8 results
Now showing1 - 8 of 8
- Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Quantifying the impact of uncertainties in weather and mobility demand forecasts on optimized household scheduling and grid loadingThe increasing electrification of heating and mobility, as well as distributed generation introduce significant uncertainty into distribution grids, complicating demand predictions and grid operation. This paper presents a Monte Carlo-based framework to quantify the impact of correlated weather and behavioural uncertainties on optimized household scheduling and grid loading. Forecast errors of temperature, solar irradiation, inflexible load, and mobility demand are modelled and propagated through a home energy management system and probabilistic power flow analysis. Results show that uncertainty in household power and grid loading is mainly driven by photovoltaic generation and electric vehicle charging, while temperature primarily affects daily energy. Dynamic tariffs reduce overall uncertainty, whereas flat tariffs lead to pronounced peaks. Despite high household-level variability, aggregation limits uncertainty at the grid level, and severe congestion events remain predictable.Publicationtype: Conference Paper12 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Managing uncertainty by leveraging flexibility in smart energy systems: AI-supported distributionally robust chance-constrained optimization(IEEE, 2025-06-20); ; ; The rapid decarbonization of the energy systems to meet climate targets is driving increased electrification across various energy sectors, with the electrification of the household heating sector posing significant challenges for electrical networks in many parts of Europe. Electric heat pumps (UP), paired with thermal storage (TS), have cemented themselves as the key technology in this transition, offering high efficiency and signif-icant flexibility for load shifting and energy storage. However, this flexibility is subject to uncertainty due to variable weather conditions and user behavior. In this paper, an AI -enhanced framework for optimizing the operation of a UP-dominated res-idential network under uncertainty is presented. The framework utilizes a Bayesian neural network to generate forecasts for UP demand based on real measured data, creating a data-driven, AI-enhanced ambiguity set in the distributionally robust chance-constrained (DRCC) optimization. This approach enhances the confidence level in the proposed dispatch plan by addressing forecast uncertainty in a statistically robust manner. By effectively leveraging the thermal system's flexibility, it mitigates the risk of constraint violations and ensures reliable grid operation. Val-idation on a residential network demonstrates that the proposed framework achieves higher robustness and reliability compared to traditional deterministic models, highlighting its potential to support the energy transition.Publicationtype: Conference Paper54 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Evaluating non-discrimination obligation in implementing congestion management measures in electrical distribution systems(IEEE, 2025-06); ; ; The increasing electrification of transport and heating sectors, alongside the rise of distributed renewable energy generation, is causing significant changes in electrical distribution grids, leading to more frequent congestion. To address this, congestion management strategies are essential to ensure grid stability while complying with the regulatory obligations of non-discrimination. However, the lack of clear definitions and standardized methodologies for evaluating non-discrimination in designing and implementing congestion management remains a critical gap. In this paper, a comprehensive framework is proposed to define and evaluate non-discrimination principles in congestion management measures. A novel indicator-based methodology is introduced that considers the customer's individual characteristics together with the curtailments to quantify systematic inequalities in the treatment of customers. The proposed methodology is validated with a case study utilizing two congestion management approaches, where flexible appliances are curtailed proportionally or with equal shares in case of congestion. Results demonstrate the framework's ability to identify discriminatory practices, providing valuable insights for fulfilling the requirement of non-discrimination in congestion management strategies in future grid operations.Publicationtype: Conference Paper51 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Coordination of Demand-Side Flexibility Resources for Preventive Congestion Management in Distribution Systems(VDE Verlag, 2025-05); ; ; A coordination function (CoF) is proposed as a measure of preventive congestion management in the distribution grid. The CoF coordinates flexible loads of customers by using information on their planned electrical demands given by their home energy management systems. Based on the aggregated demand data, the predicted grid condition is analyzed. If congestion is detected it is resolved by making adjustments to the flexible loads from their original plans utilizing the opportunities provided by the CoF. For the energy rescheduling two different methods, i.e., a quota-based approach and an equal-share approach, are evaluated. The results show that, by adjusting flexible