative relies on a revealed preference method to develop curves as a proxy to supply curves. The revealed preference method uses capacity utilization rates to determine dispatching order of fossil power plants. The EIA-86 power plant capacity and EIA-96 power plant generation data are used to calculate capacity utilization rates for every power plant in the data base. A power plant with a high utilization rate implies it is dispatched more often and is more likely to have lower system costs. Theoretically, the cumulative distribution function of capacity, sorted by utilization rate from high to low, should be a good proxy to the load and supply curve. EIA uses a modified cumulative distribution function for each type of fossil plant as a proxy for the supply curve of each plant type. This approach improves the fit between the estimated curve and the modified cumulative distribution function. For example, coal power plants in a region are sorted by utilization rates and assigned to 1 bin numbers. Plants with utilization rates greater than or equal to 9% are assigned to bin number 1. Plants with utilization rate greater than or equal to 8% and less than 9% are assigned to bin number 2. Every fossil plant in the data base is assigned one bin number. Table 1 shows the bin number, capacity in each bin, and cumulative capacity of coal and natural gas plants in Texas in July 25. For each fuel type, the supply curve in a region is estimated with cumulative capacity as the dependent variable and bin number as explanatory variable. Each supply represents the relationship between capacity and historical dispatching order. The supply curves for coal, gas, diesel fuel, and residual fuel take two functional forms. The coal supply curve takes the functional form: Log (cumulative coal) = C () + C (1) * (1/bin) (1) Where C () and C (1) are estimated coefficients The gas, diesel, and residual fuel supply curves take the functional form: Log (cumulative XXX) = C () + C (1) * bin + C (2) * bin 2 (2) Where XXX=gas, diesel fuel, or residual fuel C (), C (1), and C (2) are estimated coefficients Availability of coal power plants takes into consideration routine maintenance and planned outages. The data EIA collected show that most power plant operators reported use of small amounts of natural gas, diesel, or residual fuel; these observed data indicate that these coal power plants are not 1% available. An availability factor for each region is used to reflect this fact to improve model performance. The availability factor is.93,.94, and.96 for the Eastern trading bloc, the Western trading bloc, and Texas, respectively.

5 For the model, there are thirteen sets of supply curves for fossil power plants for each month; one set for each region. These monthly supply curves depict what plant operators will dispatch in response to variable demand levels in a typical day. These historical monthly supply curves derived from monthly data for a single year capture only seasonal dispatching patterns. Dispatching gas power plants can be influenced by the prices of residual and diesel fuels. There are two aspects in fuel switching. First, if the relative price of gas is significantly below prices of residual fuel and diesel, the utilization rate of natural gas plants could rise if the savings in system cost is positive. Second, for dual fueled gas/resid or gas/diesel plants, plant operators can switch if fuel cost savings from switching is positive. Empirical analysis shows that generation from gas plants is sensitive to the relative price of gas to residual fuel. However, generation from residual fuel is sensitive neither to gas price nor residual fuel price Regional demand for electric power Demand for electricity changes by hour, by day, and by month. This fluctuation in demand determines dispatching needs and patterns. A region without cost effective generating capacity at any point in time might import from other regions so long as transmission capacity is not a limiting factor. Demand for power in a typical day depends on economic activities and daily temperature. Its components include demand from residential, commercial, industrial, and transportation sectors. As a result, weekday activities could be different from weekend activities and the level and shape of a 24-hour load curve can also be different. Ideally, two sets of 24-hour load curves should be used to capture the effects of peak load demand on dispatching decisions; one for weekdays and the other for weekends. However, the level of effort in getting the more detailed data may outweigh the benefits of marginal improvements in out-of-sample projections. As a result, for each region only one 24-hour load curve will be used to represent an average day for each month. In this model, we use twenty four time series data to represent twenty four demand slices for a typical day of a typical month. The sum of these 24 demand slices times the number of days in the month will always be equal to a pre-determined monthly demand, which can be either historical data or projected demand. Adjustment to the load curves to remove the supply of power from hydro, nuclear, renewable, and combined heat and power (CHP) is necessary to simplify the dispatching algorithm. An ideal approach is to remove the generation of these technologies from the load curve according to their dispatching pattern if we know when these technologies are dispatched and how much. For example, we know nuclear power plants probably run continuously 24 hours a day so we subtract hourly nuclear generation from the 24-hourly load curve. For hydro power, we should allocate more generation to peak hours. For solar power, we should convert monthly to daily average and allocate generation between late morning to late afternoon. The level of effort required to make the adjustments to fine tune the shape of adjusted load curve outweighs the improvements in projecting dispatching fossil plants because solar generation is very small and wind generation cannot be determined by power plant operators. To simplify the adjustment, we convert monthly non-fossil generation to a 24-hour daily average generation curve and subtract the generation from the 24-hour load curve. The dispatching model uses the adjusted load curves to solve for generation of coal, gas, diesel, and residual fuel.

