Towards Automatic Parameterization of Online Feedback Optimization for Autonomous Low-Voltage Distribution Grids

The increasing integration of distributed energy resources in Low-Voltage (LV) distribution grids necessitates the implementation of active network management to ensure reliable grid operation. In previous work, we proposed a combination of data-driven state estimation and Online Feedback Optimization (OFO) as a solution for autonomous LV grid management. However, this approach still requires manual parameter tuning of the OFO algorithm during the commissioning phase. In this paper, we propose a three-step and model-free parameterization method for a projected-gradient-based realization of the OFO algorithm. First, we estimate OFO’s sensitivity matrix by applying least square estimation on historical smart meter measurements. Subsequently, we estimate OFO’s step size based on the estimated sensitivity matrix and a closed-form formula derived from a small-signal stability analysis. The third step comprises an adaptive step sizing procedure that enables the OFO algorithm to adapt to changing grid conditions. Our proposed model-free parameterization method thus automates the commissioning of OFO in LV grids, which is demonstrated through simulation experiments on a benchmark LV grid.