loads, the CoF can effectively resolve anticipated congestion while creating only minimal curtailments for the customers. However, this will come with significant costs of up to 10% for the customers, so there is a need for compensation for participating in a flexibility coordination. The comparison of both methods shows that a hybrid approach should be pursued in further research.Publicationtype: Conference Paper91 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , StableRLS: stable reinforcement learning for simulink(2024-02-01); ; Today's innovative systems rely heavily on complex electronic circuitry and control. This complexity increases development, testing, and manufacturing efforts. Consequently, building the first hardware prototype is often uneconomical, leading to a preference for simulation-based approaches. MATLAB, a commercial software, facilitates rapid prototyping, mathematical analysis, and simulation through its Simulink toolbox. Additionally, MATLAB offers a reinforcement learning (RL) toolbox for integration with Simulink. However, this RL toolbox is still maturing and lacks many recent RL innovations, partly because MATLAB is not alywas the first choice for machine learning in academia and industry. Despite this, many companies and researchers leverage Simulink's powerful simulation capabilities, often developing custom simulation libraries. To bridge the gap between MATLAB/Simulink and the more widely used Python ecosystem for RL, StableRLS was developed. StableRLS enables the use of existing Simulink models for RL in Python by converting these models and offering a Python user interface compatible with the Gymnasium library.Publicationtype: Research Report82 39 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Day-Ahead-Engpassprognose für elektrische Verteilnetze mittels probabilistischer Lastflussrechnung(Technische Universität Graz, 2024); ; Die Transformation zur CO2-Neutralität verursacht grundsätzliche Veränderungen unseres Energiesystems. Auf der einen Seite wird Energie zunehmend dezentral aus er- neuerbaren und damit volatilen Quellen bereitgestellt, auf der anderen Seite steigen die Lasten durch Elektrifizierung und Sektorenkopplung. Diese Entwicklungen werden zu einer Zunahme von Netzengpässen im Verteilnetz führen. Eine deterministische Prognose dieser Netzengpässe ist aufgrund der Unsicherheiten bei der Vorhersage der erneuerbaren Erzeugung und des Lastverhaltens schwierig. Eine Berücksichtigung der Unsicherheiten bei der Engpassprognose bietet dagegen die Möglichkeit, neben der erwarteten Netzauslastung auch die Zuverlässigkeit der Prognose zu quantifizieren. Dazu wird eine Methodik entwickelt, die mittels einer probabilistischen Prognose die Betriebsmittelauslastung inklusive einer möglichen Spannbreite für den Folgetag vorhersagt. Dies geschieht mittels einer probabilistischen Lastflussrechnung nach der Monte-Carlo-Methode. Die notwendigen Eingangsverteilungen für Last und Erzeugung werden dabei auf Basis von historischen Lastdaten sowie numerischer Wetterprognosen erstellt. Es wird für ein ländliches und ein städtisches Mittelspannungsnetz nachgewiesen, dass mit Hilfe des Verfahrens die Betriebsmittelauslastung gut vorhergesagt werden kann. Allerdings bilden sich die hohen Unsicherheiten der Wettervorhersage auch in der Auslastungsprognose ab. Zusätzlich wird eine übersichtliche Darstellungsmethode zur Engpassauswertung vorgeschlagen.Publicationtype: Conference Paper93 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , KoLa – Koordinierungsfunktion des Verteilnetzes und Lastmanagement für den elektrifizierten Personenverkehr(2023-12)40 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , KI-basiertes Energiemanagement für Schiffe mit Gleichspannungs-Energienetzen(2023-03-16); ; Die Veränderung des Klimas verdeutlicht u. a. die globale Relevanz des emissionsarmen Betriebs maritimer Systeme. Für diesen bedarf es einer nachhaltigen Energieversorgung an Bord, die durch die Integration erneuerbarer Energiequellen erreichbar ist. Dadurch entstehen zusätzliche Freiheitsgrade für das Energiemanagement eines Schiffs gegenüber dem Stand der Technik, bei dem alle Anlagen über große zentrale Generatoren versorgt werden. Da an Bord großenteils Gleichstrom (DC) verwendet wird, wird eine DC-Energieversorgung betrachtet, welche gegenüber Wechselstrom (AC)-Netzen Umwandlungsverluste reduziert. In der Studie wird mit Reinforcement Learning als Methodik für eine mehrdimensionale Optimierung untersucht, welche zusätzlichen Freiheitsgrade im Energienetz ausgeschöpft werden können. Dazu werden ein exemplarisches Kabinennetz eines Kreuzfahrtschiffes realitätsgetreu in MATLAB (Software) modelliert und eine optimierte Regelung für das Energiemanagement simuliert. Dabei werden zwei Regelziele verfolgt: zum einen die Optimierung der Spannungsniveaus der Quellen (Batteriespeicher, Photovoltaik, Netzeinspeisung) in Form einer Droop-Regelung, zum anderen die Optimierung der mit dem Netz ausgetauschten Leistungen der genannten Quellen. Durch die Anpassung der Spannungsniveaus lässt sich die Aufteilung der Lasten zwischen den Quellen variieren, eine Überlastung einzelner Quellen verhindern und die Lebensdauer der Komponenten im Netz maximieren. Die Studie leistet einen Beitrag bei der Ermittlung des Potenzials von KI-Anwendungen für das Energiemanagement.Publicationtype: Conference Poster not in Proceedings62 57