Ai In Swissgrid's Control Room: Automated Balancing Energy Activation - From Vision To Operational Reality

A primary responsibility of Transmission System Operators (TSOs) is frequency control, achieved by maintaining generation-load balance within the power system. Swissgrid, the Swiss TSO, traditionally managed balancing energy activation manually. However, this approach faced diminishing economic efficiency, driven by escalating imbalance situations and increasing balancing energy costs. The multi-stage activation process spanning European and national balancing markets creates complex sequential decision-making problems. In response, Swissgrid developed and operationalized a machine learning-based automation solution, becoming one of the first TSOs worldwide to implement such automation and the first with a cost-optimized approach. This paper presents a stacking ensemble forecasting method that dynamically leverages multiple model strengths based on operating conditions. Additionally, we describe our approach to optimize activation decisions across European and national balancing markets. Our automation achieved a 27% decrease in imbalance forecast error compared to system operators and C30-50 million average annual cost reduction.