12. November 2026

Uncategorized

Machine Learning for Complex System (CoSy ML)

Name: DPUI Colloq
Start: 2026-11-06T15:00:00+01:00
End: 2026-11-06T16:00:00+01:00
Location: Zoom

Formerly “Data Processing for Utility Infrastructure” (DPUI)

We develop machine learning for complex systems – physical and infrastructure systems, biomedical and population-level systems, and robotic systems.

Our work spans modelling, simulation and acceleration, prediction, optimisation and control, knowledge extraction and reasoning, and agentic AI for decision support.

We organise our research along three core methods and three application domains. The matrix below shows where they meet – every row and column leads to a page with projects, publications, open data, code, demos and the people involved.

Who we are

CoSy ML brings together researchers with backgrounds in computer science, electrical engineering, medical engineering and fluid mechanics. Each card links to the person’s PRL profile.

Group head

Dr.-Ing. Siming Bayer
Group head
siming.bayer@fau.de

Researchers

Adithya Ramachandran
Doctoral researcher

Satyaki Chatterjee
Doctoral researcher · with Diehl Metering GmbH

Julian Oelhaf
Doctoral researcher

Changhun Kim
Doctoral researcher

Karan Pahlajani
Doctoral researcher

Mohammad Moataz Tolba
Doctoral researcher · with Diehl Metering GmbH

Prabhu Nithin Gollapudi
Doctoral researcher

Prajol Shrestha
Research associate RKRedwanul Karim
Researcher LLyuhe
Visiting researcher · Sichuan University ZZengyu
Visiting researcher · Sichuan University

Research: methods × domains

Rows are our core methods, columns our application domains. Click a heading to open its page – or Details in a cell to see who works where both meet.

Core methods

Constrained & Trustworthy ML

ML for Multimodal Data Fusion & Analysis

Model Predictive Control & Reinforcement Learning

Application domains

Critical Infrastructure

Medical

Robotics

Highlights

29 October 2026 – Poster at GridFM 2026 in Porto. “Physics-Informed but Not Physics-Consistent: Error Geometry and Subspace Projection for Neural AC Power Flow” at the 6th Workshop on Foundation Models of the Electric Grid.
19 October 2026 – Three papers at IEEE PES ISGT Europe 2026, Budapest. The PROTECT-90 fault dataset, fault-inception detection in real-world disturbance data, and peak-aware short-term load forecasting.
4 October 2026 – New name: CoSy ML – Machine Learning for Complex Systems. The group formerly known as Data Processing for Utility Infrastructure continues as CoSy ML, reflecting its growing range of methods and application domains.
21 July 2026 – Panel talk at the IEEE PES General Meeting 2026. “GridSense: Ask Your Power Grid” – LLM-based power-grid analysis grounded in a grid knowledge graph, in the panel Power Grid Artificial Intelligence – European Solutions.
May 2026 – Physics-informed GNN for AC power flow at ICASSP 2026, Barcelona. Edge-aware attention and a line-search correction operator make a graph neural power-flow solver more physically consistent at inference time.
12 February 2026 – UtilityTwin recognised at the BMFTR StiPS ceremony in Berlin. Adithya Ramachandran and Siming Bayer represented the lab at the award ceremony.
24 November 2025 – UtilityTwin wins 2nd prize in the Student Prize for Social Innovations (StiPS). Nationwide prize of the Federal Ministry of Research, Technology and Space (BMFTR).

Projects & funding

Hyb_Mod_NetRunningResearch and development of hybrid network modelsSep 2025 – Aug 2028Graph databases for district heating networks that integrate topology and diverse sensor data, and hybrid network models that combine differential-equation-based and AI-driven methods for optimal control.Trustworthy ML Multimodal data Critical infrastructure Siming Bayer

Siming

Full description (CRIS)

CRIS →

GridAssistRunningAI-based congestion management and grid restorationTeilvorhaben: KI-gestütztes Engpassmanagement und NetzwiederaufbauDec 2024 – Nov 2027 · BMWE, “OptiNetD” initiativeAI assistance for transmission and distribution grid operation: physics-informed neural power flow, automated congestion and fault management, grid restoration, and a graph-based knowledge database for heterogeneous grid data – validated in a hybrid field test with Lechwerke (LEW).Trustworthy ML MPC & RL Critical infrastructure Changhun Kim

