Forecasting electricity sales over the short term: a comparison of new methodologies. Final report

Technical Report · Mon Sep 01 04:00:00 EDT 1986

OSTI ID:5288230

Engle, R F; Goodrich, R

This project examined the accuracy of a variety of new time series forecasting methods for the short term prediction of state wide monthly residential electrical sales per customer. Using a formal experimental design across ten states in the US, models were fit to historical data from 1962 through 1977, and evaluated for their performance over the following three year period through 1980. Rolling forecasts were generated on a monthly basis through the evaluation period to obtain measures of forecast accuracy for different horizons. The forecasting models included several econometric techniques and several pure time series techniques. The study concluded that there were gains in accuracy for structural econometric models using exogenous variables, for dynamic models involving lagged dependent variables and serial correlation, and for adaptive models. It was further concluded that the systematic use of model selection criteria such as the Akaike Information Criterion, and of diagnostic tests, led to improved accuracy in the models.

• Research Organization: Quantitative Economic Research, Inc., San Diego, CA (USA); Scientific Systems, Inc., Cambridge, MA (USA)
• OSTI ID: 5288230
• Report Number(s): EPRI-EM-4772; ON: TI87920015
• Country of Publication: United States
• Language: English

Related Subjects

29 ENERGY PLANNING, POLICY, AND ECONOMY
296000* -- Energy Planning & Policy-- Electric Power
ACCURACY
CORRELATIONS
ECONOMETRICS
ECONOMICS
ELECTRIC POWER
ERRORS
FORECASTING
LOAD MANAGEMENT
MANAGEMENT
MATHEMATICAL MODELS
POWER
POWER DEMAND
RESIDENTIAL SECTOR
SALES
STATISTICAL MODELS