Modeling spike occurrences in electricity spot prices for forecasting

Predicting the occurrence of extreme prices, so-called spikes, is one of the greatest challengeswhen modeling electricity spot prices. Despite the fact that recently new insights have beenachieved, the contemporaneous literature seems to be still at its beginning of understanding thedifferentmechanisms that drive spike probabilities. We therefore reconsider the problem offorecasting the occurrence of spikes, in the Australian electricity market. For this purpose, wefirst discuss properties of the price data with a focus on the occurrence of spikes. We thenpropose simple models for the probability of spikes which take these properties into account. Themodels compare favorably for in- and out-of-sample forecasts to a competing approach based on theautoregressive conditional hazard model.