Extended Abstract (120K)5.1
Developing improved tools for electric utility peak load forecasting
David J. Sailor, Tulane University, New Orleans, LA; and P. B. Breslow
Short term forecasting of electricity demand is playing an increasingly important role for utility companies as energy markets become more competitive. This is particularly true in warm summertime climates where the air-conditioning demand defines the peak load and dictates the required generating capacity. Utility restructuring in many states has provided increased incentive for the electric utilities to reexamine their operations and approaches to decision making regarding fuel mix, generating capacity and energy market activity. One key determinant of day-to-day variance in electricity demand is the weather. Temperature, wind speeds, insolation, and humidity all play important roles in determining utility loads, primarily through their impact on demand for space conditioning in the residential and commercial sectors. Accurate short term (1-3 day) forecasts of weather when appropriately combined with load sensitivity data can improve management decisions, decrease operating costs, and improve reliability.
Traditional weather forecast products are often insufficient for forecasting electric utility loads. Specifically, while peak temperatures correlate moderately well with loads, derived parameters such as cooling degree days are typically more useful for load forecasting. Furthermore, the spatial distributions and short term historical trends of weather parameters are important in determining the response of utility customers. With this issue as a motivating factor we are currently developing a new load forecasting model to be used as a tool in estimating peak loads. This model is trained using appropriately defined customer-weighted mesoscale meteorological model output and utility load data. For development purposes we are using an enhanced version of the MM5 (v3.4) mesoscale model from the National Center for Atmospheric Research and recent load data at both the metropolitan and regional scale. The result is a forecasting system specifically tuned for the application of forecasting peak loads. This approach will be described in this poster and evaluated with respect to competing approaches.
Session 5, Energy Applications
Wednesday, 16 January 2002, 3:30 PM-5:30 PM
Next paper