Changhun

Full description (CRIS)

TP3: Automated Optimal System Management in MS/HS Networks & Network Restoration

AP 3.1: Automated Bottleneck and Fault Management Based on Machine Learning Methods

AP 3.2: Stable Grid and Supply Restoration

TP5: Competence Team for Interfaces and Databases

AP 5.1: Creation of a Knowledge Database for Heterogeneous Power Grid Data Sources

Automated Grid Management and Restoration through AI and Data Integration

The increasing complexity and volatility of modern power grids necessitate intelligent, automated systems for optimal operation and rapid restoration. In TP3, we develop AI-driven solutions for automated bottleneck and fault management (AP 3.1) as well as stable and systematic grid restoration (AP 3.2) in medium- and high-voltage (MS/HS) networks. These systems leverage machine learning techniques—ranging from supervised and unsupervised learning to reinforcement learning—with a hybrid training approach incorporating domain knowledge via Physics-Informed Neural Networks (PINNs) and Known-Operator Learning. This ensures reliable, explainable decision-making in real-time operations and minimizes data requirements in safety-critical environments.

In bottleneck management, predictive and event-triggered AI models autonomously suggest topology adjustments and flexibility deployment, factoring in a wide range of grid constraints and operational targets. Fault management employs multi-source data analysis for real-time fault detection, location, isolation, and resupply, utilizing both existing sensors and proposed novel measurements. For restoration, a dedicated assistance system is developed using a two-stage training process: supervised learning of standardized grid restoration procedures followed by reinforcement learning in a high-fidelity simulation environment. A single-agent model competes across simulation instances to evolve optimized strategies for power system recovery.

To support these applications, TP5 (AP 5.1) establishes a unified graph-based knowledge database for integrating heterogeneous data sources—from GIS, SCADA, and metering systems—into a cohesive representation. This structure enables complex system queries and machine-readable context for AI models, forming the backbone of advanced operational analytics.

The entire system architecture and algorithms are tested in AP 6.3 through a hybrid field test utilizing real-time simulation with OPAL-RT and live grid snapshots from Lechwerke’s control center. Assistance systems interface with the simulation via standard protocols, ensuring vendor-independent implementation and practical transferability. Experienced grid operators validate the system in realistic operational conditions, ensuring alignment with the requirements of future grid operations and energy transition goals.

CRIS →

AI Grid ProtectionRunningCoordinated grid protection based on machine learning methods (Netzschutz-KI)Jul 2024 – Jun 2027 · DFG, project no. 535389056Learning-based protection agents for fault detection, classification, faulted-line identification and localisation; reproducible benchmarks (PROTECT-90, EvEMTBench) and reinforcement learning for coordinated fault clearance.Multimodal data MPC & RL Critical infrastructure Julian Oelhaf

Julian

Full description (CRIS)

Project page → · CRIS →

HLA matchingIn preparationModelling the Human Leukocyte Antigen (HLA) system for donor–patient matchingwith University Medicine EssenNew project: understanding and modelling the HLA system and optimising the matching between living donors and patient populations.Medical

OptiHydCompletedAI-refined thermo-hydraulic model for the improvement of the efficiency and quality of water supplyNov 2023 – Oct 2026Hybrid AI models of the relationship between water and soil temperature, built from heterogeneous data of a Bavarian water supply network – the basis for leak detection, anomaly recognition and drinking-water quality monitoring.Trustworthy ML Multimodal data Critical infrastructure Mohammad Moataz Tolba

Mohammad, Siming

Full description (CRIS)

The United Nations' goals for sustainable development have made improving quality of life and access to clean drinking water a political priority. However, in recent decades, the water cycle in Bavaria has also been significantly affected by climate change. Two important aspects of daily drinking water supply and distribution are the assurance of water quality and the increase in usage efficiency. To enhance the resilience and capacity of the water supply in general, numerical simulation, data integration, and artificial intelligence (AI) are necessary. In this project, we aim to develop an AI-refined temperature-hydraulic model using heterogeneous data sources from a Bavarian water supply network. Hybrid AI methods are employed to model the complex relationship between water and soil temperature. The resulting model will serve as the basis for various real applications such as leak detection, anomaly recognition, and monitoring of drinking water quality, with the overarching goal of increasing the efficiency and quality of the water supply while simultaneously contributing to the containment of the impact of climate change on drinking water supply

