Projecting Calibration Variables into the Future

The process of generating a forecast using ENERGY 2100 involves reviewing, analyzing, and determining the best method of projecting the calibration variables. An initial set of projection methods is set up as a default. Experience generating energy forecasts with the model has enabled refinement of the initial projection methods chosen; however, each year these methods are considered starting points for the full process of developing the forecasted calibration values. 

A multitude of automated forecasting methodologies are built into the model, and other methods are developed as needed. The initial projection method assigned for each variable (by area, industry, fuel, and enduse) are listed below.

Default Projection Methods of Calibration Variables

Variables

Projecting Calibration Variable into Future

Default Methods

MMSM0

Non-Price Factor
  • Use average historical market share value (Res/Com/Ind)
  • Use average historical non-price factor (Trans)

CERSM

Lifestyle Multiplier
  • Assign future values equal to exogenous values scaled to the last historical year’s value

CUF

Capacity Utilization
  • Trend future values to an exogenous value

FsPEE

Feedstock Process Efficiency
  • Assign future values equal to exogenous values scaled to the last historical year’s value

DEMM

Maximum Device Efficiency Multiplier
  • Assign future values equal to exogenous values scaled to the last historical year’s value

PEMM

Maximum Process Efficiency Multiplier
  • Assign future values equal to exogenous values scaled to the last historical year’s value


Given model results from the initial projection methods, users can conduct an analysis (combined with expert knowledge or expectations of the future energy market) to determine the best methods for projecting the calibration variables. Analysts review and analyze the historical values of the calibration variables, the impact of the historical energy data on the calibration variables, and the impact of the initial values on the forecast. Modifications to the methods used may be made for specific areas, industries, fuels, or enduses by creating adjustment files which overwrite the initial methodology assigned. 

The model is re-calibrated when there are changes to model structure or as new data becomes available. The calibration process is automated so the only manual effort is to review and resolve the issues introduced by new model structure or new data. 

In the current version of ENERGY 2100, the demand equations for the U.S. forecast are calibrated to match the U.S. EIA’s Annual Energy Outlook (AEO) forecast. Historical U.S. data are input from U.S. government sources (SEDS, SEPER, Form 860, etc.). The demand equations for Canada are calibrated throughout the historical period, and the calibration variables are projected using the default methods above, with further adjustments as desired by analysts.