ADMM Performance Enhancement Techniques for DC Optimal Power Flow Applications

This paper presents two novel techniques to improve the convergence performance of distributed DC optimal power flow (DC OPF) when solved using the alternating direction method of multipliers (ADMM). In this framework, the OPF problem is decomposed into multiple subproblems, each representing a portion of the overall power grid. While most existing approaches penalize only the voltage angles of boundary buses in the ADMM Lagrangian function, we propose augmenting the Lagrangian penalty by incorporating a surrogate for tie-line power flow. Additionally, we reformulate the nodal power balance constraints at boundary buses to reflect each subproblem's willingness to exchange power through tie-lines with neighboring subproblems. These two techniques are applied to both sequential and parallel variants of ADMM. Case studies on various power systems demonstrate that, while each technique individually reduces the number of ADMM iterations, their combination yields the most significant enhancement in convergence for both sequential and parallel ADMM.