Generation of realistic scenarios for multi-agent simulation of electricity markets. (1st December 2016)

• Record Type: Journal Article
• Authors: Silva, Francisco; Teixeira, Brígida; Pinto, Tiago; Santos, Gabriel; Vale, Zita; Praça, Isabel
• Is Part Of: Energy. Volume 116:Part 1(2016)
• Page Start: 128
• Page End: 139
• Publication Date: 2016-12-01
• Subjects: Data-Mining -- Electricity markets -- Knowledge discovery -- Machine learning -- Multi-agent simulation -- Scenarios generation; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Energy consumption -- Periodicals; 333.7905
• DOI: 10.1016/j.energy.2016.09.096
• Languages: English
• ISSNs: 0360-5442

Abstract

Abstract: Most market operators provide daily data on several market processes, including the results of all market transactions. The use of such data by electricity market simulators is essential for simulations quality, enabling the modelling of market behaviour in a much more realistic and efficient way. RealScen (Realistic Scenarios Generator) is a tool that creates realistic scenarios according to the purpose of the simulation: representing reality as it is, or on a smaller scale but still as representative as possible. This paper presents a novel methodology that enables RealScen to collect real electricity markets information and using it to represent market participants, as well as modelling their characteristics and behaviours. This is done using data analysis combined with artificial intelligence. This paper analyses the way players' characteristics are modelled, particularly in their representation in a smaller scale, simplifying the simulation while maintaining the quality of results. A study is also conducted, comparing real electricity market values with the market results achieved using the generated scenarios. The conducted study shows that the scenarios can fully represent the reality, or approximate it through a reduced number of representative software agents. As a result, the proposed methodology enables RealScen to represent markets behaviour, allowing the study and understanding of the interactions between market entities, and the study of new markets by assuring the realism of simulations. Highlights: Automatic generation of electricity market simulation scenarios. Knowledge discovery and analysis of real electricity markets data. Innovative clustering approach to group market players according to their similarity. Representation of market players through software agents. Generated scenarios provide a reliable representation of reality even when using a reduced number of agents.