The modeling and forecasting of extreme events in electricity spot markets

Primary concerns for traders since the deregulation of electricity markets include both the selection of optimal trading limits and risk quantification. These concerns have come about as a consequence of the unique stylized attributes of electricity spot prices, such as the clustering of extremes, heavy tails and common spikes. We propose self-exciting marked point process models, which can be defined in terms of either durations or intensities, and which can capture these stylized facts. This approach consists of modeling the times between extreme events and the sizes of exceedances which surpass a high threshold. Empirical results for four major electricity spot markets in Australia show evidence of dependence between the occurrence times of extreme returns. This finding is directly related to the future behavior of the stochastic intensity process for price spikes. In addition, the proposed approach also provides more accurate one-day-ahead value at risk (VaR) forecasting in electricity markets than standard stochastic volatility models.