Power flow calculations play a critical role in ensuring the reliability and efficiency of power grids. However, classical methods used for AC power flow analysis often lack robustness and explainability when faced with power grid data inconsistencies, which frequently occur in operational contexts (e.g., long term AC planning).

We present a nonlinear-optimization-based approach, robust to data inconsistencies and identifying them through a penalization of the standard power flow equations. This mechanism enables both the feasibility of the equations system and the identification of inconsistencies, making the method effective regardless of the presence of data errors.

This method has been published as open-source code within the Power System Blocks (PowSyBl) library, part of the LF Energy Foundation, and dedicated to electrical grid modeling, visualization and simulation. It leverages an external nonlinear optimization solver, recently interfaced with the OpenLoadFlow module, to perform the computation. The reliability of the approach is validated on both academic and real-world networks, deliberately perturbed to introduce data errors that cause traditional approaches to fail.