UB IlmenauIlmenau : Universitätsverlag Ilmenau, 2024. - XII, 172 Seiten.
(Ilmenauer Beiträge zur elektrischen Energiesystem-, Geräte- und Anlagentechnik - IBEGA ; 37)
DOI 10.22032/dbt.59235
URN urn:nbn:de:gbv:ilm1-2023000383
Zugl.: Dissertation, Technische Universität Ilmenau, 2023
With the increasing feed-in and distribution of renewable energies in the power system, the number of converters is growing. This can lead to the destabilization of the system due to several factors, one of which is the controller parameter of the converter. This requires parameterization and optimization of the mentioned parameters. The presented work is devoted to the main research question of how the stability of a converter-dominated distribution grid is improved for varying operating points by the dynamic adaptive control method. To carry out the parameterization of the controllers, the Reinforcement Learning (RL) agent for adaptively creating proper controller parameters according to the system states to improve and ensure stability is proposed. In this work, a digital twin (DT) of the studied network is first constructed with the help of parameter estimation. The DT model is utilized to generate training data for an artificial-neural-network-based state estimator, which is dedicated to arcuately and efficiently determining the system's stable state. More-over, a small signal stability indicator (SI) using the damping ratio of the dominant eigen-value for the DT model is developed. With the SI indicating to the RL agent the system stability margin during the RL training, the agent can ultimately output optimal controller parameters for the converters. Numerical case studies are used to verify the viability of the proposed approach that network stability can be improved by the proposed adaptive control method.
Zugriff auf den Volltext über die dbt:
https://nbn-resolving.org/urn:nbn:de:gbv:ilm1-2023000383