The Energy Information Administration (EIA) of the U.S. Energy Department (DOE) developed the Short-Term Integrated Forecasting (STIFS) model to generate short-term (up to 24 months) monthly forecasts of U.S. supplies, demands, imports, stocks, and prices of various forms of energy. The purpose of this report is to define the electricity model in STIFS and describe its basic properties. This report documents the May 1998 version of the electricity model equations in STIFS.
This report is written for persons who want to know how short-term energy markets forecasts are produced by EIA. The report is intended as a reference document for model analysts, users, and the public.
The STIFS model determines monthly aggregate U.S. electricity demand by four major sectors and provides a national-level supply balance. In STIFS, U.S. electricity supply is comprised of two major components: domestic net electricity generation (that is, electric power actually transmitted to the transportation grid by electric utility-owned and nonutility-owned power plants) and net electricity imports (principally from Canada). Generation sources (fuels used in power production) identified in STIFS are coal, petroleum, natural gas, nuclear power, hydroelectric and other renewables, including wind and solar, wood and waste, and geothermal. A catchall category representing the total of transportation and distribution losses of electricity and other items, including any pure statistical discrepancy between electricity supply and electricity demand, rounds out the demand/supply balance.
The electricity module of STIFS is structured so as to be highly recursive in the following sense. Demand by sector is determined, leading to a calculation of aggregate electricity demand. Working backward from total demand, a certain level of transmission and distribution losses is calculated. Demand plus transmission and distribution losses equals total electricity gross supply. Contributions from non-utility generators (NUGs) and imports (both of which are determined outside of STIFS and are therefor exogenous) are subtracted from total supply. This yields a net electricity generation total for domestic electric utilities. Total generation and a number of exogenous factors determine most of the categories of electricity production by fuel source, except that relative prices help determine the portions of electricity generation contributed by fossil fuel-fired units.
Electricity demand is measured in terms of monthly sales divided by the number of days in the month. For electricity demand, reported monthly sales are not strictly related to consumption in the month that they are reported. This is because reported sales are on a billing-cycle rather than calendar month basis. A particular month's reported electricity sales volume actually represents consumption by customers for part of the current month and part of the previous month. The current version of STIFS models electricity sales on an 'as-reported' basis, but includes lagged as well as current values for some key determinants (particularly weather) to compensate for the billing lag problem.
Total electricity demand is calculated for four broad sectors: (1) residential; (2) commercial; (3) industrial; and (4) "other" (public street and highway lighting, other sales to public authorities, sales to railroads and railways, and interdepartmental sales.) The main determinants of electricity demand in the STIFS estimating equations are: household growth (residential sector); changes in commercial employment (commercial); growth in manufacturing output (industrial) or in real GDP (other); weather (residential and commercial); general seasonal factors (all); and trends in demand intensity (residential, commercial, and other). Trends in the intensity of electricity use in the aggregate sectors are the net result of several (possibly counteracting) factors, such as cumulative efficiency changes, demographic shifts (for example from geographical areas with typically high electricity intensities to areas with low ones, or vice versa), increased penetration of electricity-using equipment (such as computers, electronic appliances, etc.), and so on. STIFS does not account for the separate components of these long-run trend factors, but only provides a quantification of the net result.
The principal determinants of short-term demand variations in the residential sector (ESRCPUS) are weather factors, although a significant trend in consumption per household persists in raising demand from year to year. For weather, non-zero parameter values for heating degree-days (ZWHDPUS) or cooling degree-days (ZWCDPUS) are allowed only during the season in which particular weather impacts are meaningful for the aggregate data. Thus, heating degree-days affects residential demand only from October through April. This technique is designed to improve the credibility of the separate estimates of heating and cooling degree-day effects on electricity demand, which might otherwise be confounded due to a close (negative) correlation between aggregate cooling and heating degree-days. In addition, growth in the total number of households in the United States adds proportionately to electricity demand, other things being equal.
|
While weather also measurably affects demand in the commercial sector (ESCMPUS), this factor is less important than strong, persistent trends in the intensity of electricity use due to the continued introduction of electronic equipment such as microcomputers into the work place. For a given level of expected electricity use per commercial employee per day, average daily consumption of electricity in the commercial sector is expected to increase (or decrease) proportionately with employment, leading to a somewhat more cyclical short-run pattern of underlying demand (that is, demand corrected for noneconomic factors such as weather) than is to be expected in the residential sector.
|
Of the major electricity demand sectors, industrial demand (ESICPUS) is undoubtedly the most cyclical, being so dependent upon movements in the level of industrial output (ZOMNIUS), with a particularly evident differential effect from variability in paper industry output (ZO26IUS). Seasonal shifts in industrial demand are apparent, although typical measurements of weather variability generally are not as significant a set of determinants of electricity demand in the industrial sector as in the residential and commercial sectors.
|
For the "other" electricity demand category (ESOTPUS), the simple assumption that, other things being equal, demand growth will be proportional to growth in the overall economy (in terms of real GDP) provides a good basis for estimating short-run demand changes. Aside from seasonal variation, there is no apparent tendency for this particular category of electricity demand to grow at noticeably different rates from those for the overall economy.
|
Identities for re-transforming estimated adjusted electricity demand values into daily consumption rates:
ESRCPUS = ESRCPUSQ * KQHMPUS
ESCMPUS = ESCMPUSQ * EMCMPUS
ESOTPUS = (ESOTPUSQ/1000) * GDPQXUS
CCOEPUS = ESOTPUS - TROEPUS
(where TROEPUS = exogenous estimated transportation electricity demand).