6 2.2.3 Solving for dispatching of fossil plants This sub-section discusses the solution algorithm, adjustments to the load curve, model outputs, and the computing platform. The solution algorithm for the electricity dispatching model is straightforward; for each 24-hour time period, the regional supply is equated to the regional demand. The equilibrium bin number, now a continuous variable, then determines generation by fuel type. Figure 3 illustrates the process. First, the model aggregates oil, gas, and coal supply curves to get a total fossil fuel supply curve for a region. Given a regional demand, the model equates supply to demand and solves for equilibrium bin numbers and net generation. The equilibrium bin number then determines generation from oil, gas, and coal plants. Mathematically, the following three equations show monthly supply, demand, and market clearing generations. Supply = S IJ (3) Where I = coal, gas, diesel fuel, or residual fuel J=regions Demand = D J (4) Supply = Demand (5) For Texas, monthly supply is simply the sum of four monthly supply curves: coal, gas, diesel fuel, and residual fuel. The January supply curves will be used, in conjunction with a 24- hour daily average January demand, to solve for 24 generation levels of coal, gas, diesel, and residual fuel. Daily coal generation is the sum of 24 hourly generations, and monthly generation equals daily generation times the number of days in the month. For the Western trading bloc, there are three regions: Mountain, Pacific minus California, and California. The monthly regional supply curve is the sum of three regions supply of coal, gas, diesel, and residual fuel. The regional 24 hourly demands is the sum of demand in these three regions. For the Eastern trading bloc, there are nine regions and the calculation is the same as the Western trading bloc except the number of regions is nine instead of three. The dispatching model outputs include regional power generation by coal, natural gas, diesel fuel, and residual fuel for the forecasting time horizon of 24 to 36 months. Fuel use can be computed based on net generation and the average heat rate of power plants. Table 2 illustrates the demand and supply relationship and their implications on hourly trade flows. For each hour, the sum of generation from all regions in a trading bloc must equal the sum of demand. For each region, the difference between hourly demand and hourly generation is imports or exports. In addition, the sum of generation of 24 hours for a region is the average daily generation, which should not deviate from the historical average daily generation if the model dispatches power according to the historical pattern.

