PACR method boosts AC power flow restoration by 80% cost improvement

Power grid operators rely on DC optimal power flow (DCOPF) for its speed and scalability, but DC dispatches often violate nonlinear AC power flow equations and operational limits. A new paper from Michael A. Boateng and colleagues presents PACR (Parameter-Optimized AC Power Flow Restoration), a method that bridges this gap. PACR uses differentiable surrogates for distributed slack (active-power balancing) and PV/PQ switching (reactive-power regulation), each with trainable parameters like participation factors, voltage setpoints, and steepness. These parameters are learned offline by differentiating through the AC restoration equations using the implicit function theorem, then fixed for inference. The result: a lightweight restoration step that converts any DCOPF dispatch into an AC-feasible operating point.

PACR was evaluated on IEEE, ACTIVSg, and PEGASE test systems, comparing its performance against conventional single-slack AC recovery and full ACOPF. On the largest case (9,241 buses), PACR improved the cost difference by 80% over the conventional baseline and reduced solving time by 75% relative to ACOPF. This opens the door to using fast DCOPF in real-time operations without sacrificing AC feasibility, especially for large-scale grids. The method is architecture-agnostic and could be integrated into existing EMS software. Future work may explore online adaptation and extension to contingency analysis.