Robustness assessment of primal–dual gradient projection-based online feedback optimization for real-time distribution grid management

The increasing deployment of distributed energy resources causes voltage and congestion issues in distribution grids. Recently, online feedback optimization (OFO) emerges as a promising real-time solution approach. OFO uses measurements as feedback and employs optimization algorithms as feedback controllers to steer the distribution system towards optimal operating points. OFO does not need an accurate grid model nor consumption data of non-controllable loads and affords fast implementation, which make it particularly suitable for real-time distribution grid management. This paper aims to provide an extensive robustness assessment of OFO based on the primal–dual gradient projection (PDGP) algorithm under practical distribution grid operational conditions. To quantify system performance, we use metrics including active power curtailment ratio, voltage and loading constraint violations, normalized reference power tracking error, and distance to the deterministic-case trajectory. Simulations conducted on a 136-bus medium-voltage grid using second-scale data reveal that the algorithm demonstrates satisfactory robustness to time-varying generation and loads, grid model inaccuracy, measurement errors, and communication failures, but is susceptible to systematic communication delays and unnoticed topology changes particularly involving tripping of cables at the beginning of distribution feeders. Potential solutions to these shortcomings are discussed. • An extensive robustness assessment of online feedback optimization is provided. • Four metrics are developed to quantify system performance under uncertainties. • Congestion management and balancing services provision are considered. • Simulations confirm and challenge existing findings and provide new insights. • Potential solutions to shortcomings of the approach are discussed.