CRIS →

UtilityTwinCompletedAn intelligent digital twin for utility networks with heterogeneous data and AI algorithmsSep 2021 – Aug 2024 · Bavarian State Ministry of Digital Affairs, IuK Bayern (DIK0325/01)Digital twin for energy and water supply networks combining high-resolution sensor data, big data and machine learning. 2nd prize in the BMFTR Student Prize for Social Innovations 2025.Multimodal data Critical infrastructure

Ady, Siming

Full description (CRIS)

In the UtilityTwin research project, an intelligent digital twin for any energy or water supply network is to be researched and developed on the basis of adaptive high-resolution sensor data (down to the sub-second range) and machine learning techniques. Overall, the technology concepts BigData and AI are to be combined in an innovative way in this research project in order to make positive contributions to the implementation of the energy transition and to counteract climate change.

Project page → · CRIS →

Collaboration partners

FAU institutes

Institute of Electrical Energy Systems (IEES), FAU – power-system expertise; joint work on neural power flow

Industry & utilities

Diehl Metering GmbH – industrial doctoral projects
LEW Distribution Grid GmbH (Lechwerke AG), Augsburg – GridAssist – field test with live grid snapshots
Water and heat utilities – e.g. the Bavarian water supplier in OptiHyd

Clinical partners

University Medicine Essen – HLA matching project

Universities & research institutes

Sichuan University – visiting researchers

Publications

2027

Journal Articles

Oelhaf J., Kordowich G., Bergler C., Maier A., Jäger J., Bayer S.:
Deep Learning Architectures for Fault Analysis in Power System Protection: A Reproducible Evaluation on Public Waveform Data
In: Electric Power Systems Research Volume 265, Part C (2027), Article No.: 114169
ISSN: 0378-7796
DOI: 10.1016/j.epsr.2026.114169
URL: https://www.sciencedirect.com/science/article/pii/S0378779626014574
BibTeX: Download

2026

Journal Articles

Kordowich G., Oelhaf J., Bayer S., Maier A., Kereit M., Jäger J.:
Feature Selection for Fault Prediction in Distribution Systems
In: Electric Power Systems Research 261 (2026), Article No.: 113498
ISSN: 0378-7796
DOI: 10.1016/j.epsr.2026.113498
BibTeX: Download
Oelhaf J., Kordowich G., Perez Toro PA., Bergler C., Jäger J., Maier A., Bayer S.:
A Standardized Framework for Machine Learning in Power System Protection
In: International Journal of Electrical Power & Energy Systems 181 (2026), Article No.: 112169
ISSN: 0142-0615
DOI: 10.1016/j.ijepes.2026.112169
URL: https://www.sciencedirect.com/science/article/pii/S0142061526006113
BibTeX: Download
Oelhaf J., Pashaei M., Kordowich G., Bergler C., Maier A., Jäger J., Bayer S.:
Robustness Evaluation of Machine Learning Models for Fault Classification and Localization in Power System Protection
In: IET Conference Proceedings 2026 (2026), p. 188-193
ISSN: 2732-4494
DOI: 10.1049/icp.2026.1934
BibTeX: Download
Chatterjee S., Ghumkar S., Ahbab MM., Ramachandran A., Tenbrinck D., Maier A., Semmelmann K., Bayer S.:
Early Detection of DMA-Level Leaks in Water Networks Using Robust Regression Ensemble Framework
In: Water 18 (2026), Article No.: 563
ISSN: 2073-4441
DOI: 10.3390/w18050563
BibTeX: Download
Ramachandran A., Chatterjee S., Neergaard TFB., Oberndoerfer M., Maier A., Bayer S.:
A deep learning framework for heat demand forecasting using time–frequency representations of decomposed features
In: Energy and AI (2026), Article No.: 100704
ISSN: 2666-5468
DOI: 10.1016/j.egyai.2026.100704
BibTeX: Download

