Angebote für Abschlussarbeiten

Sie können Ihre Bachelor- und/oder Masterarbeit bei uns in den Bereichen Regelungs- und Netzleittechnik anfertigen. Wir freuen uns über Ihr Interesse und bieten Ihnen folgende aktuelle Themen an:

Versuchsstand: Kaskadenregelung (B.Sc.)
InhalteVerantw.
Für das Labor Regelungstechnik ist ein Beispielversuch zur Analyse und Synthese einer Kaskadenregelung zu entwickeln.

Zunächst ist zu praktischen Anwendungen, dem typischen Verhalten und den Vorteilen von kaskadierten Regelkreisen zu recherchieren. Anhand der Ergebnisse ist aus den im Labor vorhandenen Modell- und realen Regelstrecken ein geeigneter Prozess zusammenzustellen bzw. zu wählen.

Für diesen Prozess soll eine Kaskadenregelung entworfen und getestet werden. Dazu sind geeignete Verfahren zur Prozessanalyse und Modellierung anzuwenden. Für Simulationszwecke steht eine Matlab/Simulink-Umgebung zur Verfügung. Die Ergebnisse sind am realen Versuchsstand zu überprüfen.

Zur Regelung sind notwendige Reglertypen auszuwählen und mit einem geeigneten Verfahren zu parametrieren. Vor dem Test am realen Versuchsstand ist das Regelkreisverhalten durch eine Simulation zu überprüfen. Als Regler können sowohl Software-, Modell- als auch reale industrielle Regler verwendet werden.

Nach erfolgreichem Test sind die einzelnen Schritte und Ergebnisse nachvollziehbar zu dokumentieren. Hinweise zur Einbindung in die Lehre und Vorschläge für Variationen der Versuchsdurchführung sollen gegeben werden.

Angebot: Arbeitsplatz und Betreuung am Fachgebiet, Hard- und Software bereits vorhanden, Arbeitsaufwand kann über ein Semester gestreckt werden.
Dr.-Ing. Uwe Rau
Versuchsstand: Regelung technischer Strecken (Durchfluss, Füllstand, Luftstrom) (B.Sc.)
InhalteVerantw.
Für das Labor Regelungstechnik ist ein Beispielversuch zur Analyse und Synthese der Regelung einer realen technischen Regelstrecke zu entwickeln.

Zunächst ist zu praktischen Anwendungen und dem typischen Verhalten der Regelstrecke zu recherchieren. Anhand der Ergebnisse ist aus den im Labor vorhandenen technischen Möglichkeiten ein geeigneter Prozess zusammenzustellen.

Für diesen Prozess soll eine Regelung entworfen und getestet werden. Dazu sind geeignete Verfahren zur Prozessanalyse und Modellierung anzuwenden. Für Simulationszwecke steht eine Matlab/Simulink-Umgebung zur Verfügung. Die Ergebnisse sind am realen Versuchsstand zu überprüfen.

Zur Regelung sind notwendige Reglertypen auszuwählen und mit einem geeigneten Verfahren zu parametrieren. Vor dem Test am realen Versuchsstand ist das Regelkreisverhalten durch eine Simulation zu überprüfen. Als Regler können sowohl Software-, Modell- als auch industrielle Regler verwendet werden.

Nach erfolgreichem Test sind die einzelnen Schritte und Ergebnisse nachvollziehbar zu dokumentieren. Hinweise zur Einbindung in die Lehre und Vorschläge für Variationen der Versuchsdurchführung sollen gegeben werden.

Angebot: Arbeitsplatz und Betreuung am Fachgebiet, Hard- und Software bereits vorhanden,
Arbeitsaufwand kann über ein Semester gestreckt werden.
Dr.-Ing. Uwe Rau
Deep Reinforcement Learning in Active Distribution Grids (M.Sc.)
ContentResp.
This master thesis will be conducted in close cooperation with the Berlin-based Start-up Qantic (https://www.qantic.com/) and can be written either in English or German.

