Exploring the dynamics of the world energy system : an agent-based - system dynamics model

Due to limitation of some energy resources, there are concerns whether the world energy system (WES) can supply sufficient energy for societies in the future. Scientists develop models to gain insights into the system. GEMBA is a system dynamics model developed by Dale to explore the global energy supply using biophysical economics approach. Biophysical economics theory analyzes the economy based on the physical properties and structures of real economic systems and it considers natural resources and their impacts on the economic processes. GEMBA like other system dynamics models adopt a top-down view on the WES. The top-down view assumes that all elements of a system have global knowledge about the system. Many of such models do not capture some characteristics of a WES such as geographical resource distribution and demand diversity. In addition, it is often not possible to analyze the emergent effects from variations in low-level elements on the system behavior in top-down analysis. We developed an exploratory agent-based model, by taking a biophysical economics lens, for bottom-up analysis of a WES and relevant natural resources. We decomposed the world into a number of geographical regions to capture resource distribution and demand diversity in the WES. Our Multi-Region World Energy Model (MRWEM) combines the GEMBA with the concept of energy-return-on-investment (EROI) for imported energy. So, in MRWEM the internal behaviors of the world regions are modeled with GEMBA and the system dynamics approach while the inter-regions behaviors are modeled with agent-based modeling approach. MRWEM exhibits a number of advantages over GEMBA. First, it provides insights on the inter-regions energy movements and trade which is impossible in GEMBA. Second, MRWEM provides flexibility in analysis as changes in the model can be done at the level of regions not the whole world. Also, MRWEM facilitates analysts to analyze the WES using different geographical decompositions. Moreover, it shows that the hybrid adoption of agent-based modeling and system dynamics is possible and insightful when the level of abstraction is very high