Master's thesis Llopis Mengual
Detection of the topology of complex energy systems using Supervised Learning from the field of Artificial IntelligenceCopyright: EBC
The energy turnaround is turning the German energy market upside down. The new challenges will make energy systems more decentralised and thus much more complex, so an analysis of the individual energy systems becomes thus a particular challenge. Sensors in energy systems provide a great amount of data, and evaluating this data is a very personnel-intensive task. It is very difficult for a human being to recognize the interrelationships between the systems, so algorithms based on Artificial Intelligence can help here.
Although plans of energy systems provide information about which influence could exist between plants, energy systems often lack a machine-readable form of these ones. However, characteristics of the individual sensors provide an indication of which other sensors and systems they are linked to. In this thesis, supervised learning approaches based on multivariate time series classification are used for the detection of the connections.
For this purpose, physical models of the energy system can be used to generate realistic data with Modelica and train Artificial Intelligence. The systems to be studied are a heat pump and a heat exchanger, with which this approach can also consider heat transfer aspects of the systems. Finally, real data from these energy systems is used to check whether the relationships are correctly recognized by the algorithm.