Performance Comparison of Alternating Direction Optimization Methods for Linear-OPF based Real-time Predictive Control

The paper contributes to improving the computational performance of controls of distributed energy resources (DERs) in distribution grids for efficient real-time control and short-term scheduling. The considered setting is a distribution grid with heterogeneous DERs controlled with model predictive control (MPC) to track a dispatch plan at its grid connection point (GCP) subject to DERs’ and grid’s constraints. The MPC control is first expressed as a quadratic programming (using linearized grid models) and, then, solved with several state-of-the-art alternating direction optimization methods: AMA, ADMM, AADMM (ADMM with adaptive penalty parameter) and their accelerated variants: FAMA, FAADMM, and FAADMM with restart rule. Performance is tested in terms of computational performance, constraints satisfaction, and optimality against the centralized MPC. The case studies are the CIGRE and IEEE benchmark grid for low- and medium voltage systems hosting different numbers of controllable DERs.