A comparison study of co-simulation frameworks for multi-energy systems: the scalability problem

Abstract

The transition to a low-carbon society will completely change the structure of energy systems from a standalone hierarchical centralised vision to cooperative and distributed Multi-Energy Systems. The analysis of these complex systems requires the collaboration of researchers from different disciplines in the energy, ICT, social, economic, and political sectors. Combining such disparate disciplines into a single tool for modeling and analyzing such a complex environment as a Multi-Energy System requires tremendous effort. Researchers have overcome this effort by using co-simulation techniques that give the possibility of integrating existing domain-specific simulators in a single environment. Co-simulation frameworks, such as Mosaik and HELICS, have been developed to ease such integration. In this context, an additional challenge is the different temporal and spatial scales that are involved in the real world and that must be addressed during co-simulation. In particular, the huge number of heterogeneous actors populating the system makes it difficult to represent the system as a whole. In this paper, we propose a comparison of the scalability performance of two major co-simulation frameworks (i.e. HELICS and Mosaik) and a particular implementation of a well-known multi-agent systems library (i.e. AIOMAS). After describing a generic co-simulation framework infrastructure and its related challenges in managing a distributed co-simulation environment, the three selected frameworks are introduced and compared with each other to highlight their principal structure. Then, the scalability problem of co-simulation frameworks is introduced presenting four benchmark configurations to test their ability to scale in terms of a number of running instances. To carry out this comparison, a simplified multi-model energy scenario was used as a common testing environment. This work helps to understand which of the three frameworks and four configurations to select depending on the scenario to analyse. Experimental results show that a Multi-processing configuration of HELICS reaches the best performance in terms of KPIs defined to assess the scalability among the co-simulation frameworks.