Regional Short-term Electricity Consumption Models Prepared for 25 th Annual North American Conference of the USAEE/IAEE, Denver September 18-21, 25 Frederick L. Joutz, Department of Economics The George Washington University Washington, Dc 252 (22) 994-4899 [e-mail bmark@gwu.edu] Dave Costello Office of Energy Markets and End Use (EMEU) Energy Information Administration U.S. Department of Energy (22-586-1468), [e-mail dave.costello@eia.doe.gov]
Outline of Presentation Brief Description of Model Framework of Model Sectoral Demand Descriptions Modeling Issues Elasticity Estimates Sample Forecasts
Regional Short-term Electricity Consumption Models The Regional Short-Term Energy Model (RSTEM-EC) is designed to provide analytical and forecasting support by the nine U.S. Census Regions and four particular states (California, Florida, New York, and Texas). Time series energy-econometric models of energy consumption, supply, and prices have been built for the electricity markets. These consumption markets for each region and particular state are broken out into four sectors: residential, commercial, industrial, and other. The consumption market equations are aggregated into regional models.
Regional Short-term Electricity Consumption Models The RSTEM-EC model determines monthly regional U.S. electricity consumption by four major sectors, including the residential, commercial, industrial and other sectors. The demand equations are based on time series energy econometric techniques. The specifications attempt to deal with issues related to dynamics, stationarity, integration, cointegration, seasonality, and structural breaks
Outline The models can generate short-term (up to 24 months) monthly forecasts of U.S. consumption in stand alone form and interacting with the regional electricity supply and price models. The RSTEM-EC consumption projections are linked to the Electricity Supply Model, The Supply Model uses a combination of inputoutput energy process models and econometric time series models to develop the generation of electricity, load dispatch, fuel consumption, and approximations of wholesale electricity trading and pricing across four major North American Electricity Reliability Council (NERC) regions..
RSTEM 13 Electricity Demand and Supply Regions AK Pacific Census Regions & Divisions of the United States Pacific HI WA OR Pacific NV CA WEST MT ID WY Mountain CO UT NM AZ ND SD NE TX West North Central OK West South Central MIDWEST NORTHEAST MN IA East North Central KS MO IL IN AR LA WI KY TN East South Central AL MS SOUTH MI OH WV South Atlantic GA Middle Atlantic PA SC FL VA VT NY NC ME New England NJ DE NH MA CT RI MD LEGEND REG ION DIVISION STATE
RSTEM Regional Electricity Demands (Excl. Own Use and Direct Sales) From electricity supply model GI Model Regional Macro Data Inputs Regional Electricity Production Costs? Regional End-Use Prices Regional Distribution Cost Factors NOAA/EMEU CNEAF Regional Weather Assumptions Residential Commercial Industrial Other Demand Demand Demand Demand To U.S. Electricity Balance Module
Residential Model There are economic, seasonal, and trend issues which need to be addressed in the residential models. The primary determinants of residential electricity consumption in a region are: occupied housing units or households; share of occupied housing units or households using electricity as the primary energy source for heating; share of occupied housing units or households with installed air conditioning; cooling degree-days; heating degree-days; real delivered residential per-unit electricity charges; and real personal income per household.
Commercial Model The primary determinants of commercial electricity consumption in a region are: employment, hours worked or real wage disbursements in commercial employment (as a proxy for commercial sector output); cooling degree-days; heating degree-days; real commercialsector per-unit delivered electricity charges; and self-generating capacity in the commercial sector. Autonomous trends in commercial electricity use intensity are additional factors in commercial demand related to commercial building shell efficiencies, average commercial floor-space per unit of output, and penetration of electricity-using equipment in commercial establishments.
Industrial Model The primary determinants of industrial electricity consumption in a region are: employment, hours worked or real wage disbursements in industrial employment or industrial output as measured by the Federal Reserve Board; real industrial-sector per unit delivered electricity charges; cooling degree-days; heating degree-days; and selfgenerating capacity in the industrial sector. Autonomous trends in industrial electricity use intensity are additional factors in industrial demand related to energy efficiency trends in industrial processes and equipment, shifts in regional industrial output patterns among industry sectors of varying levels of electricity use intensity per unit of output, and penetration of general electricity-using equipment in industrial establishments.
Modeling Issues The sectoral demand model specification begins by analyzing the time series properties of the energy and economic variables. Autocorrelation functions and partial autocorrelation functions of the series in (log) levels, first differences, and annual differences are examined to look for dynamic patterns and integration. Plots of the series in (log) levels, first differences, annual differences, and monthly seasonal stacks are examined for trends, seasonal effects, shifts in the data generating process, and outliers. Unit root tests are conducted to test for (seasonal) integration and cointegration. Error-correction models are developed where appropriate. The principles and approach of the General-to-Specific Modeling have not been fully implemented in the modeling effort to-date because of time and resource constraints.
General-to-Specific Modeling Empirical models are at best approximations of the true data generating process (DGP). There are identifiable structures in the model that are interpretable in light of economic theory. A congruent model has the following six properties. The residuals must be white noise for the model to be a valid simplification of the DGP. The model must be admissible on accurate observations. For example nominal interest rates and prices cannot be negative. The conditioning variables are at least weakly exogenous for the parameters of interest in the model. Forecasting models require strong exogeneity, while policy models require super exogeneity. The parameters of interest must be constant over time and remain invariant to certain classes of interventions. This relates to the purpose of the model in the previous criteria. The model must be able to explain the results from rival models; it is able to encompass them.
