arXiv subjects

Hannes M. H. Wolf

Publications and source records attributed to Hannes M. H. Wolf.

4 recordsLinked to original sources

arXiv2026

Graph Neural Ordinary Differential Equations for Power System Identification with Transfer Learning Applications

With the shift towards decentralized energy generation, the increasing complexity of power systems renders physics-based modeling challenging. At the same time the growing amount of available measurement data opens the door for obtaining data-driven models. A modern method to do so are neural ordinary differential equations (NODEs), offering a framework for continuous-time system identification. Recent extensions, so called graph NODEs impose a structural inductive bias that has the potential to improve generalization of the learned representations. In this work, we employ and extend them to develop message-passing graph NODEs (MPG-NODEs) for identification of coupled systems with heterogeneous node dynamics and heterogeneous edge couplings. This includes state-of-the-art machine learning architectures to infer latent representations of unmeasured states, local node and edge embeddings to account for heterogeneity as well as an autoregressive scheme to allow for piecewise constant control inputs. We apply MPG-NODEs to identify voltage and frequency dynamics of power systems and compare them to a monolithic NODE under identical measurement assumptions. Our case study indicates that while achieving similar performance, our proposed MPG-NODEs are more flexible with transfer learning options that allow to add or remove edges and nodes with little retraining.

eess.SY↗

arXiv2026

Data-Driven Active Power Flow Modeling: A Behavioral Systems Approach

The increasing decentralization of power systems driven by a large number of renewable energy sources poses challenges in power flow optimization: Partially unknown power line properties can render model-based approaches unsuitable. With the increasing deployment of sensors, data-driven methods rise as a promising alternative, offering flexibility to adapt changes and deal with unknown properties. In this paper, we propose a novel data-driven representation of nonlinear active power flow equations for radial grids based on Willems' Fundamental Lemma. Our approach allows for direct integration of input/output data into active power flow optimization, enabling cost minimization and constraint enforcement without requiring explicit knowledge of the electrical properties of the grid. Moreover, we derive a computationally tractable convex relaxation and show in a numerical case study that our approaches yield results that are identical to optimal active power flow formulations with known parameters.

eess.SY↗

arXiv2025

Augmented Neural Ordinary Differential Equations for Power System Identification

Due the complexity of modern power systems, modeling based on first-order principles becomes increasingly difficult. As an alternative, dynamical models for simulation and control design can be obtained by black-box identification techniques. One such technique for the identification of continuous-time systems are neural ordinary differential equations. For training and inference, they require initial values of system states, such as phase angles and frequencies. While frequencies can typically be measured, phase angle measurements are usually not available. To tackle this problem, we propose a novel structure based on augmented neural ordinary differential equations, learning latent phase angle representations on historic observations with temporal convolutional networks. Our approach combines state-of-the art deep learning techniques, avoiding the necessity of phase angle information for the power system identification. Results show, that our approach clearly outperforms simpler augmentation techniques.

eess.SY↗

arXiv2024

Identification of Power Systems with Droop-Controlled Units Using Neural Ordinary Differential Equations

In future power systems, the detailed structure and dynamics may not always be fully known. This is due to an increasing number of distributed energy resources, such as photovoltaic generators, battery storage systems, heat pumps and electric vehicles, as well as a shift towards active distribution grids. Obtaining physically-based models for simulation and control synthesis can therefore become challenging. Differential equations, where the right-hand side is represented by a neural network, i.e., neural ordinary differential equations (NODEs), have a great potential to serve as a data-driven black-box model to overcome this challenge. This paper explores their use in identifying the dynamics of droop-controlled grid-forming units based on inputs and state measurements. In numerical studies, various NODE structures used with different numerical solvers are trained and evaluated. Moreover, they are compared to the sparse identification of nonlinear dynamics (SINDy) method. The results demonstrate that even though SINDy yields more accurate models, NODEs achieve good prediction performance without prior knowledge about the system's nonlinearities which SINDy requires to work best.

eess.SY↗