Total electricity utility sales (ESTCPUS) is then determined by the identity:
ESTCPUS = ESRCPUS + ESCMPUS + ESICPUS + ESOTPUS
In addition, STIFS accounts for an estimated amount of electricity produced by nonutility generators and consumed for own account (ESNTPUS).
Thus, total electricity demand is:
ESTXPUS = ESTCPUS + ESNTPUS
The supply of electricity as presented in STIFS refers to electricity produced by regulated electric utilities (ELEOPUS) or nonutility generators (NTEOPUS) plus any net imports (ELNIPUS). Electric power generation is broken down into components by power source (that is, by fuel category). Generation from coal, petroleum, natural gas, nuclear power, hydroelectric power and geothermal and other renewable power sources is covered. However, all information relating to the operation of NUGs is provided from outside of STIFS and is therefor exogenous.
Modeling the Canadian electricity market is outside of the scope of STIFS and, therefore, imports are taken as exogenous. A major factor in the availability of electricity imports from Canada is the condition of watersheds affecting Canadian hydroelectric output. The exogenous information on imports is constructed to take account of most-likely outcomes for such key Canadian supply factors. Special scenarios, which include variations in assumed electricity imports, can be used to investigate the significance of greater or lesser availability of Canadian supply.
In the short run, nonutility supply is determined largely by available capacity because additional nonutility supply (capacity) entails building new units which take a number of years to complete. The capacity information used to determine supply in the Short-Term Energy Outlook is obtained using the EIA Form 867, Annual Nonutility Power Producer Report, which reports existing and planned nonutility units. The utilization of this capacity (to calculate nonutility sales to utilities and generation for own-use) are determined based on history. However, the utilization of these units can be varied to reflect different scenarios.
Net electricity generation from nuclear power plants and from hydroelectric facilities is exogenous. Detailed analysis of short-term hydroelectric and nuclear power availability is done regularly for routine STIFS runs, but the results are generally assumed to be insensitive to any alternative scenarios considered in simulations of STIFS. For nuclear power, under normal circumstances, only unforeseen downtime would alter expectations about generation patterns, since these plants provide almost exclusively baseload rather than incremental or peaking power capacity. A similar argument applies to hydroelectric power, except that the unknown factor is the relative abundance (or lack) of rainfall and snowpack to feed watershed levels. The nuclear and hydroelectric power forecasts for STIFS are supplied by EIA's Office of Coal, Nuclear, Electric and Alternate Fuel, using the Short-Term Nuclear Annual Power Production Simulation (SNAPPS) model and the Short-Term Hydroelectric Generation Model (STHGM).
In the electricity model, the demand side is linked to the supply side by exploiting observed regularities in the difference between reported total electricity demand (ESTXPUS) and total electricity supply, as defined by the sum of the three main supply components mentioned above (ELEOPUS + NTEOPUS + ELNIPUS). This difference (TDLOPUS) includes any transmission and distribution loss of electrical energy, errors related to the nature of the billing-cycle problem, and other discrepancies related to the independent measurement of electricity supply and demand. It is assumed that, other things being equal, TDLOPUS will be proportional to ESTCPUS. More specifically, the ratio of the two variables (labeled TDLOFUS) is modeled as being constant except for some regular seasonal variation.
|
TDLOPUS = - (TDLOFUS * ESTCPUS )
Once TDLOFUS and TDLOPUS are calculated, total electricity generated by electric utilities and NUGs (TSEOPUS) can be determined.
TSEOPUS = ESTXPUS + TDLOPUS - ELNIPUS
Electric utility output is calculated as
ELEOPUS = TSEOPUS - NTEOPUS
Given exogenous estimates for nuclear and hydroelectric power as well as functionally independent estimates for renewable sources of supply other than hydropower, the requirements for generation from fossil fuel sources can be determined.
Wind- and solar-powered electricity generation (WNEOPUS) as well as the wood and waste category (WWEOPUS) exhibit some seasonality but no discernible trend, on balance appearing to have remained about flat since 1989 or 1990.
|
|
Geothermal-based generation (GEEOPUS) has exhibited a significant downward trend since at least 1987, reflecting the much less favorable tax treatment afforded such projects following the Tax Reform Act of 1986.