7 The model uses EViews software to estimate the coefficients of the supply curves and solve the model for generation by fuel type. It takes about 6 seconds to solve for the monthly forecast over a 48 month time horizon. 3. Performance of the Model The current regional electricity model used 25 annual capacity and monthly generation data to create monthly electric power supply curves of coal, natural gas, diesel, and residual fuel for all 13 regions. Monthly sales of electricity for 24, 25, and 26 at the state level are aggregated to the regional level. For each region, monthly sales data were then converted to a 24- hour load curve. The model uses hourly load to solve for generation by coal, natural gas, diesel fuel, and residual fuel. Figures 4, 5, and 6 show the model results; they compare projected average daily generation from coal and natural gas to their historical counterparts. Overall, the model is capable of capturing the seasonality of electricity markets. In general, the model has the tendency to use more coal and fewer gas and oil-fired units. While projected generations from coal and gas track historical patterns well, projected generation from diesel and residual fuel are not very satisfactory. Several factors may contribute to this deficiency. First, dispatching decisions for oil based generators may be different from that of coal and natural gas; it is very likely that these so-called peak units may also be used as transition units to smooth out the changes in load because running a large generator as spinning reserve may be more costly if the load shape and dispatching options can be well defined. Second, the load curves in each region were aggregated from different regions or states. As a result, these curves may have flattened load shapes and prevented the model from dispatching correctly. Third, generators may have reported inconsistent data to EIA because of the relatively small shares of oil-based units in the whole generation mix. Fourth, EIA may have adopted a methodology in the estimation of generation for plants which are not in the monthly sample. An experienced model user could make adjustments to the bin numbers assigned to diesel and residual fuel units and then try to match the historical generation pattern. However, the fact that oil-based units are a very small fraction of total generation makes it less attractive to spend too much effort to make the improvement. There are many different measures to evaluate the performance of a model. The selection of measurements, therefore, depends on the intended use of the model. EIA will use the model to provide monthly forecasts and gain insights into the use of natural gas by the power generation sector. EIA will also use the model to simulate the effects of unplanned plant outages on generation patterns and trade flows; scenario analysis could include the effects of above normal cooling degree days on power generation and use of natural gas and petroleum based products. The goodness of fit of the model can be measured by the mean absolute percentage error (MAPE). Equation 6 shows the computation of MAPE. 1 * T + h t= T + 1 ( yˆ t yt) / yt / h (6)

8 Where ŷt is projected value y t is historical value h is number of observations Table 3 shows the mean absolute percentage error for the Eastern trading bloc, the Western trading bloc, and Texas. The MAPE calculated from the 12 months of 25 for gas are 2.1%, 2.88%, and 4.14% for Texas, Western trading bloc, and Eastern trading bloc, respectively. MAPE increased when we include the twelve months of 24 out of the sample forecast. The 24 month MAPE calculated from 24 and 25 projections are 3.4%, 3.51% and 7.44%. The Western trading bloc has the smallest MAPE for coal. The performance of the Eastern trading bloc is not as good as the other two regions because of the concentration of diesel and residual fuels in the East and the inability of the model to select the dispatching patterns satisfactorily. Another measure of the performance of the model is the percentage difference within the forecasting time horizon. It is relevant especially for the natural gas market. If natural gas use is consistently above normal, injections into underground storage would be below normal. As a result, natural gas prices in the spot and future markets may increase in anticipation of potential shortages. In addition to the MAPE, the other measure can be twelve month percentage error. Equation 7 reports the calculation. T + h ( yˆ t t= T + 1 yt) / yt (7) Table 4 shows the month percentage difference. Texas was the smallest. Both Texas and the West show differences smaller than 1%. However, the East shows 2.35% for coal and 4.6% for gas. For the East, the differences for coal and gas are positive and it reflects the effects of underestimating generation from diesel and residual fuel. 4. Analysis of Dispatching patterns and Trade Flows The model provides insights into two important areas. First, in a tight natural gas market the ability to predict dispatching of power from gas plants can help analysts understand natural gas market conditions and the direction of gas price changes. Second, the modeling methodology and the historical data allow analysts to assess regional trade and generation patterns empirically. Trade flows reflect the extent transmission lines are utilized, which are essential to investment decisions in improving the delivery reliability of the power system. Information on fuel availability, prices, and the implied utilization rate of transmission lines can help decision makers determine optimal investment choices in an evolving and complex energy market. The rest of this section will analyze these two specific areas. Natural gas and petroleum liquids power plants play a crucial role in some regions. Table 5 shows the share of natural gas in each of the 13 regions for July 25. Natural gas power plants are important to six of the thirteen regions, especially in the peak demand seasons. In the Eastern trading bloc, New England, West South Central, Florida, and New York, share of natural gas ranged between 32 and 47 percent. In California and Texas, share of natural gas exceeded 5