Motivation and background
In Germany, distribution grids play an important role in the integration of renewable energies. Originally designed as passive systems, they are developing into active energy grids in which not only many decentralized generation plants are connected, but also sector coupling, and other flexibility options will be implemented in the future. As a result, grid operation is becoming increasingly complex. In addition, there are often only a few measuring devices and sensors installed in low and medium-voltage grids that allow visibility of the condition of the current grid state. To be able to guarantee safe grid operation, in addition to the classic network expansion, a limitation of the feed-in power or a purchase of reactive power, among other things, are available. By means of a so-called Q/U control (Volt-Var-Control), node voltages can be influenced, e.g. in order to minimize network losses and to avoid voltage profile violations. The application of artificial neural networks (e.g. Deep Reinforcement Learning) to Q/U control is currently the subject of scientific work - in these publications the state of the network or its change of state as a result of actions (e.g. switching actions) is conceived as a sequential decision process (Markov Decision Process). The network state (e.g. the node voltages) can be changed here by actions that are clearly defined in advance (e.g., specification of a power factor). Furthermore, a metric is introduced to evaluate the quality of a respective action and the resulting network state (as a reward function). An algorithm now learns by interacting with the system to choose the actions that maximize the reward in the end. Previous work on the use of reinforcement learning in active distribution networks generally does not consider the technical and regulatory characteristics of German regulatory framework. Currently, the possibilities for active control of distribution grid operators in Germany still seem limited, as the installed resources are only capable of performing distributed Q-U control to a limited extent. On the other hand, German distribution system operators have specific instruments at their disposal to ensure grid stability (e.g. feed-in management and redispatch 2.0), which have national specificities. Accordingly, the research question of the final thesis could be: How can existing approaches in reinforcement learning be adapted to a distributed volt-var control in active distribution grids, which at the same time also comply with the German legislation.

Scope and working packages
A reinforcement learning algorithm is to be developed and implemented in the Python programming environment. Here, existing open-source implementations as well as models of Qantic GmbH can be built upon - basically, the following work steps are recommended:
  1. Familiarization with the current state of the art of volt-var control in distribution networks as well as a review of selected scientific publications.
  2. Development of a simulation of a distribution network using "Pandapower" to determine the nodal voltages; if necessary, existing topologies can be used here (https://pandapower.readthedocs.io/en/develop/networks.html).
  3. Explicit mapping of the possibilities to maintain a smooth voltage profile in German distribution grids (including specific measures such as feed-in management and redispatch).
  4. Transfer of the grid model into a simulation environment usable for reinforcement learning as a Markov Decision Process with defined state and action space as well as a reward function to evaluate voltage stability.
  5. Optimization of voltage stability with an existing open-source environment of a selected reinforcement learning algorithm (e.g. https://github.com/nikhilbarhate99/PPO-PyTorch,https://github.com/ikostrikov/pytorch-a2c-ppo-acktr-gail,https://blog.varunajayasiri.com/ml/ppo_pytorch.html).
Requirements and skillset
  1. Strong interest in machine learning topics and their application to power systems
  2. First experience in the programming language Python
  3. Basic knowledge of the structure and operation of electrical networks and their control concepts
References
[1] Gao, Y.; Yu, N.: Deep Reinforcement Learning in Power Distribution Systems: Overview, Challenge, and Opportunities, 2021, IEEE Power & Energy Society General Meeting
[2] Huang, B.; Wang, J.: Deep-Reinforcement-Learning-Based Capacity Scheduling for PV-Battery Storage System, 2021, IEEE Power & Energy Society General Meeting
[3] Gao, Y.: Safe and sample-efficient reinforcement learning for power distribution system controls, 2021, IEEE Power & Energy Society General Meeting
[4] Liu, H.; Wenchuan, W.: Robust Adversarial Reinforcement Learning for Inverter-based Volt-VAR Control in Active Distribution Networks, 2021, IEEE Power & Energy Society General Meeting
M.Sc. Nicolai Lorenz-Meyer
Observer design for a DC converter connected to a microgrid (M.Sc.)
ContentDesp.
Based on the assumption that not all the states of a DC power converter connected to a microgrid can be measured, it becomes necessary to design a state observer. The observer’s goal is to estimate the system’s inaccessible states based on the system dynamics and output. This estimation is then utilized in the feedback control law for achieving voltage regulation.M.Sc. Ismael Jaramillo-Cajica
Simulink model of a photovoltaic plant and energy storage system DC microgrid with grid-aware power management strategy (B.Sc.)
ContentDesp.
Due to environmental and political reasons the use of renewable energy sources (RESs) such as photovoltaic (PV) plants and wind turbines has considerably increased in recent years. This type of RES are interfaced with active distribution networks (ADNs) via power electronic converters, which are able to operate under two modes: voltage regulation mode (VRM) or current regulation mode (CRM). One of the inherent properties of RESs is their dependency on the unpredictable weather conditions, which cause fluctuations on the generated power. For this reason, RESs are usually installed together with some sort of energy storage system (ESS) whose main task is to compensate the fluctuations of the RES. In such cases, a power mangement strategy aims at injecting the desired power into the ADN while coordinating the power flows between RES and ESS such that the state of charge (SoC) of the latter remains within allowed limits for avoiding power imbalances or damage of the equipment. In this regard, an alternative found in the literature relies on imposing  a grid-aware operation of the RES, i.e., switch among VRM and CRM of the RES in dependency of the status of the ESS and AND. Recent studies have proposed mathematical models for emulating the fluctuating power generation of PV plans such different types of weather conditions that can be successfully recreated in a realistic fashion. Based on this, this projects considers a DC microgrid (MG) composed of a PV plant, an ESS and a local load that is connected to an infinte bus representing an ADN. The goal of the project is to build the Simulink model of the considered MG such that: i) the power injection of the PV plant follows a time series generated by a suitable mathematical model; ii) the SoC of the ESS depends on the power flows from/to the RES and ADN; iii) the RES power converter model is compatible with a grid-aware operation for implementing a desired power management strategy.M.Sc. Ismael Jaramillo-Cajica
Matlab/Simulink toolbox for RMS simulations of power systems (B.Sc.)
ContentDesp.
The current changes in power networks induced by the energy transition and the introduction of power electronic technologies had made necessary the study of such systems in a dynamical manner. Contrary to standard steady state and power flow analysis, now it is required to simulate the transient behavior of the networks with a particular interest on the response of the different elements’ controllers. One of the limitations in dynamic simulation of power systems is the lack of software tools specialized in both, power systems and automatic control. This is the case of PowerFactory, where the standard tools for power system analysis and simulation are available, but the incorporation of sophisticated control algorithms is rather difficult and obscure. On the other hand, one has Matlab/Simulink, designed to simulate complex control systems, but which lacks tools for simulation of large power networks.The purpose of the thesis is to dote Matlab/Simulink of elements for simulating transmission and distribution networks. These networks are usually simulated by neglecting the transient response of transmission lines and passive elements such as transformers, and only considering dynamical behavior in a time scale of tens of milliseconds or above. The process is usually referred as RMS-simulations. For that, the main tool would be Simscape, where the idea is to implement algebraic models of balanced and unbalanced three-phase impedances, admittances, transformers, ZIP loads, and compatible dynamical models of synchronous generators, induction motors and composite loads.Dr. Juan G. Rueda-Escobedo
Modeling and control of a bidirectional DC-DC converter for battery management (B.Sc.)
ContentDesp.
Nowadays we can find Li-ion batteries not only in a great variety of electronic devices, but also in vehicles, and soon in large scale power systems. This makes the understanding of the devices in charge of handling the charge and discharge of the batteries increasingly relevant. The goal of this work is to investigate the available technologies in bidirectional DC-DC converters for battery managements, and their implementation for dynamical simulation in Matlab/Simulink together with the associated controllers.Dr. Juan G. Rueda-Escobedo
Analysis of the algorithm to control a HV transformer tap changer in an utility scale PV plant (B.Sc. / M.Sc.)
ContentDesp.
The monitoring data of a transformer tap changer, which is controlled by the power plant controller of a PV plant, shall be analyzed in Matlab in order to identify and estimate potential revenue losses. The analysis can be used to implement an improved control algorithm for the tap changer, which is then simulated and validated.Dipl.-Ing. Simo Kauth

