AI for Energy Systems
In the field of AI for energy systems, we are researching data-driven methods for forecasting, optimization, and control in decentralized, dynamic energy systems. Our focus is on self-learning control strategies (e.g., reinforcement learning), time series forecasting, and hybrid approaches that combine physical system knowledge with machine learning. By integrating AI with simulation environments and traditional optimization methods, strategies can be safely developed, tested, and made robust for practical application.
- Self-learning control of PV-battery systems, heat pumps, and flexibility solutions
- Smart charging and fleet management for electric mobility
- Load and generation forecasts for buildings, neighborhoods, and energy communities
- Optimization of storage utilization, self-consumption, and grid services
- Anomaly detection and condition diagnosis based on measurement and operational data
- Decision intelligence algorithms for control and planning tasks using reinforcement learning, MPC, and stochastic optimization
- Time series forecasts (load, generation, prices) and probabilistic predictions
- ML approaches for digital twins to model complex systems
- Offline/online evaluation: simulation-based benchmarks and A/B testing
- MLOps for energy systems: data pipelines, experiment tracking, and model monitoring
- Scalable computing infrastructure (HPC/cloud) for the training, evaluation, and production-ready operation of AI models, as well as for extensive data analysis
- Practical simulation and test environments for the safe development, validation, and verification of AI-based control and optimization strategies prior to field deployment
- Data and analytics infrastructure for the collection, processing, and evaluation of time-series, operational, and measurement data
- Development and benchmarking of AI-based control and planning algorithms (e.g., smart charging, storage operation)
- Load and generation forecasts and optimization models for operations and planning
- Data strategy, data quality, and development of ML pipelines for energy data
- Prototyping and pilot testing in simulation environments prior to field tests