This research area is dedicated to fundamental research in the field of applied information technologies and serves as the foundation for ensuring the use of modern information technologies in the other research areas.
Through close collaboration with other research areas, researchers will collect and process large datasets, which they will then use to refine mathematical models, digital twins, forecasting, machine learning, and optimization. The data obtained will be immediately verified against the performance metrics of real equipment available in the laboratories of the center’s participants. However, the resulting methods and tools are planned to eventually become domain-independent, allowing them to be applied in other industries to support relevant processes.
Key approaches:
- Development of mathematical models describing the energy balance, generation planning, and resource allocation, taking into account operational and infrastructure constraints. The models are formulated as optimization problems (linear, nonlinear, and multi-criteria) concerning load distribution, capacity planning, and trade-offs between cost, reliability, and emissions.
- Methods for cleaning, validating, integrating, and performing contextual analysis of heterogeneous data streams from measurement infrastructure, control systems, and meteorological sources will be applied. This includes the processing of semi-structured data and time series, which aligns with our existing work on processing semi-structured text and monitoring data, as well as on the automated collection of information from open sources.
- Data warehouse creation, transformation, and analytics. Designing Extract, Transform, Load (ETL) pipelines and structured data warehouses for the long-term storage and analysis of energy time series (consumption, generation, grid status), as well as analytical dashboards for real-time monitoring.
- Machine learning models for process forecasting. Tuning, training, and deploying machine learning models (based on regression, ensemble methods, and neural network architectures) to forecast demand and renewable energy production, detect anomalies in grid behavior, and predict equipment wear and failures.