Improving reproducibility in energy systems analysis: Introducing FAME - a framework for agent-based models

Abstract

In energy system modelling, agent-based models and simulations (ABMS) are clearly a minority. Yet, they capture the behaviour of heterogeneous actors in a changing landscape of energy technologies, policies and regulations. However, development of ABMS does not follow any standards and often happens in isolation. Root causes are a) the lack of an existing common ABMS framework, b) the lack of transparency of existing models and c) the lack of rigorously validated ABMS in energy systems analysis. To address the first point, we introduce a framework for distributed agent-based models of energy systems (FAME) that might serve as a blueprint and common starting point for further developments. A common framework might be one step towards establishing standards in the community in order to bring together isolated approaches with different methodologies. Our framework distinguishes between the framework (FAME) itself and the design of the application that is built upon it (AMIRIS). It is developed in Java, uses almost no external software dependencies and is general enough to be adaptable to many problems in energy systems analysis. It is planned to be available as open-source and has various advantages. One is speed, since FAME can take advantage of parallel computing architecture. A second one is its ease of use, since FAME has a highly configurable interface and good user documentation. Its main goal is – by using a common, already existing framework – to reduce development time considerably, enhance reproducibility and facilitate comparisons between models.