A Distributionally Robust Control Strategy for Frequency Safety based on Koopman Operator Described System Model

As renewable energy and power electronics become more prevalent in power systems, modeling system frequency dynamics under power deficits presents increasing complexity. While data-driven approaches can alleviate some of these difficulties, they are exposed to noise and training inaccuracies, which introduce uncertainty in modelling and prediction. To tackle the uncertainty and limited statistical knowledge of prediction errors, we propose a distributionally robust, data-driven emergency frequency control (DREFC) framework. The framework focuses on achieving a high probability of frequency stability while providing flexibility for decision-makers to adjust control conservativeness. Specifically, DREFC formulates a min-max optimization problem to determine an optimal control strategy that is robust to variations in prediction errors, modeled within a Wasserstein-distance-based ambiguity set. By analytically approximating the Value at Risk (VaR) constraints, we achieve a computationally efficient reformulations. Simulation results show that DREFC effectively guarantees frequency stability, minimizes control costs and low computation time.