Bayesian forecasting for adaptive seasonal short-term load forecasting models with a comparative study
Short-term electric load forecasting is the prediction of hourly electric power demand for a service area of a utility. These forecasts of megawatt demand are used to perform scheduling and dispatching of generation units. Over the past twenty years, numerous statistical and control methods have been proposed to generate these forecasts. Nearly all these methods, however, are inadequate in how they handle the season to season variability exhibited by the load data as a function of weather variations. In this study a structural model is proposed that can be used for all seasons. It is an adaptation of Harrison-Stevens Bayesian forecasting model. This dynamic multi linear model produces forecasts by combining results from several models using a weighting function based on prior and posterior probabilities. In addition, an adaptive form of a regression model used in association with a discriminant function is developed in this research. A comparative analysis of these two models for three different data sets is carried out using several descriptive performance measures. The results indicate that the Harrison Stevens model performs slightly better than the adaptive regression model. Furthermore, a limited investigation indicates that a Box Jenkins approach is not well-suited for this application.
- Research Organization:
- Arizona State Univ., Tempe, AZ (USA)
- OSTI ID:
- 6361760
- Resource Relation:
- Other Information: Thesis (Ph. D.)
- Country of Publication:
- United States
- Language:
- English
Similar Records
Load forecasting via suboptimal seasonal autoregressive models and iteratively reweighted least squares estimation
Regional load-curve models: scenario and forecast using the DRI model. Final report. [Forecasts of electric power loads in 32 US regions]
Related Subjects
29 ENERGY PLANNING
POLICY AND ECONOMY
ELECTRIC UTILITIES
LOAD MANAGEMENT
MATHEMATICAL MODELS
CALCULATION METHODS
COMPARATIVE EVALUATIONS
ELECTRIC POWER
ENERGY DEMAND
FORECASTING
PERFORMANCE
POWER GENERATION
PROBABILITY
REGRESSION ANALYSIS
SCHEDULES
SEASONAL VARIATIONS
WEATHER
DEMAND
MANAGEMENT
MATHEMATICS
POWER
PUBLIC UTILITIES
STATISTICS
VARIATIONS
200600* - Fossil-Fueled Power Plants- Economic
Industrial
& Business Aspects- (1990-)
296000 - Energy Planning & Policy- Electric Power