GEEOPUS = exp (GEEO_01 + GEEO_02 * TIME)
It remains to determine the portions of electric utility output contributed from coal-, natural gas- and petroleum-fired generating units. Generation from these sources is estimated in a 3-step procedure: 1) initial values for coal generation (CLEOPUSX) and natural gas generation (NGEOPUSX) are estimated jointly. In contrast to the assumptions applied to nuclear power and hydroelectric generation, the amount of fossil fuel-based power generation (CLEOPUS, NGEOPUS and PAEOPUS) is assumed to be strongly influenced by changes in the overall load on the power generating system (ELEOPUS):
|
[Click here for Regression Results]
|
Given some level for aggregate system load (ELEOPUS), coal is affected negatively by significant changes in nuclear power and hydroelectric power availability. The extent of these negative effects will tend to vary across regions of the country, tending to limit the accuracy of the above equation when estimating the effects of highly localized changes in nuclear or hydroelectric power. STIFS, however, does make an important correction to this problem in the case of hydroelectric power by including a separate response parameter for shifts in hydroelectric power availability in the Pacific region (HYEOPAC): impacts of high hydroelectric output there are likely to affect coal less than in other regions. A similar correction (for opposite reasons) appears in the equation for gas-fired generation (NGEOPUSX) below.
|
Once coal-based and natural gas gas-based generation is determined, all that remains is to calculate generation from petroleum (PAEOPUSX), according to the following identity:
PAEOPUSX = ELEOPUS - CLEOPUSX - NGEOPUSX - HYEOPUS - NUEOPUS - GWEOPUS.
Because of potential quantity constraints on certain natural gas supply variables, electric utility fossil fuel-based generation variables are calculated as temporary variables, which are indicated by the "X" at the end of the variable names (NGEOPUSX, CLEOPUSX, etc.). Outside of the electricity model, STIFS checks for whether or not initially calculated total natural gas demand is within assumed deliverability limits. (See the Natural Gas Model section for details on gas supply constraints.) If initially calculated natural gas demand exceeds the supply constraints, demand cutbacks may automatically be enforced (in the electric utility and industrial sectors only), unless accommodating changes in inventory patterns or higher price trajectories (or both) are instituted. The STIFS model automatically calculates final demand and supply quantities that may be equal to or less than quantities initially calculated.
Once a final (possibly truncated) level for natural gas consumed at electric utilities (NGEUPUS) is determined in the natural gas portion of the model, final fossil fuel-based generation amounts are determined as follows:
NGEUPUSX = NGEOPUSX * NGEOKUS / NGEUKUS;
NGEUPUS = NGEUPUSX * (1 - (NGTCPUSX - NGTCPUS) /(NGINPUSX+NGEUPUSX+NGACPUSX+NGLPPUSX));
NGEOPUS = NGEUPUS * NGEUKUS / NGEOKUS;
PAEOPUSX = ELEOPUS-CLEOPUSX-NGEOPUSX-HYEOPUS-NUEOPUS-GWEOPUS;
CLEOPUS = (CLEOPUSX/(CLEOPUSX+PAEOPUSX))*(ELEOPUS-NGEOPUS-HYEOPUS-NUEOPUS-GWEOPUS);
PAEOPUS = (PAEOPUSX/(CLEOPUSX+PAEOPUSX))*(ELEOPUS-NGEOPUS-HYEOPUS-NUEOPUS-GWEOPUS);
For all normal model runs, based on historical averages, generation from distillate fuel (DKEOPUS) is assumed to be require 45,000 barrels per day of light fuel oil, petroleum coke is assumed to be 5 percent of the requirements remaining, and residual fuel is assumed to meet the rest
DKEOPUS = 0.045 * DFTCZUS / DFEOKUS;
PCEOPUS = 0.05 * (PAEOPUS-DKEOPUS);
RFEOPUS = PAEOPUS - DKEOPUS - PCEOPUS;
RFEUPUS = RFEOPUS * RFEOKUS / RFTCZUS;
DKEUPUS = DKEOPUS * DFEOKUS / DFTCZUS;
PCEUPUS = PCEOPUS * PCEOKUS / PCTCZUS;
DFEPPUS = DKEUPUS - (LAG1(DKSEPUS) - DKSEPUS)/ZSAJQUS;
RFEPPUS = RFEUPUS - (LAG1(RFSEPUS) - RFSEPUS)/ZSAJQUS;
XTDSEL = DFEPPUS;
On rare occasions, the share of total oil-based generation from distillate-fired units (DFSHR) has been seen to rise sharply to levels significantly above normal average levels, which are generally quite low. When certain regions of the country are faced simultaneously with little or no excess generating capacity and high gas demand by firm customers, extraordinarily high reliance on relatively expensive peaking units may be resorted to as a short-run electricity supply option. If this happens during the winter heating season (which would be expected to be the case), significant impacts on fuel oil prices and quantities could result depending on the general supply and demand conditions in those markets. The effect of any propensity to heavily utilize distillate fuel oil for electricity generation in times of extreme peak demand can be handled through special simulations of STIFS in which DKEOPUS is allowed to assume extreme values.
File last modified: August 13, 1998
This page's URL: http://www.eia.doe.gov/emeu/steo/pub/document/textel.html