Release Date: October 14, 2020

EIA bases its energy forecasts and projections on data from numerous sources, combined with assumptions and mathematical methods. To understand the projections, their proper uses, and their limitations, customers need information on the underlying concepts and assumptions. EIA’s new Handbook of Energy Modeling Methods explains, in plain language, the processes EIA uses to produce long- and short-term projections of key energy market statistics.

To generate energy forecasts and projections, EIA develops and maintains three major modeling systems:

• The World Energy Projection System (WEPS) generates the annual global energy market projections, currently extending to 2050, published in the International Energy Outlook (IEO).
• The Short-Term Integrated Forecasting System (STIFS) generates the forecasts EIA publishes in the monthly Short-Term Energy Outlook (STEO). The monthly forecasts extend to the end of the calendar year after the publication year.
• The National Energy Modeling System (NEMS) generates the annual U.S. energy market projections, currently extending to 2050, published in the Annual Energy Outlook (AEO).

Each system is modular, and its components project energy supply, demand, or conversion sectors, such as natural gas supply, residential demand, and electricity conversion. The Handbook of Energy Modeling Methods contains a section for each module of each modeling system. Because we are still developing the Handbook, not all sections listed below are currently available.

The plain language module descriptions are in Part A of the Handbook. Part B is a technical appendix that provides in-depth descriptions of the main methods used in the three modeling systems.[1]

Handbook of Energy Modeling Methods

Part A. Plain Language Documentation

World Energy Projection System (WEPS)

• Overview
• Convergence
• Coal Supply
• Commercial Demand
• District Heat
• Electricity
• Greenhouse Gases
• Global Activity
• Industrial Demand
• Hydrocarbon Supply (coming)
  
• Residential Demand
• Transportation Demand (coming)
  

WEPS Documentation for Computer Programmers

Short-Term Integrated Forecasting System (STIFS)

• Overview
• Macroeconomics and Weather (coming)
• Carbon Dioxide Emissions (coming)
• International Liquid Fuels
• Crude Oil Prices
• Crude Oil & Natural Gas Production (coming)
• Motor Gasoline Consumption
• Non-gasoline Petroleum Products Consumption (coming)
• Petroleum Product Prices
• Petroleum Refining
• Petroleum Market Balances (coming)
• Hydrocarbon Gas Liquids
• Biofuels
• Natural Gas
• Coal
• Electricity Supply
• Electricity Consumption
• Electricity Retail Prices
  
• Distributed Solar

STIFS Documentation for Computer Programmers

National Energy Modeling System (NEMS)

• Overview (coming)
  
• Integrating Module (coming)
  
• Emissions Policy Module (coming)
  
• Macroeconomic Module (coming)
  
• International Module (coming)
  
• Residential Demand
• Commercial Demand (coming)
  
• Industrial Demand
• Transportation Demand (coming)
  
• Coal Supply (coming)
  
• Electric Power (coming)
  
• Natural Gas Market
• Oil and Gas Supply (coming)
  
• Liquid Fuels Market (coming)
  
• Renewables Module (coming)
  

NEMS Documentation for Computer Programmers

Part B. Technical Appendix

To module regional, national, and global energy markets, EIA draws methods from a variety of disciplines, including statistics, econometrics, and operations research. EIA also develops and adapts specialized energy modeling methods.

The STEO modules, along with the energy demand modules of the NEMS and WEPS, rely primarily on statistical modeling methods. In some cases, EIA has adapted and extended established methods to accommodate long-term projection. To project prices of hydrocarbon gas liquids, for example, the NEMS uses a multi-process Bayesian dynamic module, adapted to incorporate components that rely on judgment-based parameters.

The NEMS and WEPS supply modules rely heavily on econometric and operations research methods, primarily constrained optimization by linear or quadratic programming.

Key Modeling Methods

1. Linear Regression
2. Solving for Equilibrium
3. Electricity Load Shapes
4. Cointegration Analysis
5. Time Series Benchmarking
6. Mathematical Programming

Footnote
[1] Computer code for the NEMS, WEPS, and STIFS is available online. STIFS code appears within the modular User Guides.