Regional STEO Elasticity Estimates (Draft USAEE) Region East South Central East North Central Mid-Atlantic Mountain New England Pacific South Atlantic West North Central West South Central California Florida New York Texas Price Elasticities Short-run -.9 -.2 -.3 -.7 -.2 -.14 -.5 -.11 -.1 -.15 -.15 -.1 -.1 Long-run -.4 -.6 -.59 -.6 -.62 -.55 -.22 -.23 -.33 -.48 -.3 -.8 Income Elasticities Short-run.64.34.32.17.34.9.11.8.7.1.15.75.33 Long-run 1.27 1.4.72.44 1.4.81.47.45.55.33.48 1.29.57
Region Regional STEO Elasticity Estimates (Draft USAEE) Cooling Degree Day Elasticities Heating Degree Day Elasticities Short-run Long-run Short-run Long-run East South Central.27 1.16.22.95 East North Central.5.23.18.37 Mid-Atlantic.33.56.13.28 Mountain.3 1.52.17.43 New England.5.14.18.37 Pacific.13 1.38.14.64 South Atlantic.4.64.11.19 West North Central.45.72.24 2.2 West South Central.68 5.5.1.8 California.16.51.11.34 Florida.16.56.11.28 New York.9.27.5-.14.5 Texas.19.33.8.13
California Commercial Consumption per Day BkW Industrial Consumption per Day BkW 36 16 34 32 14 2 1-1 -2 3 28 26 24 2 1-1 12 1 8-3 2M1 2M7 3M1 3M7 4M1 4M7-2 2M1 2M7 3M1 3M7 4M1 4M7 Residential Consumption per Day BkW Total Consumption per Day BkW 2 1 32 28 24 2 16 15 1 5 85 8 75 7 65 6 55-1 -5-2 2M1 2M7 3M1 3M7 4M1 4M7-1 2M1 2M7 3M1 3M7 4M1 4M7
South Atlantic Commercial Consumption per Day BkW Industrial Consumption per Day BkW 8 4 9 85 8 75 7 65 6 55 15 1 5 56 52 48 44 4 36-4 -8-5 -12 2M1 2M7 3M1 3M7 4M1 4M7-1 2M1 2M7 3M1 3M7 4M1 4M7 Residential Consumption per Day BkW 12 EXTCP_S A C 26 12 8 11 1 9 8 7 6 4 24 22 2 18 4 6 2 16-4 -2-8 2M1 2M7 3M1 3M7 4M1 4M7-4 2M1 2M7 3M1 3M7 4M1 4M7
Commercial Consumption per Day BkW Middle Atlantic Industrial Consumption per Day BkW 48 24 44 23 5-5 4 36 32 15 1 5 22 21 2 19-1 -5-15 2M1 2M7 3M1 3M7 4M1 4M7-1 2M1 2M7 3M1 3M7 4M1 4M7 Residential Consumption per Day BkW 5 EXTCP_MAC 12 45 11 15 1 5 4 35 3 25 6 4 2 1 9 8-2 -5-4 -1 2M1 2M7 3M1 3M7 4M1 4M7-6 2M1 2M7 3M1 3M7 4M1 4M7
United States Residential Consumption per Day BkW Commercial Consumption per Day BkW 5 38 1 5 45 4 35 3 25 2-2 36 34 32 3 28 26-4 -6-5 2M1 2M7 3M1 3M7 4M1 4M7-8 2M1 2M7 3M1 3M7 4M1 4M7 Industrial Consumption per Day BkW 3 EXTCP_US 1.2E+7 29 28 27 1.1E+7 1.E+7 5 26 25 24 4 2 9.E+6 8.E+6-5 -2-1 2M1 2M7 3M1 3M7 4M1 4M7-4 2M1 2M7 3M1 3M7 4M1 4M7
Real GSP by Census Division (million $2) 24 2 16 12 8 4 199 1992 1994 1996 1998 2 22 24 26 CGSPC_NEC CGSPC_MAC CGSPC_ENC CGSPC_WNC CGSPC_SAC CGSPC_ESC CGSPC_WSC CGSPC_MTN CGSPC_PAC
Real Wage Disbursements by Region (Million $2) 1 9 8 7 6 5 4 3 2 1 199 1992 1994 1996 1998 2 22 24 26 CWDC_NEC CWDC_MAC CWDC_ENC CWDC_WNC CWDC_SAC CWDC_ESC CWDC_WSC CWDC_MTN CWDC_PAC
Manufacturing Output Indexes (1997=1.) 15 14 13 12 11 1 9 8 7 6 199 1992 1994 1996 1998 2 22 24 26 IPMFGC_ENC IPMFGC_ESC IPMFGC_MAC IPMFGC_MTN IPMFGC_NCR IPMFGC_NEC IPMFGC_NER IPMFGC_PAC IPMFGC_SAC IPMFGC_SOR IPMFGC_WER IPMFGC_WNC IPMFGC_WSC
Model Structure -1of 2 The regional electricity model has 13 regions, which are grouped into three trade zones: Eastern Interconnection Western Interconnection Texas Each region has its own demand and supply representations A 14 th region (Hawaii + Alaska) is treated separately (and simply) for completeness in deriving U.S. aggregate demand and supply
Forecast Model Data Flow