enlite.ai FLEX topology optimisation

von enlite.ai

AI Application

Niedrige Vertrauensstufe

enlite.ai (Vienna) develops FLEX, a reinforcement-learning and planning stack that proposes grid-topology changes (busbar splitting, substation reconfiguration) as a near-free alternative to redispatch or curtailment. In a joint paper with TenneT presented at ACM e-Energy 2025, an RL agent was trained on TenneT's historical days for a section of the Dutch transmission grid with more than 1,000 nodes, then a planning phase produced a Pareto set of 24-hour day-ahead switching plans in four to seven minutes, from which the operator chooses the trade-off. On TenneT data it solved 31 of 31 in-distribution days versus 24 for the best expert strategy, and 137 of 181 out-of-distribution days versus 123. enlite.ai states plainly that this was validated retrospectively and is not running in live control. A TenneT lead data scientist co-authored the work.