Model-Free Robust Online Feedback Optimization for Voltage Regulation in Distribution Grids

Voltage regulation in distribution grids often suffers uncertainties such as model mismatches and parameter changes due to the installation of renewable energy sources. This article proposes a model-free robust online feedback optimization (OFO) framework that addresses these challenges. By reformulating the classical optimization problem into an incremental input form under box constraints, an OFO method that solely relies on measurement data is developed. To handle unknown input–output constraint relationships, an adaptive forgetting factor least-squares algorithm for gradient estimation is adopted, which can swiftly track the gradient information. Sound theoretical proof is laid out to show the convergence property of the proposed method. Simulations using a modified IEEE 22-bus system and real distribution grid data validate the proposed method’s effectiveness and robustness.