Die Arbeiten können sowohl auf Deutsch als auch Englisch angefertigt werden.

Für weitere Auskünfte wenden Sie sich bitte an die genannten Ansprechpartner. Falls Sie eigene Themenvorschläge haben sollten, diskutieren wir diese gerne mit Ihnen.

Unsere Webseite verwendet Cookies. Diese haben zwei Funktionen: Zum einen sind sie erforderlich für die grundlegende Funktionalität unserer Website. Zum anderen können wir mit Hilfe der Cookies unsere Inhalte für Sie immer weiter verbessern. Hierzu werden pseudonymisierte Daten von Website-Besuchern gesammelt und ausgewertet. Das Einverständnis in die Verwendung der technisch nicht notwendigen Cookies können Sie jeder Zeit wiederrufen. Weitere Informationen erhalten Sie auf unseren Seiten zum Datenschutz.

Erforderlich

Diese Cookies werden für eine reibungslose Funktion unserer Website benötigt.

Statistik

Für den Zweck der Statistik betreiben wir die Plattform Matomo, auf der mittels pseudonymisierter Daten von Websitenutzern der Nutzerfluss analysiert und beurteilt werden kann. Dies gibt uns die Möglichkeit Websiteinhalte zu optimieren.

Name Zweck Ablauf Typ Anbieter
_pk_id Wird verwendet, um ein paar Details über den Benutzer wie die eindeutige Besucher-ID zu speichern. 13 Monate HTML Matomo
_pk_ref Wird benutzt, um die Informationen der Herkunftswebsite des Benutzers zu speichern. 6 Monate HTML Matomo
_pk_ses Kurzzeitiges Cookie, um vorübergehende Daten des Besuchs zu speichern. 30 Minuten HTML Matomo