A Data-Driven State Estimation Based on Sample Migration for Low-Observable Distribution Networks

This paper proposes a data-driven state estimation based on sample migration for low-observable distribution networks, addressing the challenge of traditional state estimators being unsuitable for distribution networks with low observability. The state estimation model is trained using historical measurement data from distribution networks with high observability. Measurements updated for low-observable distribution networks are supplemented by transferring samples from high-observable distribution networks using sample migration techniques, resulting in a state estimation model suitable for low-observable distribution networks. Test results demonstrate that the proposed algorithm outperforms traditional algorithms in both estimation accuracy and robustness aspects, such as the Weighted Least Squares (WLS) and Weighted Least Absolute Value (WLAV) methods. Furthermore, sample migration enhances the generalization ability of the state estimation model.