Performance Assessment of Linearized OPF-based Distributed Real-time Predictive Control

We consider the problem of controlling heteroge-neous controllable resources of a distribution network with the objective of achieving a certain power flow at the grid connection point while respecting local grid constraints. The problem is formulated as a model predictive control (MPC), where a linearized grid model, to retain convexity, based on sensitivity coefficients (SCs) is used to model the grid constraints. We consider and compare the modelling performance of three different update policies for the SCs: when they are updated once per day considering static injections, updated once per day considering point prediction of the nodal injections, and recursively estimated using on-line measurements. Simulations are performed considering the CIGRÉ low voltage benchmark network. Performance is evaluated in terms of convergence speed, tracking error, and constraints modeling errors. Further, we perform a sensitivity analysis on the dominant model w.r.t. the length of the predictive horizon and number of controllable units.