As physical systems become increasingly complex and data-rich, intelligent, data-driven tools are essential for making informed decisions in real time. Applications such as power grids, supply chains, transportation networks, and environmental monitoring all involve interconnected entities that evolve over time and space. Accurate forecasting and decision support across such systems can drive efficiency, reduce risk, and optimize operations.

This project is conducted in collaboration with an industrial partner and focuses on applying Artificial Intelligence (AI) methods to model and forecast spatio-temporal patterns in graph-structured physical systems. Students will work with high-resolution datasets from domains such as energy systems, including grid topology, hourly production and load time series, and renewable availability profiles. A central goal is to build a decision-support model that learns from historical data and can simulate or predict system stress under dynamic conditions.

You will explore architectures that combine time series modeling and graph neural networks (GNNs), investigate temporal embeddings (e.g. seasonal/cycle time encodings), and benchmark transformer-based and GNN-based models on forecasting and intervention tasks. Students will collaborate on model experimentation, algorithmic development, and building a prototype for practical demonstration.

Possible Tasks You may work on one or more of the following topics:

• Forecasting congestion, curtailment risk, or performance bottlenecks in physical networks
• Integrating temporal attention and graph architectures (e.g. ST-GNNs, TGCN, ST-Transformer)
• Investigating cycle embeddings and temporal encoding strategies
• Creating a decision-support dashboard to simulate interventions and outcomes

Required Skills

• Proficiency in Python
• Familiarity with data and ML tools such as Pandas, NumPy, PyTorch, Scikit-learn, PyG/DGL, TensorFlow, or similar
• Foundational understanding of time series modeling or graph-based learning is a plus

After successful completion of the project there is a possibility for an internship with the partner.