Strategic behaviour analysis and modelling of electricity prices of Australian national electricity market

The Australian electricity market has a unique structure for determining wholesale prices. Market prices have been high, volatile, average prices have risen dramatically since market deregulation in 1997 to 2017-2018 and there has been significant concerns over future electricity supply security and investment. This dissertation provides three economic studies addressing challenging areas for development in the Australian National Electricity Market, namely price formation, generators' bidding behaviour and decomposing the price development in this market. This dissertation embodies a comprehensive dataset pertaining to each five-minute dispatch interval. Seven different data files were obtained from Australian Energy Market Operator. Advanced programming scripts have been developed to process the data, rectify the issues found, extract generators' specific attributes like fuel type and capacity, track the history of their bidding activity, identify the associated market information and to concatenate all of them as one dataset. The first paper examines an important issue in the Australian National Electricity Market (NEM): what are the quantitative implications of the bidding/rebidding rules in the current market design? We apply and extend existing literature on electricity markets with the objective of quantifying the market effects of generator behaviour. We utilise bid data in conjunction with price and quantity observations to determine whether there are consistent variations in equilibrium prices occurring year-round in the NEM as a result of rebidding. We achieve this by emulating a modified version of the market dispatch algorithm. This allows us to examine the dynamics of market outcomes from the time of initial offer to the time of dispatch. We analyse market outcome dynamics in conjunction with the corresponding number of rebids to elucidate any relationships that are present. The results are of policy importance given the flaws of the market design are found out and draws attention for immediate consideration. The second paper investigates how market signals in dispatch equilibria in the Australian National Electricity Market (NEM) impacts the electricity generating firms' bidding behaviour. We examine the information observed by generators at each five-minute auction to analyse how generators restate their initial and subsequent offers in response to such information. Utilizing the constructed high frequency dataset which consists of the intra-day supply bids of each generator, we illustrate that firms actively respond to the market signals in dispatch equilibria by rotating their supply curves within each trading interval. We disaggregate the effect of dispatch equilibria by the type of the generator. Further, we extend our specifications to account for difference in responses to change in prices arising from generators' bidding behaviour over dispatch intervals. The third paper extends existing theoretical frameworks describing electricity markets where each generator provides the Market Operator (MO) with a supply schedule in advance. The MO combines these with demand forecasts to produce equilibrium prices and instructs firms on their dispatch. We incorporate the possibility that generating firms may rebid (or revise) their supply schedule prior to dispatch - an important feature of markets in many countries which has not previously been included in theoretical models. We show that a dominant firm can gain substantially by manipulating its bids, and take advantage of the opportunity to submit rebids. In the Australian National Energy Market (NEM) where settlement prices are an average of six dispatch prices, it can, for example, withhold capacity at lower prices for the first bid in a period, creating a price hike, and then add capacity at lower prices to ensure dispatch. Using data from the Australian NEM we provide the first empirical evidence consistent with the hypothesized theoretical behaviour in the observed data