https://www.mdu.se/

The increasing integration of renewable energy sources into power systems has introduced greater variability and uncertainty in electricity generation. Battery energy storage systems (BESS) have emerged as a critical solution, enabling energy balancing and providing ancillary services to maintain grid stability. However, participating effectively in electricity markets requires accurate demand forecasting and intelligent battery scheduling. This thesis proposes a data-driven framework that integrates ensemble-based deep learning forecasting with optimization modeling to improve battery operation and participation in electricity markets. Electricity consumption is forecasted using a robust ensemble model that combines LSTM, CNN, and GRU architectures, trained on high-resolution, minute-level data from an industrial site. These forecasts serve as inputs to a Mixed-Integer Linear Programming (MILP) model designed to schedule battery operations for optimal engagement in the Frequency Containment Reserve for Normal operation (FCR-N) market. The optimization model incorporates technical battery constraints, fuse limitations, and regulatory requirements to ensure operational feasibility and to maximize economic returns. The framework is validated using real-world data provided by CheckWatt AB, illustrating its ability to enhance bidding strategies and improve revenue generation. The results demonstrate the strength of combining ensemble-based forecasting with mathematical optimization in enabling intelligent, market-responsive battery control. Overall, the proposed framework provides valuable insights into optimal battery scheduling strategies for energy aggregators and battery operators participating in dynamic electricity markets.