An agent-based simulation study of competition in the Victorian gas market

Agent-based simulation is used to model the deregulated Victorian gas market. The model characterises the full range of market participants: producers, transmission operators, underground storage operator, wholesale traders, market manager (regulator) of the daily balancing and gas supply price setting system, distributors, retailers, and customers of various types and sizes in numerous regions. The model is domain specific and characterises participants' behaviour in their market trading and business terms. Participants have resources and capacities, set retail prices, hold and negotiate supply contracts, decide if and when to increase capacities, determine and make daily wholesale price and volume offers, and set retail prices to customers, who have the capacity to swap retailers. All decisions are made in terms of market regulations, contract terms, seasonal patterns, market shares, prices, costs and profits. The model provides a data rich platform that allows market agents to consider the combined interactions of both micro- and macro-decisions on their overall performance in multiple sectors. The model has few control agents and relies only on the mild synchronous control of agent interactions to achieve dynamic changes in the market over short and long periods. The model relies on the diversity of agents' interactions and simple mental models to make micro-level business decisions rather than on using learning, gaming or control techniques. The agents' myriad of interactions generates emergent market and industry trends as well as outlooks for individual market agent's profit performance. The model provides a simulation framework that a market participant can use to gain strategic insights to future market competition and how to be successful in a competitive market. Participants can use the model to assess the strengths and weaknesses of their own competitive positions in the market and to consider alternative strategies by use of a look-ahead capability for testing a range of strategies to guide dynamic implementation of varying strategies and plans. The model is classified as a 'proof of concept' model and one of very few that has attempted and succeeded in modelling a multi-sector, complex, adaptive commodity market with micro-level characterisation. The model is broadly calibrated within the limits of available data and, while outputs have not been extensively or statistically validated, retrospective model predictions compare well to actual history and a number of longer term model predictions provide rationally bounded outlooks of key industry criteria such as supply prices and demand. The model platform has been used by the CSIRO to build their similarly structured NEMSIM (electricity) model.