Regionalizing the ShortTerm Energy Outlook Meeting of the American Statistical Association ASA Commi

1
Regionalizing the Short-Term Energy
OutlookMeeting of the American Statistical
Association (ASA)Committee on Energy Statistics

• Margot Anderson
• Director, Energy Markets and End Use
• Energy Information Administration
• April 28, 2005

2
Topics

• Purpose and Objectives
• Schedule
• Overview and Key Components
• Challenges

3
Purpose and Objectives

• Maintain a national perspective for EIAs
  short-term outlook while providing regional
  detail to better understand U.S. energy
  markets.
• Add relevance Provide EIA customers with
  outlook relevant to their state or region.
• Add depth Gain better understanding of
  local/regional supply problems (pipeline/transmiss
  ion constraints, regulations, etc.), which
  national model ignores or distorts.
• May add accuracy National-level forecast based
  on regional forecasts may be more accurate .

4
Schedule

• Began, mid-2003 Completion, late summer 2005.
• Ongoing Model development, testing,
  documentation
• Expert Review Spring/Summer 2005
• Web-Products
• Continue monthly release (expand output?)
• Create downloadable PC version?
• Expand web presentation and query system to
  include regional detail?
• Provide all documentation

5
Overview and Key Components

• RSTEM expands EIAs existing Short-Term
  Integrated Forecasting System model (STIFS) to
  include regional demand and price determinations
  for most energy products.
• Econometric model captures demand, inventories
  and prices and is used to forecast energy markets
  for key U.S. sectors and regions.
• RSTEM allows more detailed treatment of the key
  trends and forces in major domestic energy
  markets than was possible under STIFS.
• The frequency of the model database remains
  monthly, and the forecast horizon remains at 12
  to 24 months ahead.

6
RSTEM Components

• Electricity
• Demand
• Prices
• Load Generation
• Dispatch Model
• Natural Gas
• Prices and Storage
• Supply
• Demand

• Petroleum
• Heating Oil
• Gasoline
• Propane
• Macro Bridge

7
Degree of Regionalization Varies

• Significant regional detail for natural gas and
  electricity
• Strong regional differences AND available
  regional data.
• Census Division detail on electricity and natural
  gas demand.
• Market area price determination for spot natural
  gas.
• Census-Division (and some State) level
  electricity supply determination.
• More limited regional detail
• PADD-level pricing for key petroleum products
  (gasoline, 2 distillate, propane).
• Census Division detail on household
  heating/cooling costs

8
Variable Counts STIFS RSTEM
9
Judging Model Adequacy

• Used generally accepted criteria
• Evaluated data for measurement error
• Checked overall fit (R-squared)
• Tested independence of residuals
• Examined residuals for systematic patterns
• Tested significance of coefficients
• Checked for bias in forecasts (mean absolute
  percent error)
• Etc.

10
Selected RSTEM Challenges

• Determining how electricity generation will be
  dispatched in different regions each month.
• Avoiding aggregation errors.
• Maintaining model and data quality.
• Developing publicly accessible outputs and
  products.
• Developing appropriate diagnostic and model
  evaluation tools.