Normalizing Flows-based Probabilistic Learning-Aided Distribution System State Estimation

This paper proposes a new distribution system state estimation (DSSE) solver constructed based on a strong generative network, the normalizing flow (NF). The proposed NF-based DSSE solver provides full statistical information on the state of distribution networks through the probability density function (PDF), ensuring its robustness on Gaussian noises and effectiveness for different small-scale data sizes. The proposed solver is validated by the IEEE 123-bus test system and real-world data from London, England, and are compared with three existing DSSE solvers in the literature, showing the superior performance. The proposed DSSE framework demonstrates a quantifiable enhancement in robustness against Gaussian noises, achieving a 30% improvement in accuracy over previously presented DSSE solvers.