A Cloud-native Implementation of the Simulation as a Service-Concept Based on FMI

Abstract

Providing modelling and simulation capabilities as a service promises to increase their value by improving accessibility for non-expert users and software agents as well as by leveraging cloud-computing technology to scale simulation performance beyond the capabilities of a single computer. In order to reach this potential, implementations must align their design with the architectural styles of cloud computing applications and the web in general. We present an open-source, cloud-native Simulation as a Service (SIMaaS)-implementation that gives access to models and allows simulating them on the web. The implementation uses Functional Mockup Units (FMUs) for co-simulation as an executable form of a model and relies on FMPy for simulation. It is realized as a microservice in the form of a REST-based HTTP-API. Functionality and performance are demonstrated by using the service to create ensemble forecasts for PV systems and to search for an optimal parameter set using a genetic algorithm. Conceptual limitations and the resulting opportunities for further work are summarized.