Title: Modelling Australian electricity price using indicator saturation
Authors: James Reade - University of Reading (United Kingdom)
Shixuan Wang - University of Reading (United Kingdom) [presenting]
Abstract: Indicator saturation on electricity price series from the National Electricity Market (NEM) in Australia is employed in order to model the stylized facts of electricity prices, which include extreme spikes, seasonality, level-shifts, and autocorrelation. Standard modelling techniques in the literature to cope with these characteristics tend to focus on regime-switching models, which is constrained by the limited number of different, uncertain structural breaks, and the numerical estimation which could be possibly not converged. In particular, we develop an iterated procedure to capture outliers and shifts of differing magnitudes and economic importance. Based on a range of model evaluation tools, we find that indicator saturation method outperforms the regime-switching models with various settings. In additional to the statistical superiority, we further link our results of significant spikes, level-shifts, and trends to the policy changes in the NEM and provide recommendations for the development for the electricity markets in Australia.