Humboldt-Universität zu Berlin - Mathematisch-Naturwissenschaftliche Fakultät - Institut für Informatik

Die Verteidigung findet digital per Zoom statt. Eine Zoom-Einladung finden Sie hier. (nur mit Informatik-Account)

Introduction:
Energy modeling is the creation of a comprehensive model about an energy system, detailing and calculating various metrics including financial costs, operating hours, greenhouse gas emissions, and more. It can be used to assure compliance with regulations and requirements, optimizing operation modes, and finding ways of minimizing waste. It is often an essential and extensive part of complex engineering projects.
Despite its apparent usefulness, it is a very computationally demanding and small changes can ripple recursively throughout the model, often overwhelming the capabilities of personal workstations. Compute power is limited to the present machine. Its computation time often limited to office hours. Power outages and unforeseen shutdowns might interrupt a calculation, as personal computers are rarely equipped with an uninterruptible power supply. Office computers are hardly designed to run compute heavy tasks. These machines are often set up to be just powerful enough to run an MS Office Suite. As shown in [DS22, page 3] some energy models are too complicated to reach usable results in even a 24-hour period. A possible mitigation is transferring the computational work to a server that has sufficient compute power to run the models and then inform the user about their completion.
This work takes a look at the requirements and problems arising from moving an open source energy modeling software by the name of open energy modelling framework (oemof) into a scalable, multi-user cloud environment. For this goal, a communication application programming interface (API) with the software was written. The various problems that arose in the process will be highlighted and their solutions showcased.