This thesis introduces a methodology composed of a modelling approach and a computing support for the distributed simulation and the analysis of complex systems with large disaggregated data flows. The proposed methodology is influenced by both the system to simulate and the computing architecture used as a simulation support. This ensures a solution where the technical aspect fits the conceptual one. The modelling approach integrates a hierarchical description of the internal organization of a complex system, the design of a logical graph representing a structural decomposition of the system. A physical graph then establishes a binding between the conceptual representation level and the underlying distributed system. The Atlas platform is a support for discrete events time-driven distributed simulation whose dynamic properties replicate the behaviour of large disaggregated data flows. One of particularities of this platform resides in the concept of physical migration by group. This give a suitted solution to achieve and coordinate migration of large data flows, an important aspect for the simulation of societal and urban systems, particularly for transportation systems. The method and its platform is illustrated by an application in transportation that models and simulates people flows between the different halls of an airport terminal.
