Related Links 

Forcasting Page

Overview
Electricity Module Comparisons

• Regional Detail

• Competitive Pricing

• Estimation of Stranded Costs

• Estimation of Reserve Margins

This chapter compares the modeling approaches of NEMS and POEMS. It describes the general methodologies and points out the most significant differences between the models. With the exception of their electricity modules, NEMS and POEMS share the same components. In addition, they use many of the same data sources, and both are designed to produce annual forecasts through the year 2020. The primary differences lie in the treatment of electricity trade, dispatching, and pricing. A comparison of the two overall systems is given in Table 2, and a comparison of the electricity modeling differences is given in Table 3.

Table 2. Differences Between NEMS and POEMS

Table 3. Comparison of Major Differences Between EMM and TRADELEC^™

NEMS was developed by the EIA in 1991-1992 and was first used to produce the Annual Energy Outlook in 1993. NEMS consists of four demand modules (residential, commercial, industrial, and transportation), two conversion modules (petroleum and electricity markets), four supply modules (natural gas, oil, coal, and renewable fuels), a macroeconomic feedback module to represent the effects of the energy markets on the overall economy, and a module to represent the interaction between domestic and international energy markets. It produces a general equilibrium solution for energy supply and demand in the United States.

POEMS was developed with a more detailed representation of wholesale electricity markets than is incorporated in NEMS. A more disaggregated electricity module, TRADELEC^™, is used in place of the NEMS Electricity Market Module, to allow the examination of alternative assumptions for wholesale electricity markets (Table 2). Some of the data used in TRADELEC^™ are not available publicly, and the model is proprietary.

The overall decision process in TRADELEC^™ is similar to that in the EMM. The primary differences are in the regional detail and the methodologies used to estimate competitive electricity generation prices. In addition, the accounting of stranded costs and the estimation of reserve margins are somewhat different. These differences are summarized in Table 3.

Regional Detail

The EMM operates at the North American Electricity Reliability Council (NERC) region and subregion level. In total, the EMM represents 13 electricity supply and demand regions. TRADELEC^™ operates at the Power Control Area (PCA) level and represents 114 electricity supply and demand regions. Both models include historical data on more than 6,000 electricity generating plants.

The regional representations in both models have advantages and disadvantages. The more disaggregated nature of TRADELEC^™ makes it more suited for locational analysis issues, such as adding a new generating plant or transmission line. It also can be used to examine the impacts of alternative transmission pricing schemes--for example, pancaked pricing versus postage stamp pricing. Also, the additional regional detail allows the end-use service loads to be calibrated to PCA-level load data, which are readily available.

Both NEMS and POEMS rely on demand models that provide projections of the demand for electricity at the level of the nine Census divisions. NEMS uses fixed factors to allocate these demands to the 13 electricity regions. Similarly, POEMS uses fixed factors to allocate these demands to the 114 regions used in TRADELEC^™. The fixed allocation factor assumption is likely to be more reasonable at the 13 region level used in the NEMS than at the 114 region level used in POEMS. The variation in demand growth rates for the smaller POEMS regions is likely to be larger than the variation among the larger NEMS regions. In addition, gathering the data needed (some of which is proprietary) and solving the POEMS are resource intensive. The more aggregate nature of NEMS reduces the data and processing resources required compared to those needed for POEMS. In addition, NEMS, because it has fewer regions to solve, is able to incorporate a more detailed representation of electricity dispatch decisions (108 slices of time versus 72 in POEMS) and make use of publicly available data.

Competitive Pricing

Both the EMM and TRADELEC^™ base competitive generation prices on the marginal costs of producing power. However, the derivation of marginal costs is somewhat different in the two models. In TRADELEC^™, the marginal costs for each PCA and time period are based on the bid price of the next most expensive plant in the merit order. In other words, when plants place bids to meet a given demand, the winning plants are paid the bid price of the lowest cost plant that did not win--what is often referred to as a second price auction. The bid price for each plant (or plant group) consist of its fuel costs plus a user-specified fraction of its total O&M costs. In the Supporting Analysis CECA Competitive case, the fractions of total O&M costs assumed to be included in the bid price differ by plant type. For turbines the fraction of total O&M that is included is 100 percent; for combined-cycle and fossil steam units it is 50 percent; and for nuclear plants it is zero.

TRADELEC^™ uses a network algorithm to solve for the flow of power between PCAs. The trading algorithm accommodates both transfer limits between PCAs and the relative marginal costs (including cost of transmission losses and fees) of power between regions to identify the equilibrium price and flows of power. In addition, POEMS adds a charge to ensure that new turbine plants built by the model recover their costs. This charge is derived by calculating the revenue received by new turbines when they are paid the bid prices as described above. If the revenue is insufficient to cover the total costs of the facilities, a charge is added to increase the revenue they receive. This charge is assumed to be paid to both turbine and combined-cycle units, which have quick start capability.