9 percent. Simulation results demonstrate that the model can capture the effects of load changes on generation from gas power plants. As a result, gas demand in the power generation sector can be assessed with the rest of the natural gas market and provide a more comprehensive understanding of the overall demand and supply. In a tight gas market facing an above normal number of cooling degree days, the model can estimate the potential changes in gas injection into storage, which is closely watched by the commodity market. This can have profound implications on prices in the gas futures market. Economics of Short term power supply and trade flow. In the short run, power plants that are responsive to changes in demand will compete based on marginal cost. An importing region will determine when to import and how much to import based on the marginal cost of producing power, cost of imported power, and demand level. An exporting region can export power and compete only if the delivered electricity, which includes the cost of producing power, the transportation cost, transmission loss, is less than the marginal cost of power produced in the importing region. The marginal cost of producing power can change as hourly demand increases in each region. As a result, trade patterns could also change by the hour. To capture the hourly generation pattern and match historical monthly total generation for Florida, New England, New York, West South Central, supply curves in these regions are shifted. For New England and Florida the gas supply curves are shifted to the left to reduce the overall generation level to the historical monthly generation level. For New York, the natural gas supply curves are shifted to the right to increase the penetration of natural gas in the region. The EIA model has two trading blocs that report trade flows. Regional imports and exports can shed light on the characteristics of the market in each trading bloc, the economics of dispatching, and possible constraints facing an import region. For each trading bloc, model results include an average 24-hour daily as well as monthly regional load, generation, and imports/exports. The monthly trade patterns provide an indication of utilization rate of transmission capacity. However, the hourly trade flow shows the utilization pattern of transmission capacity. Note that this paper will focus on the pattern and not the absolute level of transmission capacity utilization rate as the average daily load curves in a region tend to understate the fluctuation of actual load curve. EIA publishes monthly sales and generation data. For the Western trading bloc, the Mountain region exports and California imports power. Oregon and Washington rely heavily on hydro power and exports a small amount of power in most months except when demand is high and hydro power supply is low. Monthly import patterns give only average utilization of transmission capacity. It does not provide insights on potential bottlenecks. The hourly generation and trade provide more useful information on potential supply problems. Figure 7 shows the hourly generation and trade in July 25 for the region. The Mountain region exported more power during peak demand hours and California imported more accordingly. For the California, utilization rate of transmission capacity increased as demand increased. This reliance of imported power in California highlights the importance of imports and transmission reliability to the region. The Eastern trading bloc shows a very different trade pattern. Several regions, New England, West South Central, Florida, and New York, imported relatively more power when monthly demand is low and imported less when demand is high. This import pattern reflects the

10 fact that these regions have higher market share gas plants. As a result, at a lower demand level, an export region with more low cost power plants can compete effectively with power plants in the import regions. The hourly generation and trade patterns for the Eastern Trading Bloc are also different from California. Figure 8 shows that New England, Florida and New York all show a higher level of imports during off-peak demand hours. These results may reflect the fact that imported power is cheaper than indigenous power in the early hours of each day. During the peak hour, however, importing regions have to produce more because exporting regions may not have as much to export due to higher indigenous demand and higher marginal cost. In addition, this import pattern may reflect relatively high market share of oil and gas power plants in these three regions and the high costs of gas and oil. It is cheaper to import power during the off-peak hours when the cost of imported electricity is low. This import pattern serves two functions. It lowers the costs of supplying electricity and it can reduce required storage of oils for some power plants in meeting peak load demand. In the Eastern trading bloc, regions with higher oil and gas share, such as, New England, West South Central, Florida, and New York, all imported power during the off peak hours. Each of these regions had enough generation capacity to meet its indigenous demand. However, economic considerations dictated the import and generation decisions. Clearly, regional trade reduced cost of electricity. More importantly, it also conserved important and more expensive resources for using when the demand is high. 5. Summary and Conclusions The modeling approach reported in this paper works well as a tool for projections as well as for in-depth simulation analysis. The mean absolute percentage error is less than five percent for projected generations of coal and natural gas power plants. As a simulation tool, it sheds light on the economics of regional generation and inter-regional trade. The hourly trade flow highlights the importance of recognizing the differences in each market in developing measures to improve supply and delivery reliability. Natural gas power plants are important to six of the thirteen regions, especially in the peak demand seasons. Simulation results demonstrate that the model can capture the effects of load changes on generation from gas power plants. As a result, gas demand in the power generation sector can be assessed with the rest of the natural gas market and provide a more comprehensive understanding of the overall demand and supply situations. Inter-regional trade improves the overall generation efficiency of a trading bloc. In the Eastern trading bloc, Florida, New England, and New York, imported more in months when demand is low and imported less when demand is high. These regions may have enough generating capacity to meet its indigenous demand. However, during off peak hours these regions may find it more cost effective to import power. During peak demand hours, an exporting region may also face higher indigenous demand and may find it difficult to increase total generation to meet indigenous peak load demand and export power. In addition, conversion efficiency of intermediate load and peak load power plants in an exporting region may not be very different from that of an importing region. Transmission loss and tariff may prevent an

