Automatic load forecasting. Final report

Technical Report · Sun Mar 01 04:00:00 EST 1981

DOI:https://doi.org/10.2172/6804946· OSTI ID:6804946

Nelson, D. J.; Vemuri, S.

A method which lends itself to on-line forecasting of hourly electric loads is presented and the results of its use are compared to models developed using the Box-Jenkins method. The method consists of processing the historical hourly loads with a sequential least-squares estimator to identify a finite order autoregressive model which in turn is used to obtain a parsimonious autoregressive-moving average model. A procedure is also defined for incorporating temperature as a variable to improve forecasts where loads are temperature dependent. The method presented has several advantages in comparison to the Box-Jenkins method including much less human intervention and improved model identification. The method has been tested using three-hourly data from the Lincoln Electric System, Lincoln, Nebraska. In the exhaustive analyses performed on this data base this method produced significantly better results than the Box-Jenkins method. The method also proved to be more robust in that greater confidence could be placed in the accuracy of models based upon the various measures available at the identification stage.

• Research Organization: Nebraska Univ., Lincoln (USA). Dept. of Electrical Engineering
• OSTI ID: 6804946
• Report Number(s): EPRI-EL-1758
• Country of Publication: United States
• Language: English

Related Subjects

200300 -- Electric Power Engineering-- Power Transmission & Distribution-- (-1989)
24 POWER TRANSMISSION AND DISTRIBUTION
29 ENERGY PLANNING, POLICY, AND ECONOMY
292000* -- Energy Planning & Policy-- Supply
Demand & Forecasting
ELECTRIC UTILITIES
ENERGY SYSTEMS
FORECASTING
LOAD ANALYSIS
LOAD MANAGEMENT
MANAGEMENT
MATHEMATICAL MODELS
POWER DEMAND
POWER SYSTEMS
PUBLIC UTILITIES
TEMPERATURE DEPENDENCE
WEATHER