Satyaki Chatterjee, M.Sc.

Master’s Project – Physics-informed graph neural networks for urban flood simulation
Term: 13.02.2026 – ongoing

Urban flooding causes billions in damage each year and endangers lives, yet physics-based hydraulic simulators (e.g., MIKE FLOOD, SWMM) take hours per scenario—too slow for real-time planning. This project develops a graph neural network surrogate that approximates coupled 1D drainage and 2D surface flow simulations, enabling flood predictions in seconds.

Research questions:

• Generalization: Can the model predict water levels for unseen rainfall events, including intensities between trained scenarios?
• Mass conservation: Does enforcing nodal conservation improve accuracy and reduce autoregressive error versus global conservation?
• 1D–2D coupling: Does explicitly modeling underground drainage nodes improve predictions, especially when conservation is enforced at coupling interfaces?

For more details:
Project page

MA Thesis – Discovering seasonal behavioral structure in water demand through consumer clustering
Term: 01.05.2026 – ongoing

This thesis develops AI methods to cluster and track residential water consumption patterns using IoT smart-meter data, enabling interpretable behavioral segmentation for water utilities.

• Method development and benchmarking for robust consumer segmentation
• Temporal tracking and transition detection of behavioral changes
• Interpretability and stability analysis of clusters

Applications include adaptive tariffs and targeted demand-management strategies.
For more details:
Thesis page

MA Thesis – H2OArmor: A dynamic data-driven leak detection framework for varied digital maturity levels in water utilities
Term: 08.01.2025 – 08.07.2025

This thesis develops H2OArmor, a scalable machine-learning pipeline for DMA-level leak detection using heterogeneous data sources.

• Framework development: Flexible ML pipeline adaptable to different utilities
• Model integration: Ensemble-based leak detection methods
• Automated onboarding: Scalable integration of new methods
• Confidence scoring: Unified leak-detection decision score

For more details:
Thesis page

Students interested in a thesis or project are requested to apply via e-mail with their CV, grades, and relevant projects/GitHub. For an industrial thesis, please include a detailed research proposal.

Since 03/2025:
Ph.D. candidate at the Pattern Recognition Lab, Diehl Metering GmbH

10/2019 – 12/2022:
M.Sc. in Elite Master Programme of Advanced Signal Processing and Communication Engineering,
Friedrich-Alexander University Erlangen-Nürnberg

• Chatterjee, S.; Ghumkar, S.; Ahbab, M. M.; Ramachandran, A.; Tenbrinck, D.; Maier, A.; Semmelmann, K.; Bayer, S. (2026).
  Early detection of DMA-level leaks in water networks using robust regression ensemble framework.
  Water, 18, 563.
  https://doi.org/10.3390/w18050563
• Ramachandran, A.; Chatterjee, S.; Neergaard, T. F. B.; Oberndoerfer, M.; Maier, A. K.; Bayer, S. (2026).
  A deep learning framework for heat demand forecasting using time–frequency representations of decomposed features.
  Energy and AI, 24, 100704.
  https://doi.org/10.1016/j.egyai.2026.100704
• Souza Oliveira, D.; Ponfick, M.; Braun, D. I.; Osswald, M.; Sierotowicz, M.; Chatterjee, S.; Weber, D.; Eskofier, B.; Castellini, C.; Farina, D.; Kinfe, T. M.; Del Vecchio, A. (2024).
  A direct spinal cord–computer interface enables the control of the paralysed hand in spinal cord injury.
  Brain, 147(10), 3583–3595.
  https://doi.org/10.1093/brain/awae088
• Chatterjee, S.; Ramachandran, A.; Neergaard, T. F.; Maier, A. K.; Bayer, S. (2022).
  Heat demand forecasting with multi-resolutional representation of heterogeneous temporal ensemble.
  NeurIPS 2022 Workshop on Tackling Climate Change with Machine Learning.
  https://www.climatechange.ai/papers/neurips2022/46
• Chatterjee, S.; Bayer, S.; Maier, A. K. (2021).
  Prediction of household-level heat consumption using PSO enhanced SVR model.
  NeurIPS 2021 Workshop on Tackling Climate Change with Machine Learning.
  https://www.climatechange.ai/papers/neurips2021/42
• Choudhury, S.; Bandyopadhyay, S.; Chatterjee, S.; Dutta, R.; Dutta, S. (2017).
  A hybrid approach to enhance the security of automated teller machine.
  Proceedings of the International Conference on Communication and Networks.
  Advances in Intelligent Systems and Computing, vol. 508. Springer, Singapore.
  https://doi.org/10.1007/978-981-10-2750-5_71

AquaPositive: Data-Driven Strategy Recommendation of Water End-usage

Term: March 1, 2025 – March 31, 2028

Description: Water scarcity is a pressing global issue, exacerbated by factors such as climate change, population growth, and unsustainable water consumption patterns. To mitigate this crisis, it is imperative to introduce effective and sustainable water management strategies that extend to the very end of the consumption chain: the end consumers themselves. The AquaPositive project seeks to develop an intelligent and informed system that empowers end consumers to play a pivotal role in alleviating water scarcity by modifying their consumption behaviors. This approach is anchored in the understanding that informed behavioral changes can significantly contribute to conserving water resources. Educating end consumers about the significance of water conservation and its direct impact on mitigating water scarcity is fundamental. Informed consumers are more likely to adopt responsible water usage practices. However, Analyzing historical consumption patterns and behavioral deviations among different consumer groups provides valuable insights. Understanding when and how end-consumers tend to use excessive water allows for targeted intervention strategies. Tailoring recommendations to individual consumers based on their consumption behavior profiles empowers them to make conscious choices. These recommendations can include insights on optimal usage times and areas where water can be saved without compromising their needs. Therefore, there are two outcomes that the research project would try to address:
1. When and where, how much water is needed to maintain sustainable operations of end consumer groups
2. When and where, how much water can be saved by employing different strategies
The information as an outcome from these two questions would yield an informed strategy as ‘Action Recommendation’ for water end-consumers. The Aqua Positive project aims to develop an AI-driven recommendation system to guide end consumers in efficient water consumption, considering historical consumption, consumption patterns, population growth, climate change, and weather fluctuations.

Data Processing for Utility Infrastructure (DPUI)
Time Series Intelligence (TSI)