11 exporting region from sending more expensive power to other regions where less efficient power plants may be able to compete. Analysis of hourly trade flow from model results can improve stakeholders understanding of investment needs. In California, the utilization of transmission capacity increases as hourly demand increases. Congestion may pose a problem and one of the solutions is to expand transmission capacity. In the Eastern trading bloc, almost every region has enough generating capacity to meet its own indigenous demand. The primary function of inter-regional power flow is improving the overall generation efficiency of the trading bloc. The focus for the trading bloc should be improving the reliability of the transmission capacity because it actually serves two functions: improving generation efficiency and improving supply reliability during peak hour by trade. Clearly, the utilization rate of a given transmission line determines the economics of additional investment. However, the utilization pattern determines if a delivery system is congested and poses a problem to the reliability of supply. Differences in market conditions may require different strategies to ensure reliability. For regions where inter-regional delivery capacity is not binding, the issue may be on fuel supply reliability and intra-regional delivery reliability of electric power. Fuel supply, prices, and storage capacity may be as important as power delivery reliability in these regions. In regions such as California, inter-regional transmission lines are critical and need to be addressed to ensure adequate supply from exporting regions. Intra-regional delivery reliability, which is not analyzed here, can also be examined using the same modeling methodology. The current U.S. electricity market can be characterized by aging infrastructure, evolving load centers, and heightened links between the power sector and natural gas markets. Many reports discussed investment requirements to improve the reliability of the power system. Solutions to the infrastructure and demand problems, however, are not unique and may vary with regions. Regions such as Florida, New England, and New York, have enough indigenous capacity so each region can meet most of its own demand. However, fuel availability and environmental constraints may limit the use of certain types of power plants. California, on the other hand, faces transmission constraints and may find answers in options such as adding generation capacity near the load center, expanding transmission capacity, adopting DG technologies, or adopting demand-side management. Answers to the questions of where to invest, what to invest in, and how much to invest, lay clearly in empirical data on regional demand, generation capacity, fuel availability, and power flow. References Electricity Monthly, Energy Information Administration, various issues. Electricity Transmission in a Restructured Industry: Data Needs for Public Policy Analysis, Energy Information Administration, DOE/EIA-639, December 24. Paul L. Joskow, Market for Power in the United States: An Interim Assessment, The Energy Journal, 26; 27, 1. National Transmission Grid Study, U.S. Department of Energy, May 22

12 Figure 1: A Simplified Algorithm for the Electricity Model Generation Capacity + Monthly Generation Electric power market Generation = demand Monthly Sales + Power Control Area 24-hourly load Regional Supply Curves by fuel type Trading Block Supply Curves by fuel type Hourly Hourly Market clearing Market prices clearing prices Hourly Regional Generation By fuel Electric Power Electric Power Imports/Exports Imports/Exports Regional average daily 24-hour load by month Trading Block 24-hour load Figure 2: Electricity Model Demand and Supply Regions AK AK Pacific Census Regions & Divisions of the United States Pacific HI HI WEST WA WA MT MT OR OR Pacific CA CA ID ID NV NV WY WY Mountain UT UT CO CO NM NM AZ AZ MIDWEST ND ND MN NORTHEAST MN ME VT ME VT N ew ew ngl nd SD WI E ngl a nd SD West WI NH NH North East MI NY MI NY Central IA IA North Middle NE NE Central Atlantic CT RIMA CT RIMA OH OH PA PA NJ NJ KS MO IN KS MO IN IL IL WV WV VA VA KY KY NC NC out OK East TN t la nt ic AR South S out h OK TN SC TX Central A t la nt ic AR SC TX West South Central LA LA MS AL MS AL GA GA FL FL SOUTH DE DE MD MD LEGEND REGION REGION DIVISION DIVISION STATE STATE