Conference Contributions

Kim C., Conrad T., Karim R., Oelhaf J., Riebesel D., Arias Vergara T., Maier A., Jäger J., Bayer S.:
Physics-Informed GNN for Medium-High Voltage AC Power Flow With Edge-Aware Attention and Line Search Correction Operator
2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (Barcelona, Spain, May 3, 2026 - May 8, 2026)
In: 2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2026
DOI: 10.1109/ICASSP55912.2026.11463241
URL: https://ieeexplore.ieee.org/document/11463241
BibTeX: Download

Unpublished Publications

Oelhaf J., Kordowich G., Kim C., Andrea Perez-Toro P., Bergler C., Maier A., Jäger J., Bayer S.:
Controlled Comparison of Machine Learning Models for Fault Classification and Localization in Power System Protection (Conference contribution, accepted)
DOI: 10.48550/arXiv.2510.00831
BibTeX: Download
Oelhaf J., Pashaei M., Perez Toro PA., Kordowich G., Bergler C., Maier A., Jäger J., Bayer S.:
Fault Inception Detection in Real-World Disturbance Data for Power System Protection (Conference contribution, accepted)
IEEE PES ISGT EUROPE 2026 (Budapest, October 19, 2026 - October 22, 2026)
DOI: 10.48550/arXiv.2606.23111
BibTeX: Download
Oelhaf J., Luce A., Bergler C., Maier A., Bayer S.:
Offline Reinforcement Learning for Distribution-Grid Protection (Conference contribution, accepted)
IEEE Power & Energy Student Summit (PESS 2026) (Karlsruhe, October 21, 2026 - October 23, 2026)
DOI: 10.48550/arXiv.2609.24703
URL: https://arxiv.org/abs/2609.24703
BibTeX: Download
Chattopadhyay S., Oelhaf J., Schoening A., Deuschel J., Bhattacharyya B., Bergler C., Maier A., Bayer S.:
Peak-Aware Short-Term Load Forecasting Across Distribution Grid Aggregation Levels (Conference contribution, accepted)
IEEE PES ISGT Europe 2026 (Budapest, October 19, 2026 - October 22, 2026)
DOI: 10.48550/arXiv.2609.18588
URL: https://arxiv.org/abs/2609.18588
BibTeX: Download
Oelhaf J., Kordowich G., Bergler C., Maier A., Jäger J., Bayer S.:
PROTECT-90: A Fault Dataset for Power System Protection (Conference contribution, accepted)
IEEE PES ISGT EUROPE 2026 (Budapest, October 19, 2026 - October 22, 2026)
DOI: 10.48550/arXiv.2606.24298
BibTeX: Download

2025

Journal Articles

Oelhaf J., Kordowich G., Pashaei M., Bergler C., Maier A., Jäger J., Bayer S.:
A Scoping Review of Machine Learning Applications in Power System Protection and Disturbance Management
In: International Journal of Electrical Power & Energy Systems 172 (2025), Article No.: 111257
ISSN: 0142-0615
DOI: 10.1016/j.ijepes.2025.111257
BibTeX: Download

