Conrad T (2024)
Publication Language: English
Publication Type: Other publication type
Subtype: Special issue of a journal
Publication year: 2024
Edited Volumes: HIGHVOLT KOLLOQUIUM - ZWISCHENRUF ʼ24
Open Access Link: https://app.highvolt.de/
The objective of this thesis was to implement an AI system to analyse the static voltage
stability of power grids and assess its ecacy in achieving faster computation times com-
pared to existing approaches.
To carry out this implementation, a training dataset of randomly generated cases for
a 5 bus grid was created using the load ow calculation and dV ◁dQ sensitivity analysis
in the PowerFactory program. Various articial neural networks were optimised using
the pruning function in the DataEngine programme and implemented using the Python
library Keras. The evaluation focused on applicability, the impact of reducing features
and cleaning the dataset, as well as examining variations in power grid size.
The thesis indicated that, despite obtaining high-accuracy results during training, the
models lacked sucient generalisation ability. Expanding the dataset did not enhance
this ability and the model necessitated more time for both data production and training
when the power grid was expanded.
Regardless of the utilization of diverse approaches, including MLPs and Kohonen maps,
adequate generalisation was not achieved. Future research should concentrate on ex-
panding the features, testing multiple articial neural networks and integrating load ow
calculations with AI. It is suggested that the methodology and approach to data cleaning
should be re-evaluated and further research should explore contemporary optimisers or
additional articial neural network models.
APA:
Conrad, T. (2024). AI-based static voltage stability analysis of power grids.
MLA:
Conrad, Timon. AI-based static voltage stability analysis of power grids. 2024.
BibTeX: Download