13 Figure 3: An Illustration of Solution Algorithm oil gas coal Total fossil supply Bin # Generation demand Figure 4: Comparison of Eastern Trading bloc Generation from Coal and Natural Gas (Unit=1 mwh) M1 4M7 5M1 5M7 6M1 6M7 Coal_hist COAL_proj NG_hist GA S_proj

14 Figure 5: Comparison of Western Trading bloc Generation from Coal and Natural Gas (Unit=1 mwh) M1 4M7 5M1 5M7 6M1 6M7 Coal_hisp COAL_proj NG_hist GAS_proj Figure 6: Comparison of Texas Generation from Coal and Natural Gas (Unit=1 mwh) M1 4M7 5M1 5M7 6M1 6M7 Coal_hist COA L_proj Gas_hist GAS_proj

15 Figure 7: Simulated Generation to Meet Load in the Western Trading Bloc In July 25 (Unit=million kw) LOAD8_7 GEN8_7 IMP8_7 LOAD9_7 GEN9_7 IMP9_7 LOAD1_7 GEN1_7 IMP1_7

16 Figure 8: Simulated Hourly Generation to meet Load in the Eastern Trading Bloc In August 25 (Unit=million kw) 25 Simulated 2 Hourly Generation 4 1 to Meet Load in the Easten Trading Bloc in August 25 (unit=million kwh) LOAD1_8 GEN1_8 IM P1_8 LOAD2_8 GEN2_8 IMP2_8 LOAD3_8 GEN3_8 IMP3_ LOAD4_8 GEN4_8 IM P4_8 LOAD5_8 GEN5_8 IMP5_8 LOAD6_8 GEN6_8 IMP6_ LOAD7_8 GEN7_8 IM P7_8 LOAD11_8 GEN11_8 IMP11_8 LOAD12_8 GEN12_8 IMP12_8

17 Table 1: An Example of Coal and Gas Capacity and their Cumulative Distribution (Unit: Mega Watt) Bin Number Coal Gas Cumulative coal Cumulative gas Table 2: An Illustration of Model Outputs for Each Trading Bloc Demand&Generation/ Region Region 1. Region J Trading Bloc Total Generation Hour 1 Hour Hour 24 r r = J G r r= 1 r = J G r r= 1 = J r= 1 G 1 = = r= 2 = = = 24 r = = r= 1 Trading Bloc Total Hourly Demand r J D1 r 1 r J D2 r r 1 r J D 24 r Generation per day h = 24 h= 1 G h1 h = 24 h= 1 G hj G hr = D hr

18 Table 3: Mean Absolute Percentage Error for Projected Generation from Coal and Gas East West Texas Coal Gas Coal Gas Coal Gas % 4.15% 1.92% 2.88% 2.39% 2.1% 24/5 3.6% 7.44% 2.44% 3.51% 3.32% 3.4% Table 4: Twelve month percentage difference East West Texas Coal Gas Coal Gas Coal Gas % 4.6%.62% -.83% -.23%.24%

19 Table 5: Total Generation, Generation from Gas and Petroleum Liquids, July 25 (unit=thousand megawatt hours) Region number & Regions Total Generation Natural Gas Oil Natural Gas share Oil share 1 New England % 8.41% 2 Mid-Atlantic % 3.7% 3 East North Central %.3% 4 West North Central %.49% 5 South Atlantic % 2.58% 6 East South Central %.62% 7 West South Central %.72% 8 Mountain %.4% 9 Pacific %.9% 1 California %.19% 11 Florida % 15.29% 12 New York % 14.87% 13 Texas %.5%