Forecasting medium- and long-term peak electricity demand

25 June 2007

  • Where: International Symposium on Forecasting, New York

Abstract: Peak electricity demand forecasting is important in medium and long-term planning of electricity supply. Extreme demand often leads to supply failure with consequential business and social disruption. Forecasting extreme demand events is therefore an important problem in energy management.

Electricity demand at a given time is subject to a range of influences, including the ambient temperature, recent past temperatures, time of day, day of week, holidays, economic conditions and so on. I develop a semi-parametric model for half-hourly demand incorporating such weather, calendar and economic variables. The model is used to forecast upper-tail percentiles of electricity demand over a ten year horizon.

In order to forecast electricity demand using the model, we need to forecast all explanatory variables including temperature. To simulate future temperatures at half-hourly intervals, and thereby obtain the forecast distribution of temperature, a seasonal bootstrap method has been developed.

The method is demonstrated using half-hourly South Australian demand data from 1997-2006 with forecasts obtained for 2007-2016.

 Download pdf

« Stochastic population forecasts using functional data models | Graduation address »