Electricity Spot Price Modelling: Benefits and Limitations of Selected Methodologies

This thesis investigates and assesses selected methods for stochastic electricity spot-price modelling. The focus is on day-ahead and intra-day electricity prices. The day-ahead and intra-day markets make up what is known as the spot market. To facilitate modelling, historical electricity pricing data is decomposed into distinct components: intra-weekly and intra-daily trends, longer-term trends (seasonal and annual) and sudden spikes or jumps occurring with periods of hours. These residual high frequency variations are assumed to be stochastic in nature and can be modelled assuming an Ornstein-Uhlenbeck process with parameters derived from the decomposed data. Alternatives to both the long and short period detrending approaches discussed in selected literature are proposed to address the shortcomings in these published methods and two implementations of Ornstein-Uhlenbeck are discussed. The value in this analysis lies in interrogating the decomposed, detrended, daily and hourly data in combination with other datatypes (for example, temperature, major sporting events, acts of nature) to try to understand why price patterns and anomalies are observed and subsequently use this relationship to predict spot prices with more confidence in the future.