Conference Contributions

Oelhaf J., Kordowich G., Kim C., Perez Toro PA., Maier A., Jäger J., Bayer S.:
Impact of Data Sparsity on Machine Learning for Fault Detection in Power System Protection
33rd European Signal Processing Conference (EUSIPCO 2025) (Palermo, September 8, 2025 - September 12, 2025)
In: 2025 33rd European Signal Processing Conference (EUSIPCO) 2025
URL: https://ieeexplore.ieee.org/document/11226584
BibTeX: Download
Kordowich G., Oelhaf J., Maier A., Bayer S., Jäger J.:
A Graph Neural Network-Based Approach for Power System Protection
IEEE Kiel PowerTech (Kiel, June 29, 2025 - July 3, 2025)
In: PowerTech 2025, Kiel: 2025
DOI: 10.1109/PowerTech59965.2025.11180650
URL: https://ieeexplore.ieee.org/document/11180650
BibTeX: Download
Ramachandran A., Neergaard TFB., Arias Vergara T., Maier A., Bayer S.:
Water Demand Forecasting of District Metered Areas through Learned Consumer Representations
The 33rd European Signal Processing Conference (EUSIPCO 2025) (Palermo, Italy, September 8, 2025 - September 12, 2025)
In: Water Demand Forecasting of District Metered Areas through Learned Consumer Representations 2025
URL: https://eusipco2025.org/wp-content/uploads/pdfs/0001842.pdf
BibTeX: Download
Ramachandran A., Maier A., Bayer S.:
Digitization Framework of Water Utility Documents for Digital Twins
21st International Computing & Control in the Water Industry Conference (Sheffield, United Kingdom, September 1, 2025 - September 3, 2025)
In: Digitization Framework of Water Utility Documents for Digital Twins 2025
BibTeX: Download
Oelhaf J., Kordowich G., Maier A., Jäger J., Bayer S.:
Unsupervised Clustering for Fault Analysis in High-Voltage Power Systems Using Voltage and Current Signals
Fault and Disturbance Analysis Conference (Atlanta, GA, May 5, 2025 - May 6, 2025)
In: Fault and Disturbance Analysis Conference 2025 2025
DOI: 10.48550/arXiv.2505.17763
BibTeX: Download
Mederer C., Kordowich G., Oelhaf J., Maier A., Bayer S., Jäger J.:
Verification of Neural Network Based Power System Protection Schemes
19th IET Conference on Developments in Power System Protection (DPSP Europe 2025) (Bilbao, April 1, 2025 - April 3, 2025)
DOI: 10.1049/icp.2025.1062
BibTeX: Download
Oelhaf J., Kordowich G., Perez Toro PA., Arias Vergara T., Maier A., Jäger J., Bayer S.:
A Systematic Evaluation of Machine Learning Methods for Fault Detection and Line Identification in Electrical Power Grids
ICASSP 2025 - 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (Hyderabad, April 6, 2025 - April 11, 2025)
In: ICASSP 2025 - 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), New York City: 2025
DOI: 10.1109/ICASSP49660.2025.10890544
BibTeX: Download
Pahlajani KM., Ramachandran A., Maier A., Bayer S.:
Link Prediction on Water Distribution Networks using Graph Neural Networks
21st International Computing & Control in the Water Industry Conference (Sheffield, September 1, 2025 - September 3, 2025)
In: Link Prediction on Water Distribution Networks using Graph Neural Networks 2025
Open Access: https://orda.shef.ac.uk/articles/conference_contribution/Link_Prediction_on_Water_Distribution_Networks_using_Graph_Neural_Networks/29920967?backTo=/collections/CCWI_2025_Paper_Repository/7991153
BibTeX: Download

2024

Journal Articles

Stecher D., Neumayer M., Ramachandran A., Hort A., Maier A., Bücker D., Schmidt J.:
Creating a labeled district heating data set: From anomaly detection towards fault detection
In: Energy 313 (2024), Article No.: 134016
ISSN: 0360-5442
DOI: 10.1016/j.energy.2024.134016
BibTeX: Download
Ramachandran A., Mousa H., Maier A., Bayer S.:
Week-Ahead Water Demand Forecasting Using Convolutional Neural Network on Multi-Channel Wavelet Scalogram †
In: Engineering Proceedings 69 (2024), Article No.: 179
ISSN: 2673-4591
DOI: 10.3390/engproc2024069179
BibTeX: Download

Conference Contributions

Ramachandran A., Neergaard TF., Maier A., Bayer S.:
Advancing Heat Demand Forecasting with Attention Mechanisms: Opportunities and Challenges
NeurIPS 2024 The Thirty-Eighth Annual Conference on Neural Information Processing Systems (Vancouver, Canada, December 10, 2024 - December 15, 2024)
In: Advancing Heat Demand Forecasting with Attention Mechanisms: Opportunities and Challenges 2024
URL: https://www.climatechange.ai/papers/neurips2024/26
BibTeX: Download
Ramachandran A., Mousa H., Maier A., Bayer S.:
A Week Ahead Water Demand Forecasting using Convolutional Neural Network on Multi-Channel Wavelet Scalogram
WDSA CCWI 2024 - 3rd International Joint Conference on Water Distribution Systems Analysis & Computing and Control for the Water Industry (University of Ferrara, Ferrara, Italy, July 1, 2024 - July 5, 2024)
BibTeX: Download
Basak P., Ramachandran A., Maier A., Bayer S.:
Unveiling Consumer Behavior in District Heating Network: A Contrastive Learning Approach to Clustering
SESAAU2024 – Smart Energy Systems Conference (Aalborg, Denmark, September 10, 2024 - September 11, 2024)
BibTeX: Download

