AMI Data-Driven Strategy for Hierarchical Estimation of Distribution Line Impedances

Abstract

Monitoring, operation, and protection of distribution power grids fundamentally rely on the accurate estimation of line impedances. However, line impedance estimation is challenging due to the difficulties in modeling the dependence on line temperature and aging. This paper proposes a new data-driven strategy to estimate low-voltage (LV) and medium-voltage (MV) line impedances using an advanced metering infrastructure (AMI). In the proposed strategy, two-level optimization problems are formulated using generalized equations for voltage drops along LV and MV lines and then extended based on AMI data collected over time. Hierarchical estimation is achieved using the local and global references to the LV and MV root buses, respectively, enabling parallel estimation for individual LV grids and thus reduced computation time. Reinforcement learning is also integrated to compensate for possible measurement errors in the AMI data, ensuring robust estimation of LV and MV line impedances. The proposed strategy is tested on a three-phase unbalanced MV grid with multiple single-phase LV grids under various conditions characterized by measurement samples and errors. The results of case studies and sensitivity analyses confirm that the proposed strategy improves the accuracy and robustness of line impedance estimation at both LV and MV levels.