<p>This thesis considers the conventional SARIMA model and the EVT-GARCH model for forecasting electricity prices. However, we find that these models do not adequately capture the important characteristics of the electricity price data. A new model is developed, the EVT-SARIMA model, for forecasting electricity prices which is found to be the best at modelling the nature of the electricity prices. A time series of half-hourly electricity price data from the Hayward node in New Zealand is transformed into a daily average price series and using this resulting series, appropriate models are fitted for estimating and forecasting. The new EVT-SARIMA model is used to simulate 1000 time series of daily electricity prices, over a 90 day period, to consider strategies for managing the risk associated with price volatility. The effects of different financial instruments on the cumulative distribution functions of predicted revenue obtained using our model are considered. Results suggest that different contracts have different effects on the predicted revenue. However, all contracts have the effect of reducing variability in the predicted revenue values and thus, should be used by a risk manager to reduce the range of probable revenue values. The quantity traded and which contracts to use is dependent on the objectives of the risk manager.</p>
Statistical Analysis of New Zealand Electricity Prices: A Risk Manager's Perspective