2022

Journal Articles

Moataz M., Reiter N., Al-Mqbas M., Metelkin A., Latifi A., Jovanović J., Delgado A.:
Statistical interpretation of LDA measurements in naturally developing turbulent drag-reducing flow using invariant theory
In: International Journal of Heat and Fluid Flow 93 (2022), Article No.: 108856
ISSN: 0142-727X
DOI: 10.1016/j.ijheatfluidflow.2021.108856
BibTeX: Download

Conference Contributions

Chatterjee S., Ramachandran A., Neergaard TF., Maier A., Bayer S.:
Heat Demand Forecasting with Multi-Resolutional Representation of Heterogeneous Temporal Ensemble
NeurIPS 2022 Workshop Tackling Climate Change with Machine Learning (Hybrid, December 9, 2022 - December 9, 2022)
In: NeurIPS 2022 Workshop: Tackling Climate Change with Machine Learning 2022
Open Access: https://www.climatechange.ai/papers/neurips2022/46
URL: https://www.climatechange.ai/papers/neurips2022/46
BibTeX: Download

Join us & colloquium

November

Friday

Core methods ↓Application domains →	Critical Infrastructure →Power · Energy · Water	Medical →Biomedical & population-level systems	Robotics →Learning-based robotic systems
Constrained & Trustworthy ML →	Changhun, Karan, MohammadGridAssistHyb_Mod_NetOptiHyd5 publications Details Changhun Kim (Doctoral researcher) Karan Pahlajani (Doctoral researcher) Mohammad Moataz Tolba (Doctoral researcher) GridAssist AI-based congestion management and grid restoration Hyb_Mod_Net Research and development of hybrid network models OptiHyd AI-refined thermo-hydraulic model for the improvement of the efficiency and quality of water supply	Redwanul Details Redwanul Karim (Researcher)	No joint work yet – open for collaboration
ML for Multimodal Data Fusion & Analysis →	Ady, Satyaki, JulianAI Grid ProtectionHyb_Mod_NetOptiHydUtilityTwin24 publications Details Adithya Ramachandran (Doctoral researcher) Satyaki Chatterjee (Doctoral researcher) Julian Oelhaf (Doctoral researcher) AI Grid Protection Coordinated grid protection based on machine learning methods (Netzschutz-KI) Hyb_Mod_Net Research and development of hybrid network models OptiHyd AI-refined thermo-hydraulic model for the improvement of the efficiency and quality of water supply UtilityTwin An intelligent digital twin for utility networks with heterogeneous data and AI algorithms	Lyuhe, Zengyu Details Lyuhe (Visiting researcher) Zengyu (Visiting researcher)	No joint work yet – open for collaboration
Model Predictive Control & Reinforcement Learning →	Julian, Changhun, KaranGridAssistAI Grid Protection1 publication Details Julian Oelhaf (Doctoral researcher) Changhun Kim (Doctoral researcher) Karan Pahlajani (Doctoral researcher) GridAssist AI-based congestion management and grid restoration AI Grid Protection Coordinated grid protection based on machine learning methods (Netzschutz-KI)	No joint work yet – open for collaboration	Prajol, Prabhu Details Prajol Shrestha (Research associate) Prabhu Nithin Gollapudi (Doctoral researcher)

Who we are

Group head

Researchers

Research: methods × domains

Core methods

Application domains

Highlights

Projects & funding

Collaboration partners

FAU institutes

Industry & utilities

Clinical partners

Universities & research institutes

Publications

All publications of the group (CRIS)

2027

Journal Articles

2026

Journal Articles

Conference Contributions

Unpublished Publications

2025

Journal Articles

Conference Contributions

2024

Journal Articles

Conference Contributions

2022

Journal Articles

Conference Contributions

Join us & colloquium

Open positions and theses

Colloquium

Contact