The EMM uses a similar overall approach for the components of price. Marginal costs are calculated for each unit and used in a linear programming formulation for the operations of each of the 13 NERC regions. The EMM does not use a second price auction approach. Instead, the marginal energy price during each time period is set to the sum of the fuel plus variable O&M costs of the last plant dispatched. Trade between regions is constrained by limits on the transmission system. There are no transmission system constraints within regions. Without the representation of intraregional transmission constraints, the opportunity exists to dispatch plants at higher capacity factors than the transmission constraints may allow, thereby understating the cost of electricity generation. As a surrogate for intraregional constraints, maximum utilization rates for coal plants are imposed and phased in over a 5-year period. The EMM solves for the least-cost use of all available generation equipment to meet the demand for electricity, subject to any operational, environmental, and transmission constraints. Interregional trades are adjusted for losses, and a hurdle rate is used to account for the cost of transmitting power between regions.

The EMM uses a methodology different from TRADELEC^™ to calculate competitive generation prices. For each time period, the generation component of price consists of: (1) the marginal operating costs, (2) those taxes determined to be marginal costs, and (3) a reliability price adjustment equal to the marginal cost of unserved energy. The marginal operating costs in NEMS include fuel costs and variable O&M costs. However, in NEMS, variable O&M costs for each plant are derived from historical data. On average the variable O&M costs used in NEMS are lower than those used in POEMS. The reliability component represents the value of capacity during periods when capacity is in short supply. In other words, during periods when the demand for power is approaching the total amount of capacity available, the price of electricity will rise to account for the increased value of capacity when shortages are imminent. It is analogous to the capacity charge used in the deregulated electricity market in the United Kingdom.⁽¹⁸⁾ The theoretical underpinning for this approach comes from the work of Schweppe, Caramanis, Tabors, and Bohn.⁽¹⁹⁾

In both models, prices for transmission and distribution services are assumed to be regulated and are calculated to recover the total costs of these services. In order to parallel the Supporting Analysis, a performance-based rate algorithm was incorporated in EMM to reflect the assumed improvements in costs in the CECA Competitive case.

Estimation of Stranded Costs

Because of the more disaggregated regions represented, the asset accounting used in POEMS is done at the single utility level. This allows a more detailed accounting of stranded costs than is possible in NEMS. POEMS treats potential stranded costs resulting from generating assets, regulatory assets, and nuclear decommissioning differently. For generation assets, the net present value of the expected revenue stream of the plant is compared to the book value of the plant. When the discounted revenue stream is less than the book value of the plant, the difference represents stranded costs. On the other hand, when this value is greater than the book value of the plant, negative stranded costs result. It is assumed that all positive stranded costs are recovered over a 10-year period. When there are negative stranded costs, a portion is assumed to be returned to ratepayers. The portion returned varies by utility type. For private utilities 25 percent is assumed to be returned to customers, and for public utilities 100 percent is assumed to be returned to customers. Total regulatory assets and nuclear decommissioning costs that could be stranded in the absence of a fee for their collection are determined exogenously. The fees needed to recover regulatory assets and nuclear decommissioning costs are applied over 25 and 50 years, respectively.

NEMS does not have the ability to calculate stranded costs in as detailed a fashion as does POEMS. NEMS is able to calculate the aggregate difference in revenue available to meet fixed costs under cost-of-service versus competitive pricing. The discounted value of this revenue difference represents the net stranded costs. (Within each NEMS region positive and negative stranded costs are netted against each other.) In this analysis the more detailed estimates from POEMS were used for the calculation of regional electricity prices. For generating assets, net stranded costs are recovered over a 10-year period through a per-kilowatthour fee for all electricity sold. For regulatory assets and nuclear decommissioning costs, net stranded costs are recovered over 25 years and 50 years, respectively.

Estimation of Reserve Margins

Regional capacity reserve margins are an exogenous input in POEMS. For both cases of the Supporting Analysis, they were set to 8 percent in all regions and years, except for Florida where they were set to 4 percent. In the NEMS CECA Competitive case, the optimal amount of capacity to be built is determined as a function of the assumed value that consumers place on reliable electricity service, the marginal cost of capacity (assumed to be the cost of a simple combustion turbine), the demand profile, and the mix and performance of existing capacity in each region. In essence, the reserve margin is set at the point that the cost of the marginal unit of capacity is exactly equal to the amount that consumers would be willing to pay for the added reliability it provides.

Although it is complex and requires data that are not readily available (value of unserved energy), the approach used in NEMS takes account of the many factors that can change the amount of reserve capacity needed. For example, if the performance of existing plants improves, less reserve capacity is needed. On the other hand, if consumers invest in increasing amounts of sensitive electronic equipment (computers, telecommunications equipment, etc.), the value they place on reliability is likely to rise, and the need for reserve capacity to ensure that reliability would also increase. In the CECA Competitive case, the optimal reserve margins are lower than in the CECA Reference case because the assumed improvements in the availability of existing plants mitigate the effect of the lower reserve margin.