NCER Working Paper Series Forecasting Spikes in Electricity Prices

In many electricity markets, retailers purchase electricity at an unregulated spot price and sell to consumers at a heavily regulated price. Consequently the occurrence of extreme movements in the spot price represents a major source of risk to retailers and the accurate forecasting of these extreme events or price spikes is an important aspect of effective risk management. Traditional approaches to modeling electricity prices are aimed primarily at predicting the trajectory of spot prices. By contrast, this paper focuses exclusively on the prediction of spikes in electricity prices. The time series of price spikes is treated as a realization of a discrete-time point process and a nonlinear variant of the autoregressive conditional hazard (ACH) model is used to model this process. The model is estimated using half-hourly data from the Australian electricity market for the sample period 1 March 2001 to 30 June 2007. The estimated model is then used to provide one-step-ahead forecasts of the probability of an extreme event for every half hour for the forecast period, 1 July 2007 to 30 September 2007, chosen to correspond to the duration of a typical forward contract. The forecasting performance of the model is then evaluated against a benchmark that is consistent with the assumptions of commonly-used electricity pricing models.